Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning Human-Robot Collaboration via Heterogeneous-Agent Lyapunov Policy Optimization
Hao Zhang, Yaru Niu, Yikai Wang, Ding Zhao, Eric H. Tseng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Lyapunov projection (HALO) over HAPPO | HAPPO · OSP/SCT/SLH average success | success rate: OSP 78.0 -> 87.2 | Improved |
| add Lyapunov projection (HALO) over PCGrad | PCGrad · OSP average | success rate: 83.0 -> 87.2 | Improved |
| add HALO | HAPPO · rationality gap at 2B steps | gap V: 4.89 -> 0.09 | Improved |
| add HALO | HAPPO · gradient conflict rate | GCR: 72.5% -> 4.2% | Improved |
| add HALO | PCGrad · convergence step | steps to plateau: 1.2B vs 1.3B | No clear effect |
| use soft Lagrangian penalty on V instead of hard projection (ablation) | HAPPO · SLH-extreme (pivoting) | success rate: 74.9 -> 76.5 | No clear effect |
| add static Lyapunov projection (ablation) | HAPPO · SLH-extreme (pivoting) | success rate: 74.9 -> 79.5 | Improved |
| add adaptive step scheduling (ablation) | static projection · SLH-extreme (pivoting) | success rate: 79.5 -> 80.8 | Improved |
| add alignment rectification (ablation) | HALO · SLH-extreme (pivoting) | success rate: 80.8 -> 82.3 | Improved |
| add alignment rectification (ablation) | HALO · SLH-extreme (pivoting) | gap V: 0.24 -> 0.09 | Improved |
| add HALO to real-world SCT | PCGrad · real-world SCT | time to destination: 81.5 -> 76.2 | Improved |
| learned MARL humanoid vs robot-script | Robot-Script · real-world SCT | object drop rate: 60% -> 0% | Hurt |
Closest papers by meaning
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Adaptive Collaboration with Humans: Metacognitive Policy Optimization for Multi-Agent LLMs with Continual Learning (ICLR 2026 · similarity 0.87)
- Beyond Single Stationary Policies: Meta-Task Players as Naturally Superior Collaborators (NeurIPS 2024 · similarity 0.87)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.87)
- Cooperative Policy Agreement: Learning Diverse Policy for Offline MARL (AAAI 2025 · similarity 0.87)
- Learning Joint Behaviors with Large Variations (AAAI 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
