IntervenGen: Interventional Data Generation for Robust and Data-Efficient Robot Imitation Learning
This addresses the burden on human operators in interactive imitation learning for robotics, offering a data-efficient solution to improve policy robustness, though it is incremental as it builds on existing interactive methods like DAgger.
The paper tackles the problem of distribution shift in robot imitation learning by proposing IntervenGen, a system that autonomously generates corrective interventions from a small number of human inputs, resulting in up to a 39x increase in policy robustness with only 10 human interventions.
Imitation learning is a promising paradigm for training robot control policies, but these policies can suffer from distribution shift, where the conditions at evaluation time differ from those in the training data. A popular approach for increasing policy robustness to distribution shift is interactive imitation learning (i.e., DAgger and variants), where a human operator provides corrective interventions during policy rollouts. However, collecting a sufficient amount of interventions to cover the distribution of policy mistakes can be burdensome for human operators. We propose IntervenGen (I-Gen), a novel data generation system that can autonomously produce a large set of corrective interventions with rich coverage of the state space from a small number of human interventions. We apply I-Gen to 4 simulated environments and 1 physical environment with object pose estimation error and show that it can increase policy robustness by up to 39x with only 10 human interventions. Videos and more results are available at https://sites.google.com/view/intervengen2024.