Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularity

TL;DR

This paper introduces a modular co-evolution approach where primitive agents self-assemble into complex morphologies and learn to control them, demonstrating improved generalization in simulation over static baselines.

Contribution

It presents a novel method for agents to dynamically self-assemble and learn control policies, enabling better adaptability and generalization to environmental and structural changes.

Findings

Enhanced generalization to environment changes

Effective self-assembly of complex morphologies

Superior performance over static baselines

Abstract

Contemporary sensorimotor learning approaches typically start with an existing complex agent (e.g., a robotic arm), which they learn to control. In contrast, this paper investigates a modular co-evolution strategy: a collection of primitive agents learns to dynamically self-assemble into composite bodies while also learning to coordinate their behavior to control these bodies. Each primitive agent consists of a limb with a motor attached at one end. Limbs may choose to link up to form collectives. When a limb initiates a link-up action, and there is another limb nearby, the latter is magnetically connected to the 'parent' limb's motor. This forms a new single agent, which may further link with other agents. In this way, complex morphologies can emerge, controlled by a policy whose architecture is in explicit correspondence with the morphology. We evaluate the performance of these dynamic and modular agents in simulated environments. We demonstrate better generalization to test-time changes both in the environment, as well as in the structure of the agent, compared to static and monolithic baselines. Project video and code are available at https://pathak22.github.io/modular-assemblies/