EgoPush: Learning End-to-End Egocentric Multi-Object Rearrangement for Mobile Robots

TL;DR

EgoPush is a perception-driven reinforcement learning framework that enables mobile robots to perform multi-object rearrangement in cluttered environments using egocentric vision, without relying on global state estimation.

Contribution

We propose EgoPush, a novel object-centric latent space and active perception approach for egocentric, end-to-end learning of multi-object rearrangement on mobile robots.

Findings

EgoPush outperforms RL baselines in success rate in simulation.

EgoPush achieves zero-shot sim-to-real transfer on a mobile robot.

Design choices in EgoPush are validated through ablation studies.

Abstract

Humans can rearrange objects in cluttered environments using egocentric perception, navigating occlusions without global coordinates. Inspired by this capability, we study long-horizon multi-object non-prehensile rearrangement for mobile robots using a single egocentric camera. We introduce EgoPush, a policy learning framework that enables egocentric, perception-driven rearrangement without relying on explicit global state estimation that often fails in dynamic scenes. EgoPush designs an object-centric latent space to encode relative spatial relations among objects, rather than absolute poses. This design enables a privileged reinforcement-learning (RL) teacher to jointly learn latent states and mobile actions from sparse keypoints, which is then distilled into a purely visual student policy. To reduce the supervision gap between the omniscient teacher and the partially observed student, we restrict the teacher's observations to visually accessible cues. This induces active perception behaviors that are recoverable from the student's viewpoint. To address long-horizon credit assignment, we decompose rearrangement into stage-level subproblems using temporally decayed, stage-local completion rewards. Extensive simulation experiments demonstrate that EgoPush significantly outperforms end-to-end RL baselines in success rate, with ablation studies validating each design choice. We further demonstrate zero-shot sim-to-real transfer on a mobile platform in the real world. Code and videos are available at https://ai4ce.github.io/EgoPush/.