Phase One

Create a low-cost, small, and autonomous robot by the October 23rd, 2025 deadline. The robot must demonstrate stochastic movement and prove the viability of vibration-driven motion. Split into two groups, our team focused on the hardware/software implementation as well as the body design. The body of the prototype was created with easy deconstruction and assembly in mind. The main body is skeletonized to limit weight. The base of the model utilizes thin legs to maximize vibration. The motors are positioned orthoganally to each other. The hardware sits vertically within a fuselage to limit any snagging on the holes while inserting components. The microcontroller was programmed to read garbage from the ADC to ensure randomized, non-linear motion.

Robot body 1 Skeletonized Body
Robot body 2 Body Insert
Base Base
Microcontroller SAMD21 Microcontroller
Prototype Video

Phase Two

Build five robots using the upgraded XIAO Seeed microcontroller and velcro sticking technology. The five robots adopted a new design that optimizes space usage and creates more surfaces for sticking. These robots use ICU feedback to implement autononomous movement.
These robots are then released into varying sized arenas including a 3D printed arena of 20cm x 20cm, and foam board arenas of 30cm x 30cm and 40cm x 40cm. With each arena, the robots will be observed to gather data on their complex dynamics including peak speed, maximum and average turning radius, mean square speed, and mean square displacement over time.

Victor 2.0 Prototype VICTOR 2.0 Prototype
Victor 2.0 Legs Close Up VICTOR 2.0 Legs
XIAO Seeed Microcontroller XIAO Seeed Microcontroller
Victor 2.0 Full Formed Complete VICTOR 2.0
Prototype Video

Phase Three

Develop a printed circuit board (PCB) and integrate recursive Q-learning (AI) to create a tabular system to track robot performance. With the development of the PCB, the body design will face fewer limitations which allow for a more optimized design from the observations made in phase two. Paired with a better design, integrating Q-learning will allow the robots to walk independantly. This independence will create optimal pathing to other robots creating a faster approach to emergent behaviors.