Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision
For those working on robotics or AI, this research offers a practical solution for creating more efficient and effective robotic memory systems.
This paper introduces a new approach for robotic memory called the workspace token. It uses a VLM during training to identify task-relevant information and distills it into a lightweight latent memory. This memory can be used as a drop-in replacement for observations at deployment, solving memory-intensive tasks without needing VLM reasoning in real-time. The method was tested in both simulation and on hardware, showing improved performance and efficiency.
Complex robotic manipulation tasks frequently require a long-term memory of past events and actions. As conditioning on full histories renders policies prone to spurious correlations and degrades performance, many approaches to policy memory involve compressing historical information through expensive VLM queries in-the-loop to process only task-salient information. In this paper, we propose an alternative approach in which computationally intensive VLM queries are made during train-time to learn a lightweight latent memory that can be efficiently queried at deployment time. Our representation, which we call the \textbf{workspace token}, is trained by (1) using a VLM to identify current and historical information necessary for completing a task, then (2) distilling these into the workspace token using a set-reconstruction decoder loss. In both simulation and hardware, we show that the workspace token can be used as a drop-in replacement for observations during deployment, enabling policies to solve memory-intensive tasks without the need for VLM reasoning in-the-loop. Interestingly, we found that workspace tokens are not only more lightweight but also lead to better policy performance.