Improving Robotic Imitation Learning via Trajectory Standardization

2026-06-22

AI Summary

Imitation learning in robotics is often hindered by high noise levels and non-uniform temporal sampling in human demonstration trajectories, leading to inefficient policy learning. This work proposes Information-aware Standardized Resampling (ISR), a novel approach that integrates information intensity fields with Riemannian geometry. By constructing an information metric based on the norms of velocity and acceleration, ISR performs offline trajectory resampling via geodesic equidistant parameterization, ensuring approximately equal information distance between consecutive points. This effectively filters out redundant micro-movements while preserving high-curvature and fine-manipulation segments. Evaluated on three real-world manipulation tasks, ISR improves task success rates by approximately 25% compared to conventional time-uniform downsampling, while simultaneously reducing dataset size and lowering training costs.

Abstract

Imitation learning for robotic manipulation relies on large sets of human demonstration trajectories, which are often noisy and temporally irregular due to variable operator speed, intermittent pauses, and inconsistent action density. A common preprocessing strategy is time-uniform downsampling to shorten sequences, but it cannot effectively remove speed-induced non-uniformity or redundant pauses. This mismatch degrades data quality and hinders policy learning. To address this issue, we propose Information-Standardized Trajectory Resampling (ISR), an offline preprocessing method for effective imitation learning. ISR resamples each trajectory by enforcing approximately equal information distance between adjacent points. Specifically, we map trajectories onto an information-modulated Riemannian manifold and perform geodesic-equidistant parameterization. We construct an information-intensity field from velocity and acceleration norms: the velocity term removes small-motion redundancy, while the acceleration term preserves high-curvature and fine-manipulation phases. We evaluate ISR on three real-world manipulation tasks with mainstream imitation learning policies. Compared with the baseline time-uniform 3x downsampling, ISR improves task success rates by about 25%, remains robust across datasets collected from different operators, and reduces both dataset size and training cost. The code and videos are publicly available at https://d-robotics-ai-lab.github.io/isr.page.

Problem

• imitation learning
• trajectory standardization
• temporal irregularity
• data preprocessing
• robotic manipulation

Innovation

• trajectory standardization
• imitation learning
• information-intensity field
• Riemannian manifold
• geodesic-equidistant resampling