Timezone: »

 
Spotlight
Keyframe-Focused Visual Imitation Learning
Chuan Wen · Jierui Lin · Jianing Qian · Yang Gao · Dinesh Jayaraman

Tue Jul 20 06:25 PM -- 06:30 PM (PDT) @
in Reinforcement Learning 7 »

Imitation learning trains control policies by mimicking pre-recorded expert demonstrations. In partially observable settings, imitation policies must rely on observation histories, but many seemingly paradoxical results show better performance for policies that only access the most recent observation. Recent solutions ranging from causal graph learning to deep information bottlenecks have shown promising results, but failed to scale to realistic settings such as visual imitation. We propose a solution that outperforms these prior approaches by upweighting demonstration keyframes corresponding to expert action changepoints. This simple approach easily scales to complex visual imitation settings. Our experimental results demonstrate consistent performance improvements over all baselines on image-based Gym MuJoCo continuous control tasks. Finally, on the CARLA photorealistic vision-based urban driving simulator, we resolve a long-standing issue in behavioral cloning for driving by demonstrating effective imitation from observation histories. Supplementary materials and code at: \url{https://tinyurl.com/imitation-keyframes}.

Keywords: Reinforcement Learning and Planning · Deep RL

Author Information

Chuan Wen (Tsinghua University)
Jierui Lin (UT Austin)
Jianing Qian (University of Pennsylvania)
Yang Gao (Tsinghua University)
Dinesh Jayaraman (University of Pennsylvania)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors