Research map / Reinforcement learning / Policy gradient and actor-critic
Temporal Difference Learning for Model Predictive Control
Nicklas A Hansen, Hao Su, Xiaolong Wang
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 |
|---|---|---|---|
| Use TD-MPC instead of DrQ | DMControl image-based control · Finger Spin, DMControl 100k | return: 901 \scriptstyle{\pm 104} -> 943 \scriptstyle{\pm 59} | Improved |
| Use TD-MPC instead of DrQ | DMControl image-based control · Walker Walk, DMControl 100k | return: 612 \scriptstyle{\pm 164} -> 577 \scriptstyle{\pm 208} | Hurt |
| Use TD-MPC instead of DrQ | DMControl image-based control · Cheetah Run, DMControl 100k | return: 344 \scriptstyle{\pm 67} -> 222 \scriptstyle{\pm 88} | Hurt |
| Add latent state consistency loss rather than using no regularization (ablation) | TD-MPC · DMControl, 15 state-based tasks | return | Improved |
| Use latent state consistency rather than reconstruction or contrastive regularization (ablation) | TD-MPC · DMControl, 15 state-based tasks | return | Mixed |
| Add latent state consistency loss (ablation) | TD-MPC · DMControl state-based tasks | return | Improved |
| Learn a representation instead of using an identity mapping and state-space prediction (ablation) | TD-MPC · DMControl state-based tasks | return | Mixed |
| Add latent state consistency loss (ablation) | TD-MPC · DMControl state-based tasks | return | Mixed |
| Use TD-MPC instead of SAC | Walker Walk · Walker Walk | time to solve (h): $0.41$ -> $0.47$ | Hurt |
| Use TD-MPC instead of SAC | Humanoid Stand · Humanoid Stand | time to solve (h): $9.31$ -> $9.39$ | Hurt |
| Increase planning computational budget | TD-MPC · Quadruped Walk | return | Improved |
| Reduce planning iterations during inference | TD-MPC · DMControl tasks | return: 50% | No clear effect |
Closest papers by meaning
- Temporal Difference Models: Model-Free Deep RL for Model-Based Control (ICLR 2018 · similarity 0.88)
- Towards Parameter-Free Temporal Difference Learning (ICML 2026 · similarity 0.87)
- TD-MPC2: Scalable, Robust World Models for Continuous Control (ICLR 2024 · similarity 0.87)
- Bootstrapped Model Predictive Control (ICLR 2025 · similarity 0.86)
- Time-Aware World Model for Adaptive Prediction and Control (ICML 2025 · similarity 0.85)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · 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.
