Event
UIUC proposes CARL method with guaranteed stability for robot reinforcement learning
1 reports · 1 sources
Researchers at the University of Illinois Urbana-Champaign have proposed a contraction-aware reinforcement learning method, CARL, which combines reinforcement learning with contraction theory from control theory to provide stability guarantees for robots. The researchers say tracking performance improves by up to 15% compared with existing methods. The method was tested on three types of real robots: the wheeled TurboBot 3 Burger, a flapping-wing drone, and the hexapod Jethexa. The researchers say tracking errors can still occur, but the method provides a mathematical upper bound on the error magnitude; performance in simulation also remained better in real-robot tests. CARL is openly accessible, and the code is available on GitHub.
Report timeline
- University of Illinois proposes CARL method to keep reinforcement learning stable for wheeled, flying, and six-legged robots
Researchers at the Grainger College of Engineering at the University of Illinois Urbana-Champaign combined reinforcement learning with control theory to propose a contraction-aware reinforcement learning method, CARL, which guarantees robot stability while improving tracking performance by up to 15% compared with existing methods.
Tech Xplore Robotics ·