A Hardware and Software Testbed for Underactuated Self-Assembling Robots: Extended Abstract

We present the implementation and characterization of an affordable testbed for underactuated multi-agent, self-assembling systems. There has been recent interest into the control of nano- and micro-scale active particle systems, but these systems are often difficult to manufacture and observe, hindering control research. Our testbed offers an accessible way to experiment with different design and control approaches. The testbed is composed of an off-the-shelf rolling weaselball toy and a 3D printed external hub that modifies the agent's dynamics. The software toolbox includes simulations and code for data extraction and analysis of the weaselballs. The advantage of our testbed for studying distributed robotic systems is that these robots can be made quickly and cheaply, are relatively small, and do not require complex or expensive environments. The software in our toolbox includes a high fidelity Gazebo simulation, and Python code for analyzing trajectories from both simulation and overhead video of the system. Using this toolbox, we present useful computed properties of the system with regards to object clustering.