Research map / Reinforcement learning / Policy gradient and actor-critic
Combining Model-Based and Model-Free Updates for Trajectory-Centric Reinforcement Learning
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 |
|---|---|---|---|
| Combine an LQR-FLM update on the fitted cost approximation with a PI^2 update on residual costs | TVLG policies on gripper pusher and door opening · Simulated gripper pusher and door opening | task success | Helped |
| Combine model-based and residual model-free updates | Real-robot hockey policy · Hockey | puck enters goal | Helped |
| Combine model-based and residual model-free updates | Real-robot power plug policy · Power plug plugging | successful insertion: 60% -> every rollout | Helped |
| Use MDGPS with PILQR local policies to train a neural network | Neural network policy · Hockey, two training goals and one new goal | success rate: 90% | Helped |
| Use MDGPS with PILQR instead of direct model-free deep RL | Neural network policy · Reacher | samples for approximately the same final distance: DDPG: 25 times more samples; TRPO: 150 times more samples | Helped |
| Use MDGPS with PILQR instead of TRPO | Neural network policy · Door opening | samples to learn a successful policy: 20 times more samples | Helped |
| Use MDGPS with PILQR rather than MDGPS with PI^2 | Neural network policy · Reacher | final distance to target | Helped |
| Increase PI^2 sampling by 10 times | PI^2 · Gripper pusher | task performance: 10 times more samples | Helped |
| Increase PI^2 sampling by 10 times | PI^2 · Door opening | task success: 10 times more samples | No clear effect |
Closest papers by meaning
- ASID: Active Exploration for System Identification in Robotic Manipulation (ICLR 2024 · similarity 0.87)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.87)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.87)
- Diffusion Policy Policy Optimization (ICLR 2025 · similarity 0.86)
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 2025 · similarity 0.86)
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.