MAPO: Mixture of Experts with Alternating Policy Optimization for Universal Humanoid Motion Tracking
MAPO: Mixture of Experts with Alternating Policy Optimization for Universal Humanoid Motion Tracking
This research paper proposes MAPO, a method combining mixture-of-experts policies with alternating optimization to address divergence misestimation issues. The authors claim that MAPO successfully tracks 182 motions where PPO with MoE fails, achieving the highest tracking success rate and lowest error among compared methods.
The paper introduces MAPO, a Mixture of Experts with Alternating Policy Optimization method for universal humanoid motion tracking, claiming superior performance over PPO with MoE on 182 motions as demonstrated in the provided abstract.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
trajectories
182
View original evidence
s the router and the experts to avoid the divergence misestimation. MAPO consistently tracks 182 motions that PPO with MoE fails to track, whereas PPO with MoE tracks only one motion that MAPO fails to track. MAPO also attains the highest tracking success rate among the compared methods and the lowest tracking error across most error metrics, demonstrating its superior performance. Videos can be found on the project
Open source S4robots
1
View original evidence
s the router and the experts to avoid the divergence misestimation. MAPO consistently tracks 182 motions that PPO with MoE fails to track, whereas PPO with MoE tracks only one motion that MAPO fails to track. MAPO also attains the highest tracking success rate among the compared methods and the lowest tracking error across most error metrics, demonstrating its superior performance. Videos can be found on the project
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
s the router and the experts to avoid the divergence misestimation. MAPO consistently tracks 182 motions that PPO with MoE fails to track, whereas PPO with MoE tracks only one motion that MAPO fails to track. MAPO also attains the highest tracking success rate among the compared methods and the lowest tracking error across most error metrics, demonstrating its superior performance. Videos can be found on the project
Open source S4
arXiv:2611.02613v1 Announce Type: new Abstract: Universal humanoid motion tracking is essential for enabling humanoid robots to perform diverse tasks in human environments. To capture the wide diversity of human motions, mixture-of-experts (MoE) policies have been widely adopted, leveraging dynamic routing to enhance policy expressiveness. However, we identify that divergence between old and updated policies can be m
Open source S2
isestimated during policy-gradient updates of MoE policies, due to their dynamic routing. Because most policy-gradient algorithms regulate policy updates based on the divergence, this misestimation can lead to either overly conservative or excessively large updates, which can yield a suboptimal policy. In this paper, we propose Mixture of Experts with Alternating Policy Optimization (MAPO), which alternately optimize
Open source S3
website: https://mapo-corl2026.github.io/MAPO/
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org