End-to-End Egospheric Spatial Memory

TL;DR

This paper introduces Egospheric Spatial Memory (ESM), a novel, parameter-free, egocentric spatial memory module that enhances autonomous agents' spatial reasoning and task performance through expressive 3D representations and end-to-end training.

Contribution

The paper presents ESM, a new egocentric spatial memory module that improves training efficiency and performance for visuomotor tasks and enables seamless integration with other modalities.

Findings

ESM outperforms existing memory baselines on drone and manipulator tasks.

ESM enables effective semantic segmentation on ScanNet dataset.

The module bridges real-time mapping and differentiable memory architectures.

Abstract

Spatial memory, or the ability to remember and recall specific locations and objects, is central to autonomous agents' ability to carry out tasks in real environments. However, most existing artificial memory modules are not very adept at storing spatial information. We propose a parameter-free module, Egospheric Spatial Memory (ESM), which encodes the memory in an ego-sphere around the agent, enabling expressive 3D representations. ESM can be trained end-to-end via either imitation or reinforcement learning, and improves both training efficiency and final performance against other memory baselines on both drone and manipulator visuomotor control tasks. The explicit egocentric geometry also enables us to seamlessly combine the learned controller with other non-learned modalities, such as local obstacle avoidance. We further show applications to semantic segmentation on the ScanNet dataset, where ESM naturally combines image-level and map-level inference modalities. Through our broad set of experiments, we show that ESM provides a general computation graph for embodied spatial reasoning, and the module forms a bridge between real-time mapping systems and differentiable memory architectures. Implementation at: https://github.com/ivy-dl/memory.