Learning to Walk in Costume: Adversarial Motion Priors for Aesthetically Constrained Humanoids

2025-09-05

AI Summary

Aesthetic-driven entertainment humanoid robots—characterized by extreme physical attributes (e.g., head-to-body mass ratio up to 16%, highly asymmetric mass distribution, and severely restricted joint ranges)—struggle to achieve natural, stable bipedal locomotion. Method: This paper proposes a reinforcement learning control framework integrating Adversarial Motion Priors (AMP), customized domain randomization, and sim-to-real transfer. A task-specific reward function is designed to jointly optimize dynamic balance and aesthetic constraints, mitigating the adverse impact of high artistic design costs on motion control. Contribution/Results: To our knowledge, this is the first work achieving stable standing and human-like walking for the Cosmo robot under such extreme physical constraints. The learned policy successfully transfers from simulation to the real platform. Experiments demonstrate strong adaptability to atypical humanoid morphologies, establishing a novel paradigm for co-optimizing aesthetics and functionality in humanoid locomotion control.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Adversarial Agents

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

We present a Reinforcement Learning (RL)-based locomotion system for Cosmo, a custom-built humanoid robot designed for entertainment applications. Unlike traditional humanoids, entertainment robots present unique challenges due to aesthetic-driven design choices. Cosmo embodies these with a disproportionately large head (16% of total mass), limited sensing, and protective shells that considerably restrict movement. To address these challenges, we apply Adversarial Motion Priors (AMP) to enable the robot to learn natural-looking movements while maintaining physical stability. We develop tailored domain randomization techniques and specialized reward structures to ensure safe sim-to-real, protecting valuable hardware components during deployment. Our experiments demonstrate that AMP generates stable standing and walking behaviors despite Cosmo's extreme mass distribution and movement constraints. These results establish a promising direction for robots that balance aesthetic appeal with functional performance, suggesting that learning-based methods can effectively adapt to aesthetic-driven design constraints.

Problem

• Learning natural locomotion under aesthetic design constraints
• Addressing extreme mass distribution and movement restrictions
• Ensuring safe sim-to-real transfer for entertainment humanoids

Innovation

• Adversarial Motion Priors for natural movement learning
• Tailored domain randomization ensuring safe sim-to-real
• Specialized reward structures protecting hardware components