Research map / Reinforcement learning / Visual and representation RL

Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot

NeurIPS 2026 · Reinforcement learning › Visual and representation RL · approach: Dexterous manipulation transfer · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Position paper arguing that humanoid robots should co-design their control policies and physical morphology via bi-level optimization, and outlining feasibility strategies and open questions.

Problem

Humanoid robot development relies on manually engineered, fixed body structures and optimizes only control policies, leaving co-design of morphology and control largely unexplored for general-purpose humanoids.

Approach

Formulates humanoid body-control co-design as a bi-level optimization (inner RL policy learning, outer morphology search with cost constraints) and proposes strategic exploration, Sim2Real evaluation, and goal-conditioned meta-policies as paths to feasibility, plus arguments for necessity.

Key results

• Position paper with no new experiments; summarizes around 27 prior robotic co-design studies in a table, mostly on quadrupeds and modular robots with simple parameters.
• Argues that co-design can raise the optimum over design-policy pairs, since max over design and policy is at least max over policy for any fixed design.

Limitations

• No empirical experiments validating the proposed co-design methodology on humanoids
• Feasibility arguments rely on prior work from simpler robots rather than humanoid-specific demonstrations
• Proposed solutions (WFMs, MoE reasoning co-optimization) are acknowledged as long-term and unsolved

Experimental setup

Builds on: reinforcement learning bayesian optimization decision transformer diffusion policy

Closest papers by meaning

• BodyGen: Advancing Towards Efficient Embodiment Co-Design (ICLR 2025 · similarity 0.90)
• Hierarchical World Models as Visual Whole-Body Humanoid Controllers (ICLR 2025 · similarity 0.87)
• Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.87)
• Debate2Create: Robot Co-design via Multi-Agent LLM Debate (ICML 2026 · similarity 0.87)
• Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.86)
• DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.85)