Ongoing: OOPS for learning robots

Setup: virtual reality with realistic physics simulation: inertia, friction, obstacles, rough terrain, etc.

Robot’s goal: maximize reward, minimize pain.

Approach: Fast primitives for thinking, time-consuming primitives for acting. Use Optimal Ordered Problem Solver for combining them, to generate optimal muscle movements & 3D trajectories of snake robots and other agents

Previous slide

Next slide

Back to first slide

View graphic version