PiQSaR: Pnuematic Bio-Inspired Soft Quadrupedal Robot for Delicate Object Handling

Traditional rigid robots are unsuitable for lifting fragile, irregularly shaped objects and operating under unexpected conditions, as they can damage objects, become non-functional or break, and exhibit limited flexibility. Innovations in soft robotics overcome these issues and are impactful in fields like agriculture, ocean exploration, healthcare, etc. Our project aims to develop an advanced soft robot, PiQSaR - a Pneumatic bio-Inspired Quadrupedal Soft actuator Robot with integrated flex sensors for delicate object handling, mapping, and detection. We prototyped several iterations of soft robots, each with distinct qualities. Throughout our iterations, we improved the airflow distribution, changed components, and improved the structural integrity of robots. We experimented with silicone rubbers, 3D printed molds, lego molds, syringe tubes, and cloth/tape bases to create multiple prototypes including soft tentacles, grippers with central air chambers, and quadrupedals with separate air chambers. Under pneumatic pressure, we demonstrated the soft robots' ability to grip and lift objects. We attached flex sensors to our robot to gather flex data from PiQSaR’s legs. This real-time data provided us with the size and shape of the objects with 100% reliability. Additionally, we incorporated object detection, which utilized sensor data to recognize the objects the soft robot was grasping, achieving a success rate of 80+%. By applying the flex sensor data, we consistently mapped the object that the robot was gripping. In the future, we plan to build an industry-grade soft robot using air pumps, stronger elastomeric polymer, and expand our training datasets and algorithms for improved object detection.

We made an entire presentation that will go into further detail on everything we did and it has lots of pictures and results on our project. We also 3D printed the molds with TinkerCAD.

FYI - we are a group of three eight grade students

Apart from the materials shown in the image above, you would also need zip ties, barbed couplings, an object that the robot has to grab for testing, and a plate that can be used to make the base layer of the robot.

The procedure above shows how to make the quadrupedal soft robot with separate air chambers. You would use the same procedure for the quadrupedal soft robot with a central air chamber. There is also a diagram on how the soft robot will look like.

For this iteration, you would need to use the same procedure as Step 1, but instead of using a 3D printed mold, you have to use a lego base.

The cloth base is integrated into the soft robot for various reasons. When you do not have the cloth base, the top and base of the robot both expand in all directions instead of bending and gripping down. Since the cloth does not expand, it forces the robot to bend down and grip the object.

Attached is a video of our code working given by data from flex sensors. It detects the object that the robot is gripping via our Machine Learning code. Then, it depicts a map of the object that the robot is gripping.

Read our presentation for more information.