Event

I3L: An Iterative Interactive Imitation Learning System

1 reports · 1 sources

The research team proposes I3L, a hardware-software system that combines whole-body correction with iterative policy refinement to mitigate out-of-distribution degradation of offline imitation learning policies in long-horizon whole-body mobile manipulation. Its low-cost, open-source JoyLo+ interface, through a leader arm that follows the policy and touch-based takeover, allows operators to make coordinated corrections to various body parts; iterative deployment and retraining then let operators correct failures of the updated policy. The team evaluates I3L on three robot embodiments and eight simulated and real-world tasks, including four long-horizon whole-body mobile manipulation tasks. Controlled simulation experiments show that four rounds of HG-DAgger achieve the highest final success rate on two tasks.