Abstract

The rapid evolution of industry has driven a demand for autonomous intralogistics solutions; however, existing industrial systems (such as Kiva or Swisslog) remain prohibitively expensive and complex for smaller-scale applications. This project presents the design and prototyping of a cost-effective Load Transport Robot capable of autonomous navigation and goods retrieval on a grid-based matrix map.

The robot operates on a "fetch-and-return" logic, utilizing an Arduino Mega 2560 microcontroller and QTR-8 infrared sensors for line following and coordinate tracking. A key innovation of this design is the implementation of a pyramid-shaped docking mechanism. Unlike traditional systems that require precise sensor alignment, our solution utilizes a vertical linear actuator with a tapered pyramid head. This geometry provides passive mechanical alignment: as the actuator rises into the socket of a mobile carrier, the pyramid shape physically forces the wheeled platform to self-center, effectively compensating for navigation errors without the need for complex feedback loops.

This report details the mechanical design, electronic architecture, and control algorithms of the prototype. It further provides a comparative market analysis and a legal Freedom-to-Operate (FTO) assessment, demonstrating that the proposed fixed-chassis and mechanical docking approach avoids infringing on major industrial patents. The final prototype successfully validates the feasibility of using low-cost, coordinate-based navigation for autonomous shelf transport.

Keywords: Autonomous Mobile Robot (AMR), Intralogistics, Matrix Navigation, Passive Mechanical Alignment, Arduino Mega, Linear Actuator.

2.1 Problem Definition & Story

Modern distribution warehouses are increasingly facing complex logistics challenges that make traditional manual operations inefficient, costly, and unsafe. Warehouses are no longer capable of handling growing orders and maintaining a high level of accuracy in the era of e-commerce evolution. Hence, many common problems persist that reflect the limitations of human-dependent workflows [1, 2].

One of the most significant issues reported in the warehousing industry is slow manual pick up and internal transport, which directly reduce operational efficiency. Workers often walk long distances to retrieve items from storage locations and transport them to packing or staging areas, creating bottlenecks and increasing order cycle times. As illustrated in Figure 2.1, internal transport and picking tasks are highly dependent on human labor. These repetitive and labor-intensive activities contribute to lower throughput and higher operational costs compared to automated solutions [3].

In addition to efficiency limitations, manual warehouse operations are often associated with work-related injuries. Tasks such as lifting, bending and transporting loads expose workers to musculoskeletal disorders and physical fatigue. The transportation and warehousing sector reported higher injury rates above national averages in many regions due to the intensity of manual handling activities [3, 6].

These operational and safety challenges are further aggravated by inefficient space utilization, picking errors, and limited traceability of goods. Poorly optimized warehouse layouts and human-dependent workflows make it difficult to streamline the storage and order preparation processes, resulting in wasted time, reduced productivity, and increased costs [6].

In response to these constraints, automated warehouse robots are highly needed in upgrading the warehouse logistics system. By taking over repetitive transport and handling tasks, robotic solutions can operate with higher speed, precision, and reliability than manual labor, while simultaneously reducing physical strain on workers and improving overall warehouse performance [2, 7].

2.2. Quantification of the Need

To justify the relevance of a warehouse robot solution, it is necessary to rely on measurable data that highlight the impact of automation on warehouse performance, costs, and safety.

2.2.1. Efficiency and Productivity

Facilities that have implemented warehouse robotics report significant improvements in operational performance. Studies indicate that operational efficiency typically increases by approximately 25–30% within the first year of automation, with some warehouses achieving productivity gains of up to 50%, mainly due to reduced manual transport time and optimized internal logistics flows [7].

In addition, automated and highly digitalized warehouses are 76% more likely to achieve high inventory accuracy levels (around 99% or above) and 36% more likely to reduce annual labor costs as a result of decreased dependence on manual picking and transport activities. The reported impact of warehouse automation on productivity, accuracy, and labor costs is summarized in Figure 2.2 [8].

2.2.2. Labor Cost and Workforce Issues

Labor costs remain one of the most critical challenges in traditional warehouse operations. Manual labor typically represents 50–70% of total warehouse operating expenses, making warehouses highly sensitive to workforce availability and turnover [9]. Recruiting and training a single warehouse employee can cost between $4,500 and $6,000, while high turnover rates generate continuous recruitment and training expenses. These factors increase financial burdens on distribution centers and motivate the adoption of automation solutions that reduce labor dependency [9].

2.2.3. Safety and Injury Reduction

From a safety perspective, increased levels of warehouse automation correlate with an approximate 25% reduction in reported workplace injuries, mainly due to the reduction of manual lifting, repetitive motions, and long-distance transport tasks [10].Despite these improvements, the warehousing sector continues to exhibit above-average injury incidence rates, often exceeding 4.5 cases per 100 workers, which highlights the persistent need for safer task execution methods and ergonomic improvements [5].

2.2.4. Synthesis

These quantitative indicators clearly demonstrate the motivation for warehouse automation solutions:

• increased throughput and operational efficiency,

• reduced labor and operating costs,

• improved accuracy and error reduction,

• enhanced worker safety and ergonomics.

2.3. Target Personas

A clear distinction is necessary between the users, who interact with the robot on a daily basis, and the buyers, who make strategic and financial decisions regarding its deployment.

1. User Persona — Warehouse Operators

Warehouse operators and pickers are the primary users of the proposed robotic system. They are responsible for preparing orders and performing internal transport tasks within the warehouse environment. Their daily activities typically involve walking long distances, manually lifting and transporting goods, and carrying out repetitive operations such as scanning and inventory verification [3, 5].

These tasks expose operators to several challenges. Continuous physical effort leads to fatigue and increased risk of musculoskeletal injuries, while time pressure to meet throughput targets can negatively impact accuracy and job satisfaction which directly reduce the overall productivity[5].

As a result, warehouse operators require tools that reduce physical strain, increase task accuracy, and streamline daily operations, allowing them to focus on higher-value activities such as quality control, exception handling, and coordination tasks. A robotic transport assistant directly addresses these needs by taking over the most physically demanding and monotonous transport operations, thereby improving working conditions and operational efficiency [6, 7].

1. Buyer Persona — Warehouse Management and Distribution Center Owners

The buyers of the proposed system are warehouse managers, operational directors, and owners of distribution centers responsible for optimizing logistics performance and controlling operational costs. Unlike daily users, buyers evaluate robotic solutions from a strategic and economic perspectives as well as an opportunity for scalability.

Key decision criteria include:

• Return on Investment (ROI): The ability of the system to reduce labor costs, minimize errors, and improve long-term profitability [2, 9].

• Operational Efficiency: Improvements in throughput, order cycle times, and inventory accuracy [7, 8].

• Safety and Compliance: Reduction of workplace injuries, ergonomic risks, and associated liabilities [5, 10].

• Scalability and Flexibility: The capacity of the system to adapt to changing warehouse layouts, demand peaks, and future expansion without major infrastructure modifications [2].

Buyers are therefore not direct users of the robot, but decision-makers who must justify investment based on quantifiable gains in productivity, cost reduction, and safety performance. Addressing these criteria is essential for the successful adoption of any warehouse automation solution.

2.4 Why This Project Is Relevant

In summary, the motivation for this project arises from well-documented inefficiencies and safety risks present in current warehouse operations. Distribution warehouses are under increasing pressure to maintain high throughput, accuracy, and reliability while still relying heavily on manual labor for internal transport and handling tasks [1, 2]. Manual picking, material handling, and internal transport activities contribute significantly to high labor costs, elevated injury rates, and operational delays, directly impacting overall warehouse performance [3, 5, 9].

Automation in the form of warehouse transport robots offers a data-backed pathway to improvement. By reducing dependence on manual transport tasks, robotic solutions will enable higher productivity, lower operational costs, and improved working conditions. Studies have shown that warehouse automation can lead to measurable gains in efficiency, accuracy, and injury reduction, making it a viable and increasingly necessary solution for modern distribution centers [2, 7, 10].

By considering both the user perspective, represented by warehouse operators seeking safer and less physically demanding working conditions, and the buyer perspective, represented by warehouse management aiming for cost-effective and scalable efficiency improvements, this project addresses a real industrial problem with quantifiable benefits. As a result, the proposed warehouse transport robot constitutes a relevant, timely, and justified solution within the context of contemporary logistics and electromechanical engineering [2, 6].

3. Project Working Modes, Functionality and Requirements

3.1 Operating Context and Scope of the System

The proposed system is a load transport robot intended for indoor distribution warehouses, operating in structured environments such as logistics centers, e-commerce warehouses, and industrial storage facilities. The robot is designed to perform autonomous transport of packages along predefined paths, supporting order preparation and internal logistics operations.

The system operates on flat, controlled warehouse floors and is intended to coexist with human operators and other automated systems. Navigation relies on structured guidance methods, such as line following, which are well adapted to industrial environments where infrastructure can be prepared in advance.

3.2. Identification and Formulation of the Essential Technical Problems

The warehouse transport robot is made up of three interacting subsystems: navigation, lifting platform, and trajectory tracking. Each subsystem introduces specific technical challenges that must be addressed to ensure safe, precise, and autonomous package transport within the warehouse environment.

1. Navigation System

The navigation system plays a central role in identifying the exact position of the package on warehouse grid. As the robot needs to stop beneath the package and attach to the mobile platform. One of the main technical challenges is achieving sufficiently precise docking so that the robot can align with the package and establish attachment without lifting its entire weight. Even small positioning errors may prevent successful docking or lead to unstable transport conditions.

In addition, the robot must be able to verify that a secure attachment has been established before initiating motion. Without reliable attachment verification, there is a risk that the robot could start moving while the package remains partially or fully detached, compromising safety and system reliability.

These challenges directly affect the robot’s ability to safely pick up and transport packages.

1. Lifting Platform and Traction Force Transmission

Beyond its lifting function, the lifting platform acts as the mechanical interface through which traction forces are transmitted from the robot to the mobile platform carrying the package. The main technical challenge is ensuring reliable pulling of the load during motion without introducing excessive mechanical stress or overloading the drive motors.

This challenge becomes more critical during the changing of direction. When the robot negotiates corners, lateral forces and additional torque are applied at the robot-load interface. If these forces are too high, they can reduce traction efficiency, increase motor effort, or compromise the stability of the attachment.

Several system-level constraints must therefore be considered. The geometric compatibility between the lifting platform and the package interface affects docking accuracy and force transmission under motion uncertainties. In addition, the traction and actuation capability of the mobile platform must be sufficient to pull the combined robot-load system while maintaining stable motion and acceptable motor loading ensuring there is no over sizing of the motors to prevent fast drainage of the batteries.

Finally, the trajectory layout imposed by the navigation map directly influences the forces acting on the system. Sharp turns and abrupt direction changes increase lateral loads and torque demands, which must remain compatible with the mechanical and actuation limits of the system.

The interaction between the lifting interface, traction capabilities, and trajectory constraints is therefore essential to ensure reliable transport and long-term system durability.

1. Trajectory-Tracking Method

The trajectory-tracking subsystem guides the robot along predefined paths within the warehouse. An essential technical problem was ensuring reliable sensor measurements, as infrared sensors may produce false readings due to surface irregularities, lighting variations, or noise.

To address this issue, the system must be capable of detecting tracking errors and recovering from them during operation. This includes the ability to recalibrate or return to a known reference position when position information becomes unreliable.

Reliable trajectory tracking is necessary to guarantee smooth motion, accurate positioning, and consistent transport between warehouse zones.

1. Global View of the Technical Challenges

Although each subsystem introduces its own technical challenges, the most critical difficulties arise from their interaction during operation. Navigation accuracy, traction and lifting stability, and trajectory tracking must remain consistently coordinated to prevent positioning errors, load instability, or unsafe operating conditions. Achieving reliable interaction between the three subsystems is necessary to ensure accurate docking, stable load transport, and safe autonomous operation within a warehouse environments.

Based on the identified technical challenges, the main system constraints can be summarized and structured as follows in table 2.

3.3 Conclusion

This section defined what the robot is expected to perform, the environment in which it operates, and the main technical challenges that constrain its design. By translating these challenges into structured system constraints, a clear foundation is established for the subsequent conceptual and embodiment design phases.

Analysis of Existing Industrial Solutions

In order to position our Load Transport Robot within the current market, it is essential to analyze established competitors. This section focuses on high-performance autonomous mobile robots (AMRs) currently utilized in large-scale logistics.

Swisslog IntraMove AMR Series

One of the prominent solutions in the field of intralogistics is the IntraMove series by Swisslog. Unlike traditional Automated Guided Vehicles (AGVs) that follow fixed magnetic strips or wires, the IntraMove series is classified as an Autonomous Mobile Robot (AMR), designed for the flexible horizontal transport of goods.

System Description and Architecture

The IntraMove platform is engineered specifically for warehouse and distribution center environments. It is characterized by a low-profile chassis that allows it to slide underneath pallets or mobile racking units, lift them using an integrated hydraulic or electric mechanism, and transport them to a destination.

Key technical specifications include:

1. Payload Versatility: The series offers tiered payload capacities to suit different operational needs, specifically varying between 600 kg, 1,500 kg, and a heavy-duty model capable of up to 3,000 kg.
2. Holonomic Mobility: A defining feature of this series is its omnidirectional drive system. Typically utilizing Mecanum wheels, these robots can move sideways, rotate on the spot, and navigate tight spaces without a turning radius. This is a critical advantage over Ackermann-steering vehicles (like standard forklifts) in crowded environments.
3. Navigation and Safety: The robot employs SLAM (Simultaneous Localization and Mapping) or similar natural navigation technologies, allowing it to dynamically plan paths around obstacles. Safety is ensured through 360-degree LiDAR scanners that detect human presence and decelerate or stop the vehicle instantly to prevent collisions.

Market Application and Limitations

The primary motivation for the development of the IntraMove series was to address the rigidity of fixed conveyor systems. By decoupling the transport mechanism from the facility infrastructure, Swisslog provides a "scalable" solution; facilities can increase throughput simply by adding more robots to the fleet without construction work.

However, for the purpose of our project, it is worth noting the complexity and cost of such high-end industrial systems. While they offer high payload capacities (up to 3 tonnes), their reliance on complex fleet management software (such as Swisslog's SynQ) and expensive sensor suites makes them less accessible for smaller-scale or budget-constrained applications.

Geek+ P-Series (Shelf-to-Person AMR)

While the Swisslog solution focuses on heavy pallet transport, the Geek+ P-Series represents a different category of warehouse automation known as "Goods-to-Person" (GTP) or "Shelf-to-Person" systems. This solution addresses the picking process rather than bulk transport.

System Description and Functionality

The Geek+ P-Series robots are low-profile AMRs designed to move entire storage racks rather than individual items. The robot navigates underneath a mobile shelf, lifts it using an integrated lifting mechanism, and transports the entire shelf to a fixed workstation where a human operator picks the required items.

Key characteristics include:

1. Shelf-to-Person Methodology: Unlike traditional warehousing where humans walk miles per day to find items (Person-to-Goods), this system brings the inventory to the operator. This inversion of the workflow significantly reduces redundant walking time.
2. Swarm Intelligence: These robots operate in dense fleets. They utilize centralized fleet management to coordinate traffic, preventing gridlock and optimizing the path of hundreds of robots simultaneously.
3. High Storage Density: Because the robots are small and do not require wide safety aisles for human forklifts, storage racks can be packed more closely together, improving the volumetric utilization of the warehouse.

Relevance to Our Project

The Geek+ P-Series demonstrates the efficiency of "under-ride" lifting mechanisms, a concept we are also exploring for our Load Transport Robot. However, the P-Series relies on a highly controlled environment (often using QR codes on the floor for precise localization) and specific rack designs. Our project aims for a more general-purpose adaptability similar to the Swisslog platform, but with the compact agility seen in the Geek+ fleet.

6 River Systems "Chuck" (Collaborative Mobile Robot)

While the previous two examples (Swisslog and Geek+) focus on moving heavy loads or entire racks to minimize human movement, the Chuck robot by 6 River Systems represents a "Collaborative Mobile Robot" (CMR) or "Cobot" approach. Instead of replacing the human worker, it is designed to augment their capabilities.

System Description and Interaction

Chuck is a self-driving cart that leads human associates through the warehouse. Unlike the Geek+ system where the human stands still, Chuck guides the human to the correct location in the aisles to pick items.

Its design philosophy relies on three main pillars:

1. Directed Workflow: The robot features a large touchscreen interface that displays the item image, location, and quantity. It effectively acts as a supervisor, pacing the worker and reducing mental fatigue associated with searching for items.
2. Follow-Me and Lead-Me Navigation: The robot uses onboard sensors to navigate autonomously to the next pick location, eliminating the need for the human to push a heavy cart. Once the bin is full, the robot autonomously travels to the packing area, while a new robot immediately joins the worker.
3. Rapid Deployment: A key market advantage of this system is that it requires no infrastructure changes (like cages or magnetic tapes). It operates in existing aisles alongside humans, making it a "brownfield" solution.

Comparison to Our Project Strategy

While the Chuck robot highlights the potential for human-machine collaboration in dynamic environments, our project adopts a different operational logic closer to a "fetch-and-return" system.

Unlike Chuck, which continuously guides a worker through variable paths, our robot operates on a defined matrix map. The control system provides the robot with specific coordinates of a target object or shelf. The robot then autonomously navigates to that grid position, retrieves the object, and returns to the origin point. This relies on coordinate-based localization rather than the complex "lead-me" behaviors or flow-based navigation seen in the 6 River Systems solution.

Conclusion and Market Comparison

To summarize the current state of the art, we have compiled a comparative analysis of the leading solutions in the market. The table below highlights the trade-offs between payload capacity, navigation technology, and cost, key factors influencing our own design choices.

Technical Comparison: Market Solutions vs. Our Prototype

To further contextualize our design choices, the table below compares the technical specifications of the analyzed industrial solutions against the target specifications of our Load Transport Robot.

The comparison above demonstrates that while our robot has significantly lower payload and navigation complexity (IR sensors vs. SLAM), it excels in cost-efficiency, compactness, and maintainability. This makes it an ideal educational platform for studying core intralogistics concepts without the prohibitive costs of industrial equipment.

Patent Analysis and Freedom-to-Operate (FTO)

Market Trends and Strategic Gap

Based on our analysis of the state of the art, we have identified three prevailing trends in the current market:

1. Heavy-Payload AMRs: Excel in bulk transport (pallets) but are prohibitively costly and bulky for smaller facilities.
2. Goods-to-Person Robots: Significantly boost pick-rates but often lack the ability to stabilize goods fully autonomously without custom fixtures.
3. Safety vs. Autonomy Trade-off: Few affordable robots can autonomously navigate and stabilize goods safely without expensive sensor suites.

Conclusion: Current solutions optimize for either high-payloads or fast piece-picking, rarely both. This confirms an opportunity exists for a cost-efficient, compact AMR capable of delivering goods safely while remaining fully autonomous.

Patent Infringement Analysis

To ensure the viability of our project, we conducted a Freedom-to-Operate (FTO) analysis. The table below details relevant patents that pose a potential risk to our design and the specific engineering solutions we have adopted to avoid infringement.

FTO Statement: As summarized in the table, active claims regarding modular decks and specific lifting geometries prevent the direct copying of certain industrial designs. However, by adopting a fixed chassis and a mechanical pyramid interface to dock with a dedicated platform (rather than lifting goods directly via forks), our prototype avoids these protected claims. For all other aspects of the design, we have determined that we have full Freedom-to-Operate.

In this chapter, the conceptual design of the robot is presented. First, the main functions of the robot are identified and the possible means for each function are listed. These solutions are organized using a morphological chart. Based on the selected means, several design concepts are then generated. Each concept is evaluated and compared using different criteria . Finally, the most suitable concept is selected .

morphological chart

The load transport robot is designed to move autonomously to the location of a package, pick it up, transport it to a final target position, and deposit it. This sequence summarizes the main mission of the robot and defines the constraints of the system. To perform this task, several technical problems must be addressed, including autonomous navigation, reliable localization, load detection, stable motion, and object handling. Based on these challenges, the main features of the robot are defined as follows: navigation system, load detection , locomotion , an load handling , and control system .

Navigation system:

A first option is remote control, where the robot movement is directly controlled by a user. While this solution is simple to implement, it strongly reduces the autonomy of the robot .

A second solution is the use of QR codes placed on the floor. By scanning these QR codes, the robot can identify its current position. However, this approach requires frequent scanning and does not provide continuous guidance. The robot must constantly detect QR codes to know where it is and does not have a clear path to follow between two positions.

Another option is tracing different colored lines on the floor. In this case, the robot follows a specific color corresponding to a given destination or object. The main limitation of this solution is the high sensitivity of color sensors. Differentiating several colors reliably is difficult, and strong lighting variations can lead to detection errors. This method also requires a high contrast between colors.

The selected solution is a black line matrix on the floor, where each intersection represents a coordinate. This system allows the robot to navigate using coordination. The robot can determine its position by counting intersections and follow predefined paths. The package location and the target position can be defined using these coordinates, making this solution reliable, simple to implement, and well suited for autonomous navigation.

Detection of loads:

A first option is the use of a camera, where the robot observes the environment and detects the object visually. Although this solution is flexible, it requires image processing and increases system complexity. It also demands higher computational resources .

Another solution is the use of QR codes attached to the load. By scanning the QR code, the robot can identify the object and obtain information such as its destination. However, this approach requires the robot to constantly scan its surroundings in order to detect the QR code, which complicates the detection process.

A third option is the use of sensors to detect the presence of an object. While this solution is simple, it lacks precision, as it does not provide accurate information about the exact position of the load relative to the robot.

The selected solution is instruction-based detection. In this approach, the initial position of the robot, the position of the load, and the target position are defined directly in the control code. This method is simple, reliable, and well suited to the selected navigation system. It reduces system complexity and ensures consistent operation without requiring continuous scanning or complex processing.

Object handling:

A first option is a gripper arm. This solution allows the robot to grasp objects directly. However, it requires precise coordination and mechanical design to correctly grip the object. In addition, the object must have a suitable and regular shape, otherwise grasping becomes difficult or unreliable. This makes the system complex and less flexible.

Another option is a sweeper pickup system, where the object is pushed or collected from the ground using a sweeping mechanism. This solution is simple but offers limited control over the object and is not well suited for stable transport, especially for heavier or fragile loads.

A third solution is a sliding platform, where the robot is equipped with a moving belt or sliding surface. In this case, the object is placed on the robot, and the belt is activated to move the object by sliding. When the robot reaches the target position, the belt moves again to deposit the object. This solution allows simple loading and unloading, but it is mainly suitable when the object is already well positioned. It also offers less control over the object and requires more external intervention, making the system less autonomous.

The selected solution is a lifting platform. In this case, the robot deploys a platform that lifts or pulls the object from another mobile or fixed platform. This solution allows the robot to autonomously pick up and deposit loads with minimal human intervention. It is mechanically simpler than a gripper arm and more robust than sweeping or sliding systems, making it the most suitable choice for this project.

Locomotion system:

A two wheel system allows simple differential drive control, but it provides limited stability, especially when carrying a load.

A four wheel system offers good stability, but it increases mechanical and control complexity. More motors or a more complex transmission are required, and wheel synchronization becomes more difficult.

Tracks are mechanically complex, consume more energy, and are not necessary for the operating environment of the robot.

The selected solution is a three-wheel system. This configuration is simple to implement and allows easy use of a differential drive. It provides sufficient stability while keeping the mechanical and control design simple. For these reasons, the three-wheel locomotion system was chosen for this project.

Control:

For the control system,

the Arduino Uno was initially available and considered for the project. However, due to the large number of sensors, actuators, and connections required, the Arduino Uno reached its limits. The number of input and output pins was insufficient, making cable management and system integration difficult. The selected solution is the Arduino Mega. This microcontroller provides a larger number of input and output pins, allowing easier connection of sensors and actuators. It simplifies wiring, improves system organization.

Based on the selected solutions for each feature, the final morphological chart was established. Using this morphological chart, several design concepts combining the chosen solutions are then generated. These concepts will be analyzed and compared in the following section in order to select the most suitable one.

Concept generation

After narrowing down our scope of progress , the next task was to schematically design the different possible options for our project so that our robot would be fully capable of performingits function.

Concept A : Transport robot with a complete lifting system

Here, the idea is that during the execution of the parcel transport task by our robot, when the robot picks up a parcel, the transport is carried out solely by fully lifting the cart supporting the parcels. In this concept, the entire system would be centered on the lifting motor’s ability to keep the cart stable and to support the weight of the whole assembly.

Concept B : Transport robot with a traction system

Here, the idea is quite similar to the previous one, but instead of fully lifting the cart, we try to use the robot’s lifting platform to pull the carts carrying the parcels. Of course, caster wheels are added to the cart to facilitate the overall mobility once it is attached.

Concept C : Transport robot with a magnetic attachment system

The last idea proposed is the one that aims to minimize the risk of poor precision by using magnetic attraction for parcel transport. In other words, magnets would be installed on both the upper level of the robot and the lower part of the cart, with their attraction or repulsion being controllable, so that the robot can perform its function smoothly.

Concepts Evaluation

For each of the previously mentioned concepts, a qualitative study was carried out based on several technical, aesthetic, and safety aspects in order to definitively determine which of these three options would schematically represent our robot For this purpose, each selected aspect was assigned a weight according to its importance to us, and each concept was rated out of 5 for each aspect. The scores were then summed, allowing us to quantitatively compare our concepts. It was concluded that Concept B would be the best option according to our criteria, as it combines good maintainability, safety, technical simplicity, and fairly good performance in execution.

HIGH-LEVEL DESIGN / EMBODIMENT DESIGN

High level block diagram

A clear description of the different elements involved in the modelling of the product and its main subsystems is presented in Figure 6.1

The prototype is designed to locate a package mounted on a mobile platform equipped with wheels (see cart section). It moves underneath the platform, activates a lifting mechanism to mechanically attach to it, and then transports it to a designated drop-off location where it is released.

The navigation system relies on a line-following approach based on a reflectance sensor and a 3×3 grid map. The reflectance sensor enables the robot to detect black lines on a white background and move along the grid structure of the map. The pickup and drop-off locations are defined as coordinates in the code instructions. The robot is then able to determine its trajectory based on these coordinates.

To detect the mobile platform carrying the package, the robot uses Time-of-Flight (ToF) sensors, which allow it to position itself beneath the hole in the platform. The mechanical attachment is then ensured by a pyramid-shaped lifting block in the lifting mechanism. When the lifting mechanism is activated, the mobile platform slides along the inclined slopes of the pyramid thanks to its wheels, thereby compensating for small lateral misalignments during the robot’s positioning.

Locomotion is achieved using DC motors driving the wheels, which are controlled through a dual H-bridge motor driver allowing independent control of the wheel motors. The lifting mechanism is actuated by a DC geared motor controlled by a single H-bridge driver, providing the vertical motion of the lifting system.

All subsystems are controlled by an Arduino Mega, which acts as the main microcontroller and executes the navigation, sensing, and actuation logic. The Arduino Mega is supplied with 5 V, while the drivers are supplied with 12 V from a common DC power source.

Material selection analysis

Wheels and Tyre

The prototype uses two front driven wheels and a rear freewheel (caster). The front wheels were

selected as commercial components rather than being manufactured by 3D printing. They are

composed of a plastic rim combined with a rubber tyre. This choice was primarily motivated by

traction considerations: the rubber tyre provides a higher friction coefficient than 3D-printed plas-

tics such as PLA, while the plastic structure ensures sufficient rigidity. Using PLA wheels would re-

sult in a harder contact surface and an increased risk of slipping, particularly during line-following

and docking operations.

The wheel size was also a design constraint. The front wheels have a radius of 9 cm, which was

selected to maintain sufficient vertical spacing between the ground and the first structural level

of the robot. Given the number of components mounted beneath this level (motors, mounts, fas-

teners, and sensor supports), using smaller wheels would have significantly reduced the available

vertical space, making mechanical integration and assembly more constrained.

For the rear freewheel, a commercially available pivoting caster wheel was used. This freewheel

mainly consists of an iron structure, providing high mechanical robustness. The component was

selected based on availability rather than specific material properties. According to the manu-

facturer specifications, this caster wheel has a maximum load capacity of 50 kg, which is largely

sufficient for the mass of the prototype. The nominal height of the caster was lower than the height

of the front driven wheels. To compensate for this height difference and ensure that the robot re-

mains horizontal, the wheel was mounted using four metal threaded rods inserted through dedi-

cated holes in the wooden structure. The height adjustment was achieved by tightening hexagonal

nuts on the rods.

For the front driven wheels, no explicit maximum load capacity was provided by the manu-

facturer. Their suitability was therefore verified experimentally. During testing, the wheels were

mounted on the robot and successfully supported the full weight of the prototype during opera-

tion, with no visible deformation, slippage, or mechanical failure observed.

Chassis (first and second floor)

The material choice for the first and second floor chassis was wood. This selection was made fol-

lowing a material selection methodology based on Ashby diagrams, considering the compromise

between density and mechanical strength.

As shown in the Ashby diagram comparing density and elastic limit, wood combines a low den-

sity with a moderate elastic limit, making it suitable for lightweight structural components. For

large, flat elements such as the chassis floors, this results in a favorable stiffness-to-weight ratio,

allowing the structure to support the mounted electronic and mechanical components while lim-

iting overall mass.The chassis floors are large, thin, planar elements that are primarily subjected to

bending loads due to the weight of the mounted components. In such bending-dominated struc-

tures, stiffness is strongly influenced by the material’s elastic properties relative to its mass. As

shown in the Ashby diagram, wood provides higher stiffness for a given mass compared to plastic

materials for example, making it a better choice for the chassis.

Compared to plastic materials, for example, wood offers better geometric stability and more

predictable mechanical behavior for planar structural elements, making it more appropriate for

the first and second floors of the chassis.

Motors supports

Dedicated supports were required for the DC motors driving the wheels as well as for the geared

motor actuating the lifting mechanism. These supports play a structural role by maintaining mo-

tor alignment and by limiting vibrations transmitted to the chassis during motion.

Plastic was selected as the material for the motor supports. This choice was primarily moti-

vated by geometric considerations: the motors mounts have complex three-dimensional shapes

designed to match the motor geometry as shown in Figure 6.3. Such shape are difficult to realize

using wood, whereas the use of CAD modelling combined with 3D printing allows precise control

of dimensions and tolerances. In particular, the motor mount for the wheel DC motors includes

three holes: two holes are used to fasten the mount to the first floor of the chassis, while a third,

centrally located hole allows the insertion of an adjustment screw to control the clamping force

applied to the motor. This feature enables fine adjustment of how tightly the motor is pressed

against the chassis, helping to reduce vibrations and preventing the motor from loosening due to

these vibrations during operation.

From a structural point of view, the motor supports are subjected to static loads from the motor

weight and dynamic loads due to vibrations during operation. A more detailed validation could

have been performed through analytical stress estimation or finite element analysis (FEA) to verify

that the stresses remain below the material limits. Due to time constraints and the prototyping

nature of the project, this analysis was not carried out, and the design was instead validated during

assembly and testing.

Shaft adapter

Even if it’s not represented on figure 6.3, a shaft adapter was required to connect the output shaft

of the motor to the wheel hub. The geometry of this interface was highly specific, as it had to

match both the motor shaft profile and the internal geometry of the wheel. Due to this strong ge-

ometric constraint,a commercially available standard component could not be used. The adapter

was therefore designed using CAD and manufactured by 3D printing, which was the only practical

solution allowing full control over the geometry.

To prevent axial displacement of the wheel along the shaft, an additional part was designed

and added on the outer side of the shaft adapter. This part acts as an end cap and mechanically

blocks the wheel in the longitudinal direction while allowing torque transmission from the motor.

Reflectance sensor support

A support was designed for the reflectance sensor to ensure correct positioning relative to the

ground. According to the manufacturer’s specifications for the QTR-8RC reflectance sensor, the

sensing elements must be positioned within a distance range of approximately 3, mm to 9.5, mm

from the surface for proper operation. This requirement imposed precise control of the sensor’s

distance to the ground.Since the eight sensing elements of the reflectance sensors are not located

at the center of the sensor board, the 3D-printed support had to both hold the sensor and include

an offset opening that allows the sensing elements to face the ground. This constraint made the

design of the support more challenging.

This support could in principle have been manufactured using wood by assembling several

separate parts. However, this solution would have required splitting the geometry into multiple

components and assembling them manually, increasing complexity and alignment effort. Plastic

(PLA) manufactured by 3D printing was therefore chosen as a simpler and more practical solution,

allowing the complete geometry to be produced as a single part and ensuring straightforward as-

sembly.

Structural support rods

Structural support rods were used to ensure the rigidity and alignment of the different levels of

the prototype. M4 and M5 threaded rods were selected, as they were already available at disposal.

These rods are made of steel, providing sufficient mechanical strength and stiffness for the struc-

tural loads involved. The rods were cut to the required lengths using a manual saw and assembled

using hexagonal nuts.

Lifting mechanism

The lifting mechanism consists of a pyramid-shaped block manufactured in PLA, illustrated in Fig-

ure 6.3. This block is mounted on a metal ring with an internal thread that mates with the threaded

shaft of the DC geared motor. Due to the specific geometry required to match the threaded ring

and ensure proper engagement with the motor shaft, this part had to be custom-designed and

manufactured by 3D printing.

The block was also required to have a sufficient height to engage with the mobile platform

while remaining securely attached to the motor during the entire lifting motion. The pyramid

shape was chosen to allow the mobile platform to slide into position even in the presence of lateral

misalignment. Manufacturing this geometry using wood would have been particularly complex,

especially for the inclined surfaces.

An initial design combining PLA and flat intermediate layers was considered to reduce material

usage. However, due to the intrinsic behavior of the geared motor, which exhibits slight rotational

motion during actuation, the lifting platform could become misaligned when moving upward.

This misalignment prevented reliable engagement with the mobile platform when using a multi-

part structure. To ensure sufficient rigidity and maintain alignment throughout the lifting motion,

a single solid PLA block was therefore adopted.

Although a hybrid solution combining PLA and wood could have reduced material usage, the

fully printed PLA block was selected as a reliable and straightforward solution that ensured correct

operation of the lifting mechanism.

Cable management supports

Cable management supports were implemented to organize the wiring on the first floor of the

chassis. Small square wooden elements were used and glued onto the wooden base plate to guide

and secure the cables. This solution helped prevent excessive cable movement and improved both

reliability and readability of the wiring layout.

Electronic Components

Arduino Mega

An Arduino Mega is used as the main microcontroller instead of a standard Arduino Uno in order

to support the large number of digital inputs required by the reflectance sensor array.

Breadboard

A breadboard is used to distribute supply voltages between the Arduino Mega and the motor

drivers. It is also used to manage and interconnect the cables coming from the Time-of-Flight

sensors located on the upper level of the prototype.

Motor Drivers

The locomotion system is controlled using one dual H-bridge motor driver, allowing independent

control of the wheel motors. A single H-bridge motor driver is used to control the lifting mecha-

nism.

DC Motors

The prototype uses two DC motors to drive the front wheels of the robot. A DC geared motor is

used to actuate the lifting mechanism, providing sufficient torque for vertical motion.

Reflectance Sensor

A reflectance sensor is used for the line-following navigation method. It enables the robot to detect

black lines on a white surface and follow the paths of the 3×3 grid map.

Manufacturing process choice

3D printing

3D printing was used to manufacture parts requiring complex or highly specific geometries. This

includes the motor supports for the wheel DC motors, the support for the reflectance sensor, the

shaft adapter and its axial locking part, as well as the pyramid-shaped lifting block. These compo-

nents required precise geometric features that could not be easily produced using planar manu-

facturing methods.

Laser-cut materials

Laser cutting was used to manufacture the flat structural elements of the prototype. This includes

the first and second wooden floors of the chassis, as well as additional wooden elements used for

structural support and cable management. This process was well suited for producing planar parts

with accurate dimensions and mounting features.

Manual processing

Manual processing was required for several components of the prototype. Steel threaded rods of

type M4 and M5 were used for the structural assembly. These rods were cut to the required lengths

manually using a saw and assembled using hexagonal nuts.

Additional manual drilling was performed on the wooden plates when necessary to allow the

insertion of rods or screws. This was mainly required for alignment adjustments during assembly.

The mobile platform was also manufactured manually using wooden plates and threaded rods.

Its structure was assembled by drilling the wooden elements, inserting the rods, and securing them

with hexagonal nuts to form a rigid frame.

Assembly process

Justification of the manufacturing processes and assembly by means of diagrams and a compara-

tive analysis of the various possible options

The assembly of the prototype is organized into several structural levels, each fulfilling a spe-

cific mechanical or functional role. Figures 6.3a to 6.3d illustrate the different levels and their

integration within the complete structure.

Lower level: Locomotion and ground interaction

The lower level corresponds to the underside of the first laser-cut wooden plate and supports the

locomotion and ground-sensing components, as shown in Figure 6.3a.

Two DC motors are mounted at the front of the prototype using custom 3D-printed PLA motor

mounts, which are screwed directly onto the wooden plate. These mounts were designed with

dedicated fixation holes to ensure accurate positioning. At the rear of this level, a freewheel (caster

wheel) is mounted.

A 3D-printed PLA support is also mounted at the very front of this level to position the re-

flectance sensor at an approximate distance of 3 mm from the ground, corresponding to the opti-

mal operating distance specified in the sensor datasheet. This support is fixed using screws, and

its geometry was designed with dedicated mounting holes.

Base level: Electronics integration

The upper side of the first level hosts the main electronic components, as illustrated in figure 6.3b.

These include the Arduino Mega, the breadboard, the DC geared motor, and the two motor drivers.

Although this level appears as a single plate in the figures, in reality what we did was to assem-

ble two laser-cut wooden plates together using screws. One plate acts as a standard support, while

the second plate was laser-cut with dedicated holes at the locations of the electronic components,

so that once placed, the components are mechanically constrained and cannot move while the

prototype is in motion.

Multiple holes are present on this level. Circular holes are used for screws and threaded struc-

tural rods, while elongated oval holes are dedicated to cable management for design and efficiency

purposes.

Top level: Docking and Alignement Sensing

The Top level, shown in Figure added, supports the Time-of-Flight (ToF) sensors used for docking

and alignment with the mobile platform.

A linear hole is drilled in this plate to allow the sensor cables to pass through toward the bread-

board located on the base level. At the center of this level, a square opening of approximately 10.5

× 10.5cm is cut to allow the passage of the lifting block during vertical motion.

Intermediate Level: Lifting Mechanism

The intermediate level corresponds to the lifting mechanism itself, illustrated in figure added. The

pyramid-shaped lifting block is mounted onto a metal ring with an internal threaded hole, which

mates with the threaded shaft of the DC geared motor. The block was designed with dedicated

holes on its lower face to ensure proper fitting onto the ring, and screws are used to rigidly attach

the two components.

To prevent rotation of the lifting block when the motor is actuated, two vertical guide rods are

inserted through dedicated holes in the block. These rods are fixed into the first wooden plate

using pre-laser-cut holes and secured with hexagonal nuts. This configuration effectively forms a

custom linear actuator, allowing vertical translation while preventing rotational motion.

Structural Frame and Enclosure

The different levels are connected using threaded steel rods, which provide vertical structural sup-

port and maintain alignment between the plates. In addition, laser-cut wooden supports with

interlocking teeth are used to reinforce the structure and partially enclose the electronic compo-

nents. These supports are fixed using the same threaded rods and secured with hexagonal nuts.

Mechanical and physical constraints

The mechanical design of our robot is a central element of the project, as it directly conditions the

reliability of the system, its load capacity, its navigation accuracy, as well as its robustness in a real

industrial environment. All these aspects therefore impose realistic mechanical choices that are

easy to adjust and adapt.

geared motor specifications

• NO load max speed = 60 rpm

• Small torque = 2.6 kg.cm

• hight = 16 cm

DC motor specifications

• NO load max speed = 107 rpm

• nominal torque = 1.5 to 2 N.cm

• nominal voltage = 12V cm

Structure

The structure of the robot was designed to ensure mechanical rigidity, overall stability, and the

integration of the different subsystems of the robot (the motors, sensors, electronic components,

and the lifting platform). Above all, it must also withstand the forces generated during the traction

of the cart. The chassis is composed of rigid laser-cut wooden plates. To guarantee robustness and

allow easy disassembly during maintenance, metal bars are screwed on both sides of the lateral

faces.

The first level, as specified in previous section, is composed of two parts. The lower part of the

structure is dedicated to the mounting of the motors and the driving wheels, in order to transmit

the traction forces to the ground as efficiently as possible. The upper part of this same level sup-

ports the majority of the electronic circuitry, namely the sensors, the control electronics, and the

motor used for the mobile platform. Particular attention was paid to the alignment of the me-

chanical components in order to avoid parasitic stresses and to ensure reliable operation during

motion and lifting phases.

This modular architecture also allows future developments of the robot, in particular the addi-

tion of new sensors or the adaptation of the upper platform to different types of carts.The overall

dimensions of the robot are as follows:

• Structure height = 27 cm

• Base width = 24 cm

• Base length = 31 cm

motor and traction system

The mobility of our robot is based on a simple architecture composed of two lateral wheels and

a caster wheel placed at the rear to ensure stability. This choice was made primarily for its me-

chanical simplicity, especially for implementing differential drive navigation; The two DC motors

6are directly coupled to the driving wheels, which allows generating the traction force needed

to move the robot and pull the cart.

The rear caster wheel does not participate in propulsion; its position facilitates changes in

direction without adding constraints.

The robot’s traction capability depends directly on the torque provided by the motors and the

wheel radius. The traction force generated by a motor can be expressed by the relation:

Ftraction = T/r

where T represents the motor torque and r represents the motor torque and to ensure overall

stability, the geared motor was positioned directly above the axis of the DC motors, as this area will

bear the most load. Other components were intelligently arranged to guarantee a certain stability.

External Support: The Cart

For transporting loads (such as delivering packages in warehouses), among all the options studied,

we opted for horizontal traction of the cart on which the loads will be placed. the prototype is mainly made of wood and iron bars that maintain its frame.

On top of the first wooden plate, a space of (11 cm×11 cm) was left to allow the lifting platform

to be properly attached to the cart, ensuring appropriate dimensions for effective traction when

the system is connected.

The dimensions of the cart are as follows:

• Height = 34 cm

• Width = 24 cm

• Length = 30 cm

Circuitry and Sensors

Microcontroller

To control the actuators and collect data from the sensors, we use an Arduino Mega 2560. This microcontroller was initially selected instead of an Arduino UNO because the project was designed to use a QTR-8A reflectance sensor, which requires a high number of analog inputs that the UNO could not provide. Due to a misunderstanding during the ordering process, a QTR-8RC sensor was delivered instead, which no longer required additional analog pins. Nevertheless, the Arduino Mega 2560 was kept as it comfortably accommodates all subsystems without hardware constraints.

The robot includes a QTR-8RC line-following sensor, two Time-of-Flight sensors used to detect the presence of a platform on top of the robot, two DC motors for locomotion, one geared motor with a lead screw for the lifting mechanism, and the associated motor drivers. The Arduino Mega 2560 provides sufficient digital I/O, PWM outputs, and communication interfaces to manage these components simultaneously while ensuring reliable operation.

Line following sensor

Line following is achieved using a QTR-8RC reflectance sensor array. This sensor consists of eight infrared emitters paired with eight phototransistors. Each channel operates according to a charge and discharge principle based on RC timing rather than analog voltage measurement.

When activated, the capacitor associated with each sensor is charged through a digital output pin. After switching the pin to input mode, the capacitor discharges through the phototransistor. The discharge time depends on the amount of infrared light reflected by the surface below the sensor. A white surface reflects more infrared light, causing a faster discharge, while a black line absorbs most of the light, resulting in a slower discharge time.

By measuring the discharge time for each sensor, the system determines the position of the robot relative to the line. The combination of the eight digital measurements allows accurate detection of line position and orientation, enabling reliable line following and intersection detection.

Time-of-Flight sensors

Two Time-of-Flight distance sensors (VL53L1X) are used to detect the presence of a package on the lifting platform. These sensors are connected to the Arduino Mega 2560 via the I2C communication bus using the SDA and SCL pins.

The operating principle of the Time-of-Flight sensor is based on measuring the time taken by an emitted infrared laser pulse to travel to an object and be reflected back to the sensor. Knowing the speed of light, the distance to the object is calculated using the relation:

d=c⋅Δt2d = \frac{c \cdot \Delta t}{2}d=2c⋅Δt​where $c$ is the speed of light and $\Delta t$ is the measured round-trip time of the signal.

In this project, the sensors are not used for navigation but exclusively for package detection. A distance threshold of 100 mm is defined in the control program. When the measured distance is below this threshold, the system confirms the presence of a platform over the robot. This verification step ensures that the lifting mechanism is only activated when the robot is at the good intersection, preventing unnecessary motion.

Drive motors

The robot locomotion system relies on two DC geared motors, each driving one wheel in a differential drive configuration. The selected motors are JGB37-520 DC motors, operating at a nominal voltage of 12 V. They provide a no-load speed of 107 RPM and a stall torque of 2.6 kg.cm, which is sufficient to move the robot and its payload.

Steering is achieved by controlling the relative speeds of the two motors. When both motors rotate at the same speed, the robot moves in a straight line. When their speeds differ, the robot follows a curved trajectory. Although backward motion is not implemented, the robot still follows a differential drive kinematic model, as direction changes are obtained through forward speed variation between the two wheels.

Lifting motor

The locking mechanism of the package is actuated by a geared DC motor coupled with a trapezoidal T5 lead screw and a brass nut. This system is intended to mechanically lock the package to the robot platform for safe transport.

The robot is equipped with a rigid cubic locking element, one face of which is pyramidal. The platform on which the package is placed contains a central opening aligned with this locking element. Initially, the platform and the package are independent from the robot. When the motor is activated, the lead screw converts the rotational motion of the motor into a vertical linear displacement, causing the cubic locking element to rise through the opening in the platform. The pyramidal face creates a self-centering effect, ensuring proper alignment between the platform and the robot. Once the locking element reaches a sufficient height, the platform and the robot become mechanically locked together.

A major advantage of this locking approach is that the system does not need to provide the full energy required to lift the package, which could be significantly heavy. The mechanism only needs to generate enough force to engage the locking interface, reducing the required motor torque and overall energy consumption.

To further improve mechanical robustness, two additional lead screws are placed in parallel with the main driven screw. These screws prevent rotation of the cubic locking element during its vertical motion and share the mechanical load applied to the system. As a result, once the locking element is engaged, the torque applied to the driven lead screw corresponds to approximately one third of the total torque, the remaining load being distributed across the two passive screws. This load-sharing configuration reduces mechanical stress, improves stability during motion, and increases the reliability of the locking mechanism.

Motor driver

Motor drivers are used to interface the Arduino Mega 2560 with the motors, as the microcontroller cannot supply the required current and voltage directly.

The two drive motors are controlled using an L298N motor driver. This dual H-bridge driver allows independent control of each motor, enabling speed regulation via PWM signals and direction control. It supports motor supply voltages up to 46 V, making it suitable for the 12 V drive motors.

The lifting motor is controlled using an IBT-2 motor driver. This driver is based on a single high-power H-bridge architecture and supports supply voltages ranging from 5 V to 27 V. It is well suited for applications requiring higher current, such as the lifting mechanism. The use of dedicated drivers protects the microcontroller and ensures reliable and safe motor operation.

Motor driver

Here is the schematic of the all circuit. The only comment is that the QTR8RC was not found on the software, therefore a QTR8A was used to do the schematic. The main point which requires some attention was to know which of the digital pins of the Arduino mega are compatible with PWM but the rest was quite intuitive.

Caption: General electronic circuit

Software

The software architecture of the robot was designed to ensure robust autonomous navigation, reliable task execution, and clear separation between low-level control and high-level decision making. The program is implemented on an Arduino microcontroller and integrates line following, grid-based navigation, object detection, lifting mechanisms, and mission management.

High-Level Software Architecture

Figure presents the global software block diagram of the robot. It illustrates the interaction between perception, control, decision-making, and actuation layers.

At the highest level, the system can be divided into four main functional blocks:

1. Sensor acquisition and preprocessing
2. Low-level motion control
3. Navigation and path planning
4. Mission and state management

These blocks communicate through shared state variables and are executed sequentially inside the main control loop.

Caption: High-level software architecture of the robot

Main Control Loop

The program runs inside a continuous loop that evaluates the current mission phase and selects the appropriate behavior. The loop can be conceptually divided into two main execution paths: non-navigation phases and navigation phases.

Non-navigation phases include lifting actions and end-of-mission states, during which the motors are stopped and dedicated routines are executed. Navigation phases rely on continuous sensor feedback and real-time motor control to ensure stable movement along the grid.

Line Following and PID Control

Line following is achieved using an array of infrared reflectance sensors positioned at the front of the robot. The sensor array provides a weighted position value representing the relative offset of the robot with respect to the line.

A proportional-derivative (PD) controller is used to compute a correction term applied to the left and right motor speeds. The control law is expressed as:

u(t)=Kp⋅e(t)+Kd⋅de(t)dtu(t) = K_p \cdot e(t) + K_d \cdot \frac{de(t)}{dt}u(t)=Kp​⋅e(t)+Kd​⋅dtde(t)​where $e(t)$ represents the deviation from the center of the line. The base speed defines the nominal forward velocity, while the correction term dynamically adjusts motor speeds to maintain alignment.

To prevent actuator saturation and ensure stable behavior, the motor commands are constrained within predefined minimum and maximum speed limits.

Intersection Detection and Handling

Intersections are detected by counting the number of sensors simultaneously detecting the line. When this number exceeds a predefined threshold, the system identifies the presence of an intersection.

Upon detection, the robot temporarily disables further intersection detection to avoid double triggering. The robot then updates its discrete grid position based on its current orientation and executes a predefined intersection handling sequence.

This sequence consists of:

1. A short forward alignment phase
2. A rotation phase if required (left or right turn)
3. A line reacquisition phase before resuming PID control

The short forward alignment phase could be counter-intuitive but it improved a lot the repeatability of the robot to turn with success.

Grid-Based Navigation and Path Planning

The environment is modeled as a discrete grid, where each intersection corresponds to a grid cell. The robot maintains its current grid coordinates and orientation in real time.

To navigate between target cells, a breadth-first search (BFS) algorithm is used to compute the shortest valid path while avoiding forbidden cells. The BFS algorithm determines the next cell to reach, which is then translated into a local motion command (straight, left turn, or right turn) depending on the robot's orientation.

This approach ensures deterministic and optimal navigation on a small grid while remaining computationally lightweight.

A limitation appears due to the small scale of the grid used in this project. Because the robot often operates near the boundaries of the grid, some path decisions that are valid from a purely logical perspective may lead to less convenient maneuvers in practice, particularly near edges or corners.

This behavior does not affect the overall validity of the navigation strategy, but rather reflects the constraints imposed by the limited size of the environment. On a larger grid or in a less constrained setup, the same approach would operate without this limitation.

Mission management

The robot behavior is governed by a mission level state machine. Each mission phase corresponds to a specific objective and associated behavior.

The main mission phases include:

1. Navigation toward the pickup location
2. Lifting the load
3. Navigation toward the drop-off location
4. Depositing the load

Transitions between phases are triggered by logical conditions such as reaching a target grid cell, detecting an object, or completing a lifting action.

Navigation logic Overview

The following figure details the internal navigation logic used during movement phases. It highlights the interaction between sensor readings, grid position updates, path planning, and motor control.

Caption: Navigation and decision-making block diagram

Object detection and lifting control

Object detection is performed using time-of-flight distance sensors mounted on the top of the robot. Multiple measurements are combined to ensure robust detection and to filter out spurious readings.

Once an object is detected at the pickup location, the navigation system is paused and the lifting mechanism is activated. The lift is controlled using timed motor commands, ensuring consistent motion without requiring additional sensors.

After the lifting or depositing operation is completed, the robot resumes navigation according to the current mission phase.

Safety and Robustness considerations

Several safeguards are implemented to ensure reliable operation. Intersection detection is temporarily disabled after each detection to prevent multiple triggers.

Additionally, speed, torque and timing parameters are adjusted depending on whether the robot is carrying a load. The weight of the all thing almost doubling up, for the robot to manage to move the packages it should have enough torque. But with this high torque it goes way to fast for a good navigation when the robot is not loaded. Those parameters requires a lot of calibration and have to be changed manually depending on the load, for now the robot is calibrate to transport a load of 150g.

Software conclusion

The implemented software architecture provides a robust and modular solution for autonomous navigation and task execution on a structured grid. By combining real-time sensor feedback, grid-based planning, and a mission-level state machine, the robot is capable of performing complex sequences such as pickup, delivery, and return-to-home autonomously.

The modular structure of the code allows for future extensions, including larger grids, additional sensors, or more advanced control strategies.

The main improvement that could be interesting are developing a function adjusting the different parameter that have to be calibrated in function of the load the robot has to track. Other improvement would be to make it possible for the robot to go backwards or to do 180 degrees turns which would have greatly help the manoeuvrability on this small grid.

Critical Review

In this section, we analyze the current implementation of the Load Transport Robot and propose improvements to enhance its functionality, reliability, and scalability. This review reflects on the design choices we would reconsider if we were to restart the project, particularly regarding actuation, power management, and mechanical constraints.

Actuation and Lifting Mechanism

Currently, the robot uses a standard DC geared motor to drive the linear actuator for the lifting mechanism.

1. Motor Choice: While the DC geared motor provides sufficient torque, it lacks inherent position control. We rely on timing or external switches to determine the lift height, which can be imprecise.
2. Improvement: If we were to redesign the system, we would replace the DC motor with a high-torque stepper motor or a continuous rotation servo. This would allow for precise control over the vertical position of the lifting pyramid without relying on imprecise timing loops, ensuring the mechanism stops at the exact height required every time.

Mechanical Design and Docking Strategy

The robot utilizes a 3D-printed pyramid structure to dock with the platform. This shape was chosen specifically to solve misalignment issues: as the pyramid rises into the platform's socket, the sloped sides physically push the mobile platform (which is on wheels) into perfect alignment with the robot.

1. Shape Optimization Constraints: We considered optimizing the lifting element to be more compact or shaped differently (e.g., a lever or scissor design). However, due to the robot's double-deck architecture (two distinct levels), a more complex shape risked getting stuck between the chassis layers during retraction.
2. Conclusion: The current "blocky" pyramid design, while bulky, was the most robust solution to ensure smooth vertical travel between the robot's levels without jamming, while successfully performing its primary function of passive mechanical alignment.

Navigation and Localization

The current navigation relies on QTR-8 infrared sensors for line following and intersection counting.

1. Sensor Fusion: As identified in our testing, relative localization (counting lines) is prone to cumulative errors. A significant improvement would be integrating Absolute Localization using RFID tags or QR codes placed on the floor at key intersections. This would allow the robot to "reset" its position data periodically, eliminating drift.
2. Platform Detection: We currently use "fly sensors" (proximity sensors) strictly to verify if a platform is present above the robot. This binary check is effective and sufficient; unlike the navigation sensors, this system does not require upgrading as the fine alignment is handled mechanically by the pyramid, not by sensors.

Power Management System (BMS)

One critical component missing from the current design is a dedicated Battery Management System (BMS).

1. Current State: The robot runs directly from the power source without advanced monitoring of cell health or balanced charging/discharging protection.
2. Risk: This increases the risk of over-discharging the Li-Ion cells, potentially damaging the battery life or causing voltage instability for the motors.
3. Improvement: In a future iteration, integrating a BMS module would be a priority. This would ensure safe operation, prolong battery life, and provide the microcontroller with accurate data regarding the remaining charge, enabling the robot to alert the user before a power failure occurs.

Conclusion

The developed prototype successfully demonstrates the "fetch-and-carry" logistics concept using a matrix map. The pyramid docking system proved to be an effective mechanical solution for self-correction, eliminating the need for complex alignment sensors. However, upgrading the actuation to stepper motors and integrating a BMS would significantly elevate the robot from a functional prototype to a reliable, industrial-grade solution.

The sustainability of the proposed transport robot is evaluated assuming its deployment as an industrial warehouse system, based on the same functional concept developed in this project. Sustainability is therefore addressed through energy efficiency, durability, maintenance reduction, and long-term usability, rather than through prototype-specific optimizations.

Energy Efficiency and Control Strategies:

In continuous warehouse operation, transport robots perform repetitive tasks over long periods. For this reason, sustainability is improved by designing the system to operate at nominal efficiency points instead of peak performance. Proper motor sizing, optimized transmission ratios, and controlled operating speeds reduce unnecessary energy consumption and thermal stress.

At the control level, implementing smooth motion profiles and speed regulation minimizes current peaks, reduces mechanical wear, and extends actuator and battery lifetime. These strategies improve both energy efficiency and system reliability without increasing hardware complexity.

Durability and Maintenance Reduction:

Long-term sustainability depends primarily on extending the service life of the robot. An industrial version would prioritize robust mechanical components designed for fatigue resistance and continuous duty rather than minimal weight. Simplifying the mechanical architecture and limiting the number of moving parts reduces failure probability and maintenance frequency.

Lower maintenance requirements directly translate into reduced material consumption, fewer component replacements, and lower operational downtime over the robot’s lifetime.

Repairability and Modular Architecture:

To prevent premature system replacement, the robot should be designed with repairability as a core requirement. High-wear components such as motors, wheels, sensors, and control electronics must be easily accessible and replaceable. A modular architecture allows individual subsystems to be repaired or upgraded independently, reducing waste and extending system usability.

This modularity also enables future upgrades in control or sensing technologies without requiring a complete redesign of the mechanical structure.

System-Level Sustainability:

Beyond the robot itself, sustainability is reinforced at the warehouse level. By reducing unnecessary internal transport, idle time, and handling errors, the robot contributes to more efficient material flow and lower overall energy usage within the facility.

Conclusion:

In summary, the sustainability of the transport robot is achieved through disciplined engineering choices: energy-efficient control strategies, durable and low-maintenance mechanical design, and a modular architecture enabling repair and upgrades. These measures allow the same transport robot concept to remain reliable, efficient, and sustainable over long-term industrial operation.

This project led to the design and realization of an autonomous load transport robot inspired by warehouse automation systems. The work started with the definition of the task and operating environment, followed by the selection of suitable solutions for navigation, load handling, locomotion, and control.

After exploring different concepts, a final solution was selected and implemented. The robot was then built by assembling laser-cut mechanical parts, integrating electronic components, and programming the control logic. All subsystems were combined to enable the robot to autonomously move, pick up loads, transport them, and place them at the desired locations.

Tests showed that the robot is able to navigate, pick up a load, transport it, and deposit it at a target position. While the prototype meets the main objectives of the project, the development process also revealed several limitations. Improvements could be made in terms of precision, autonomy, and robustness, especially for long-term or industrial use.

Overall, this project provided practical experience in integrating mechanical design, electronics, and programming into a single functional system. It demonstrates the challenges and trade-offs involved in developing a complete mechatronic system from concept to prototype.

[1] Cherry’s Industrial Solutions, “Top Warehousing Issues and Challenges,” 2024.

[2] McKinsey & Company, “Automation in Warehousing and Logistics,” 2024.

[3] Oberlo, “Global E-commerce Growth Statistics,” 2024.

[4] eMarketer, “Global E-commerce Forecast 2024–2028,” 2024.

[5] TGW Logistics Group, “Ergonomics and Safety Benefits of Warehouse Robots,” 2023.

[6] Mecalux, “Main Logistics and Warehousing Problems,” 2024.

[7] Raymond Handling Concepts, “Warehouse Robotics and Productivity Gains,” 2025.

[8] Ocado Intelligent Automation, “Warehouse Automation Statistics and Performance,” 2024.

[9] Glocate, “How Warehouse Automation Reduces Labor Dependency,” 2024.

[10] ClickPost, “Warehouse Industry Statistics and Injury Rates,” 2024.

[11] OTTO Robotics, “OTTO Robots: Your Guide to the World of Robotics,” 2024.

[12] Mobile Industrial Robots (MiR), “Autonomous Mobile Robots – Discover AMRs from MiR,” 2024.

[13] Nord Modules, “Top 10 Autonomous Mobile Robot Companies,” 2024.

[14] Swisslog, “Introduction of the IntraMove AMR Series for Dynamic and Flexible Horizontal Pallet Transport,” 2025.

[15] Standard Bots, “Top 12 Warehouse Robotics Companies in 2025: Leaders, Startups, and Competitors,” 2025.

[16] World Intellectual Property Organization, “Autonomous Mobile Robot with a Single Modular Platform,” Patent WO2021019383A1, 2021.

[17] United States Patent Office, “Modular Autonomous Platform,” Patent US20200064841A1, 2020.

[18] Justia Patents, “Robotic Multi-Gripper Assemblies and Methods for Gripping and Holding Objects,” U.S. Patent No. 12,296,480, 2025.

[19] Justia Patents, “Automated Warehousing Using Robotic Forklifts,” U.S. Patent No. 8,965,561, 2015.

[20] Justia Patents, “Sreehari Kumar Bhogineni – Inventions, Patents and Patent Applications,” 2024.

Mechatronics project 2025-2026 (Group 4)

https://drive.google.com/drive/folders/17bdKvANx0NGD0AYR6ha1LbraKtaXeWKB?usp=sharing