Difference-Aware Retrieval Policies for Imitation Learning
2026-06-08 • Robotics
RoboticsArtificial IntelligenceMachine Learning
AI summaryⓘ
The authors explain that traditional imitation learning methods, which copy expert behavior, often fail when facing new, unseen situations. To fix this, they propose DARP, a method that looks at similar past situations and their actions instead of trying to learn one global rule. DARP uses nearby examples from expert data to decide what action to take, improving reliability without extra expert help or data. Their tests show that DARP performs significantly better than usual behavior cloning in different tasks like robot control and using visual information.
Imitation LearningBehavior CloningOut-of-Distribution GeneralizationSemi-Parametric Methodsk-Nearest NeighborsRobotic ManipulationContinuous ControlRetrieval-Based MethodsLocal Neighborhood StructureState-to-Action Mapping
Authors
Quinn Pfeifer, Ethan Pronovost, Paarth Shah, Khimya Khetarpal, Siddhartha Srinivasa, Abhishek Gupta
Abstract
Parametric imitation learning via behavior cloning can suffer from poor generalization to out-of-distribution states due to compounding errors during deployment. We show that reusing the training data during inference via a semi-parametric retrieval-based imitation learning approach can alleviate this challenge. We present Difference-Aware Retrieval Policies for Imitation Learning (DARP), a semi-parametric retrieval-based imitation learning approach that addresses this limitation by reparameterizing the imitation learning problem in terms of local neighborhood structure rather than direct state-to-action mappings. Instead of learning a global policy, DARP trains a model to predict actions based on $k$-nearest neighbors from expert demonstrations, their corresponding actions, and the relative distance vectors between neighbor states and query states. DARP requires no additional assumptions beyond those made for standard behavior cloning -- it does not require additional data collection, online expert feedback, or task-specific knowledge. We demonstrate consistent performance improvements of 15-46% over standard behavior cloning across diverse domains, including continuous control and robotic manipulation, and across different representations, including high-dimensional visual features. Code and demos are available at https://weirdlabuw.github.io/darp-site/.