Event
PrepTurn: A Reinforcement Learning Framework for Anticipatory Turning in Quadruped Robots
1 reports · 1 sources
Zunzhi You and co-authors propose PrepTurn, a reinforcement learning framework for anticipatory turning in quadruped robots during continuous local navigation. It maps the current and next pose goals directly to joint actions, allowing the upcoming goal to influence how the robot approaches the current goal, and uses sustained rewards to encourage the robot to complete the entire goal sequence while moving. The authors report that simulation and onboard experiments on a Unitree Go2 show that the framework achieves a high sequential traversal completion rate and reduces stop-and-go motion; automatic curriculum learning that gradually increases goal difficulty is essential for stable turning behavior. The authors also report that when the tolerance for reaching goals is relaxed, the policy exploits the degrees of freedom permitted by the task, producing emergent behaviors.
Report timeline
- PrepTurn: An anticipatory turning reinforcement learning framework for fast sequential quadruped navigation
PrepTurn is a reinforcement learning framework for anticipatory maneuvering in continuous local navigation. It maps the current and next pose goals directly to joint actions, letting the upcoming goal influence how the robot approaches the current one. The framework uses a persistent-credit reward to encourage completing the entire goal sequence while in motion, and an automatic curriculum that gradually increases goal difficulty to stabilize turning behavior; simulation and onboard Unitree Go2 experiments show high sequential traversal completion rates and reduced stop-and-go motion.
arXiv cs.RO ·