Research map / Reinforcement learning / Policy gradient and actor-critic
Reward-Adaptive Reinforcement Learning: Dynamic Policy Gradient Optimization for Bipedal Locomotion
Changxin Huang, Guangrun Wang, Zhibo Zhou, Ronghui Zhang, Liang Lin
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 multi-head critic (separate value per reward component) (ablation) | DDPG · Gazebo biped training | episodes until reward rises: 6k -> 4k episodes | Improved |
| add dynamic priority weights to multi-head critic (ablation) | MHDDPG · Gazebo biped training | episodes until reward rises: 4k -> 2k episodes | Improved |
| add multi-head critic (ablation) | DDPG · Walker2d-v3 | mean return: 2398.5 -> 3162.0 | Improved |
| add dynamic weight (ablation) | MHDDPG · Walker2d-v3 | mean return: 3162.0 -> 3217.5 | No clear effect |
| add multi-head critic (ablation) | DDPG · Hopper-v3 | mean return: 3476.5 -> 3539.3 | Improved |
| add dynamic weight (ablation) | MHDDPG · Hopper-v3 | mean return: 3539.3 -> 3600.4 | Improved |
| add multi-head critic (ablation) | DDPG · HalfCheetah-v3 | mean return: 12422.9 -> 12748.9 | Improved |
| add dynamic weight (ablation) | MHDDPG · HalfCheetah-v3 | mean return: 12748.9 -> 12895.3 | Improved |
| add hybrid dynamic policy gradient | PPO · Hopper-v3 | mean return: 2609.3 -> 3263.5 | Improved |
| add hybrid dynamic policy gradient | PPO · Walker2d-v3 | mean return: 3588.5 -> 4688.0 | Improved |
| add hybrid dynamic policy gradient | PPO · HalfCheetah-v3 | mean return: 5783.9 -> 6165.2 | Improved |
| add multi-head critic (ablation) | DDPG · Gazebo slope walking | success rate | No clear effect |
Closest papers by meaning
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.87)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · similarity 0.87)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.86)
- Time-Efficient Reinforcement Learning with Stochastic Stateful Policies (ICLR 2024 · similarity 0.86)
- Diffusion Policy Policy Optimization (ICLR 2025 · similarity 0.86)
- To the Max: Reinventing Reward in Reinforcement Learning (ICML 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.
