"PraxisGuard team on AI Infra Summit Hackathon Hackathon"
Team Idea
Modern manipulation policies (ACT, SmolVLA, Pi0.5) predict motor trajectories in chunks (e.g., 50 steps at a time). While chunking stabilizes high-frequency multi-joint control, it introduces an "action horizon blindspot." If a hazard appears mid-execution, the robot operates open-loop until the chunk finishes.
Team Members
.png&w=256&q=75)
Aryan Rajput
aryan10582568535
