Multi agent transformers enable coordinated control of soft robot arrays
Distributed Dexterous Manipulation with Spatially Conditioned Multi-Agent Transformers
RoboticsMachine Learning
Summary
Controlling lots of small robots working together to move objects is really hard because there are many ways to move and they must cooperate closely. The authors created a method using multi-agent transformers that takes into account each robot's position to efficiently learn good control strategies. Their method lets the robots improve their actions step-by-step and balances using fewer robots while still handling objects well. They tested this in both computer simulations and real life, showing accurate object manipulation even while reducing wear on the robots.
What this means in practice
- •For robotics engineers: Develop control software for arrays of soft robots that efficiently manipulate objects with fewer robots and reduced physical wear.
- •For automation integrators: Integrate flexible multi-robot systems in manufacturing settings to handle irregularly shaped parts with improved coordination.
Authors
Sarvesh Patil
Abstract
Distributed Dexterous Manipulation (DDM) is a novel paradigm that presents significant control challenges due to high action-space redundancy, inter-robot cooperation, and dynamic object-robot interactions. This paper introduces a framework based on spatially conditioned Multi-Agent Transformers (MATs) to efficiently learn robust control policies for a DDM system grounded in an array of 64 soft delta robots arranged in an 8x8 grid. Our three core contributions are: (i) an MAT with adaptive layer norm for compute efficiency, (ii) spatial contrastive embeddings to ground transformer embeddings in the spatial configuration of the robots, and (iii) an MAT-based behavior cloning method fine-tuned using Soft Actor Critic. We also propose an action selection formulation to analyze the trade-off between task performance and the number of robots utilized. Our experiments show that MATs iteratively refine their actions through the stacked attention blocks. This further informs the benefit of spatial conditioning in transformers to learn DDM policies. We demonstrate long-horizon planar manipulation tasks with objects of various geometries in simulation and real-world. Finally, we show how action selection mitigates robot maintenance by reducing wear and tear due to inter-robot collisions while maintaining the ability to manipulate objects along various trajectories in the real-world, achieving an average error of ~1.5 cm, while using ~65% fewer robots.