Efficient Robotic Policy Learning via Latent Space Backward Planning

TL;DR

This paper introduces a Latent Space Backward Planning (LBP) method for robotic policy learning that improves efficiency and accuracy in long-horizon tasks by grounding goals in latent space and recursively predicting subgoals.

Contribution

The paper proposes a novel LBP scheme that predicts intermediate subgoals in latent space, enabling more efficient and accurate long-term robotic planning compared to existing methods.

Findings

LBP outperforms existing planning methods in simulation and real-robot experiments.

LBP achieves state-of-the-art performance in long-horizon tasks.

Latent space subgoal prediction reduces computational costs and error accumulation.

Abstract

Current robotic planning methods often rely on predicting multi-frame images with full pixel details. While this fine-grained approach can serve as a generic world model, it introduces two significant challenges for downstream policy learning: substantial computational costs that hinder real-time deployment, and accumulated inaccuracies that can mislead action extraction. Planning with coarse-grained subgoals partially alleviates efficiency issues. However, their forward planning schemes can still result in off-task predictions due to accumulation errors, leading to misalignment with long-term goals. This raises a critical question: Can robotic planning be both efficient and accurate enough for real-time control in long-horizon, multi-stage tasks? To address this, we propose a Latent Space Backward Planning scheme (LBP), which begins by grounding the task into final latent goals, followed by recursively predicting intermediate subgoals closer to the current state. The grounded final goal enables backward subgoal planning to always remain aware of task completion, facilitating on-task prediction along the entire planning horizon. The subgoal-conditioned policy incorporates a learnable token to summarize the subgoal sequences and determines how each subgoal guides action extraction. Through extensive simulation and real-robot long-horizon experiments, we show that LBP outperforms existing fine-grained and forward planning methods, achieving SOTA performance. Project Page: https://lbp-authors.github.io