Research map / Reinforcement learning / Visual and representation RL
Low-Rank Modular Reinforcement Learning via Muscle Synergy
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use Solar instead of MetaMorph | UNIMALS zero-shot locomotion · Armature, density, damping, gear, module parameter, and joint angle variants | performance: 1253.39 ± 28.25 vs 843.31 ± 14.80; 1654.43 ± 24.55 vs 1089.64 ± 22.64; 1663.64 ± 26.21 vs 1113.04 ± 10.18; 1480.45 ± 30. | Helped |
| Use Solar instead of MetaMorph | UNIMALS zero-shot manipulation · Armature, density, damping, gear, and module parameter variants | performance: 792.69 ± 32.35 vs 685.36 ± 9.03; 943.07 ± 57.46 vs 823.20 ± 18.95; 984.19 ± 24.14 vs 860.70 ± 11.24; 862.04 ± 36.90 vs 7 | Helped |
| Use Solar instead of MetaMorph | UNIMALS zero-shot manipulation · Joint angle variant | performance: 314.16 ± 4.95 vs 455.49 ± 11.28 | Hurt |
| Add synergy-based masks to the first attention layer (ablation) | Amorpheus on multi-task morphologies · Walker++, Humanoid++, and Hopper++ | episode reward | Helped |
| Remove the value-derived preference vector from affinity propagation (ablation) | Solar on multi-task morphologies · Walker++, Humanoid++, and Hopper++ | episode reward | Hurt |
| Use synergy-based low-rank control and synergy-aware policy learning | Modular RL · Humanoid++ multi-task and zero-shot tasks | episode reward | Helped |
| Use synergy-based low-rank control | Single-task UNIMALS locomotion · Three sampled UNIMALS morphologies | episode reward | Helped |
Closest papers by meaning
- One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control (ICML 2020 · similarity 0.90)
- Explore to Learn: Latent Exploration Through Disentangled Synergy Patterns for Reinforcement Learning in Overactuated Control (AAAI 2026 · similarity 0.87)
- Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularity (NeurIPS 2019 · similarity 0.87)
- High-Dimensional Robotic Reinforcement Learning with Developing Synergies (NeurIPS 2026 · similarity 0.86)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.86)
- MeMo: Meaningful, Modular Controllers via Noise Injection (NeurIPS 2024 · similarity 0.85)
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.