"RoboProof team on AMD Developer Hackathon: ACT III Hackathon"
Team Idea
We’re building RoboProof, a physical-AI platform for VEX V5 robots. It connects the original Simulator/iraLIB with motion control and reinforcement learning: the robot follows a path using a Guiding Vector Field (GVF), while a TD3 policy learns to adjust its movement from simulated sensor feedback. Our goal is to help a robot turn a target or task into reliable movement—and eventually improve its game performance—without asking a language model to control motors directly. The 512-neuron, 10-hidden-layer network is our primary learning candidate, now training on an AMD MI300X. we’re measuring whether the learned policy improves movement against the existing controller, and we won’t claim success until it passes independent simulator tests. We’re looking for teammates with experience in VEX robotics and mechanisms, C++ controls (PID/LQR/GVF), reinforcement learning and PyTorch, simulator physics/testing, or product design and UX.
Team Members
.png&w=256&q=75)
Zhanibek Kassimov
kassimovjanibek55649943
.png&w=640&q=75)