BodyGen: Advancing Towards Efficient Embodiment Co-Design

TL;DR

BodyGen tackles the inefficiency of embodiment co-design by integrating topology-aware transformer design (MoSAT) with a topology-aware morphology encoding (TopoPE) and a temporal credit assignment mechanism to balance design and control rewards. It introduces an attention-based co-design network that enables centralized message passing across evolving morphologies while maintaining a lightweight model footprint. Experimental results across ten MuJoCo tasks show an average improvement of 60.03% over state-of-the-art baselines and faster convergence, validating the approach. Limitations include simulation-only validation and the need for real-world transfer with perception and execution integration; future work will address these aspects.

Abstract

Embodiment co-design aims to optimize a robot's morphology and control policy simultaneously. While prior work has demonstrated its potential for generating environment-adaptive robots, this field still faces persistent challenges in optimization efficiency due to the (i) combinatorial nature of morphological search spaces and (ii) intricate dependencies between morphology and control. We prove that the ineffective morphology representation and unbalanced reward signals between the design and control stages are key obstacles to efficiency. To advance towards efficient embodiment co-design, we propose BodyGen, which utilizes (1) topology-aware self-attention for both design and control, enabling efficient morphology representation with lightweight model sizes; (2) a temporal credit assignment mechanism that ensures balanced reward signals for optimization. With our findings, Body achieves an average 60.03% performance improvement against state-of-the-art baselines. We provide codes and more results on the website: https://genesisorigin.github.io.

Figures (14)

• Figure 1: Embodied Agents generated by BodyGen.
• Figure 2: Overview of BodyGen, which leverages an RL-based framework for joint evolving of morphology and control policy, and an attention-based network equipped with Topology Position Encoding (TopoPE) for centralized message processing.
• Figure 3: The Morphology Self-Attention (MoSAT) architecture. (a) The sensor observations from different limbs are projected to hidden tokens for centralized processing with several MoSAT blocks and generate separate actions. (b) The MoSAT network processes different morphologies in a batch manner and learns a universal control policy $\pi(\cdot|\mathcal{G})$, thus improving training efficiency.
• Figure 4: The motivation of our proposed topology-aware position encoding TopoPE. (a) During the co-design procedure, the agent's morphology keeps changing. (b) A typical traversal-based PE in previous works resulted in inconsistency across mythologies. (c) TopoPE can better adapt to similar morphology structures using a reasonably alignable manner.
• Figure 5: BodyGen leverages an actor-critic paradigm for policy optimization.