Ep#105: Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning

Podcast: RoboPapers · Chris Paxton and Michael Cho

Opublikowano: 2026-09-18

Imitation learning, especially with interventions, has driven so much recent robotics progress. However, improving a policy via targeted interventions until it reaches a useful and deployable success rate is a time and labor intensive process. Instead, wouldn’t it be great if policies could improve on their own? That’s what Varun Giridhar and Animesh Garg join us to talk about. In Q-Planning, they start with a large policy like pi-0.5, and add a Q-function estimator to predict value instead of just actions, then use both successful and failed rollouts to update this Q-function online, then use it t...