Towards Robust Execution of Long-Horizon Whole-Body Control Tasks

Recent progress in robot learning has significantly advanced robotic capabilities on short-horizon skills and well-defined tasks. However, despite these advances, robots, such as humanoids and robotic arms, continue to struggle when deployed on long-horizon, complex tasks in the real world. Topics of interest of this workshop include on-policy progress estimation for multi-stage tasks, run-time failure recovery, hierarchical and memory-augmented policies trained under long-horizon rollouts, sim-to-real transfer with execution-time fine-tuning, and leveraging foundation models for task planning, perception, and feedback within closed-loop, on-policy execution.