<ul data-eligibleForWebStory="true"><li>Interactive Imitation Learning (IIL) enables agents to learn behaviors with human interventions, but this can be demanding for supervisors.</li><li>Proposed Adaptive Intervention Mechanism (AIM) in robot-gated IIL to reduce cognitive load on supervisors.</li><li>AIM uses a proxy Q-function to determine when to request human demonstrations based on agent's alignment with human actions.</li><li>Proxy Q-function assigns high values for deviations and decreases as agent's performance improves, allowing real-time assessment.</li><li>Expert-in-the-loop experiments show AIM reduces expert monitoring in continuous and discrete control tasks.</li><li>AIM outperforms Thrifty-DAgger by 40% in terms of human take-over cost and learning efficiency.</li><li>AIM identifies safety-critical states for expert intervention, leading to better quality demonstrations and reduced expert interaction.</li><li>Code and demo video for AIM available at https://github.com/metadriverse/AIM.</li></ul>

Robot-Gated Interactive Imitation Learning with Adaptive Intervention Mechanism

Discover more