Search papers, labs, and topics across Lattice.
This paper presents MinInter, a novel trajectory selection method that minimizes interpolation during data augmentation for imitation learning. By selecting source demonstrations that require the least interpolation for each sampled initial configuration, MinInter enhances the quality of synthetic demonstrations, leading to improved success rates in both data generation and policy execution across various manipulation tasks. Experimental results demonstrate that MinInter outperforms the SkillGen framework, particularly in challenging scenarios characterized by high variance and long-horizon tasks.
Minimizing interpolation during data generation can significantly boost the effectiveness of imitation learning, yielding higher success rates in complex manipulation tasks.
Imitation learning enables robots to acquire complex manipulation skills from demonstrations, but its effectiveness is limited by the cost of collecting high-quality data. Trajectory-level data augmentation methods alleviate this challenge by recombining expert demonstrations under varied initial states. However, such methods typically insert interpolations or other non-expert transition segments between disjoint parts, and such non-expert segments could reduce the quality of the generated data. This paper introduces Minimizing Interpolation (MinInter), an effective trajectory selection method that, for each sampled initial configuration, chooses the source demonstration requiring the least interpolation to form a complete trajectory. By explicitly minimizing interpolations during data generation, MinInter produces higher-quality synthetic demonstrations while remaining compatible with existing data generation frameworks. Experiments on 12 manipulation tasks with 26 variants from the MimicGen benchmark show that MinInter consistently improves both data generation success rates and policy success rates, with the largest gains on contact-rich, long-horizon and high-variance settings. Compared to the recent SkillGen framework, MinInter achieves higher policy success rates despite its conceptual simplicity, underscoring the value of interpolation minimization for data augmentation.
