SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World

TL;DR

This paper introduces SPOC, a transformer-based approach that imitates shortest-path planners in simulation to train embodied agents capable of effective navigation and manipulation in real-world environments using only RGB sensors.

Contribution

The paper presents a novel imitation learning method that leverages shortest-path planning in simulation, enabling scalable training of embodied agents for real-world tasks.

Findings

Agents can navigate and manipulate objects in real environments using only RGB sensors.

Imitating shortest-path planners in simulation is effective for training real-world embodied agents.

The approach scales to diverse environments with extensive training data.

Abstract

Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too slow and ineffective for long-horizon tasks. While IL with human supervision is effective, collecting human trajectories at scale is extremely expensive. In this work, we show that imitating shortest-path planners in simulation produces agents that, given a language instruction, can proficiently navigate, explore, and manipulate objects in both simulation and in the real world using only RGB sensors (no depth map or GPS coordinates). This surprising result is enabled by our end-to-end, transformer-based, SPOC architecture, powerful visual encoders paired with extensive image augmentation, and the dramatic scale and diversity of our training data: millions of frames of shortest-path-expert trajectories collected inside approximately 200,000 procedurally generated houses containing 40,000 unique 3D assets. Our models, data, training code, and newly proposed 10-task benchmarking suite CHORES are available in https://spoc-robot.github.io.