Papers¶
Patterns implemented¶
| Blend | Primary sources |
|---|---|
| Residual RL | Johannink et al. ICRA 2019; Silver et al. 2018 |
| Policy warm start | Mansard et al. ICRA 2018; RL-guided MPPI (arXiv:2606.25123) |
| Learned terminal cost | Bhardwaj et al. ICLR 2021; POLO, ICLR 2019 |
| Expert distillation | DAgger, AISTATS 2011; Guided Policy Search, ICML 2013 |
Surveys and context¶
- Synthesis of Model Predictive Control and Reinforcement Learning: Survey and Classification, 2025 (arXiv:2502.02133) — classifies the MPC×RL landscape; blendmpc implements its "MPC and RL as separate interacting modules" family.
- For the complementary "MPC as function approximator" family (Gros &
Zanon), see
mpcrlandmpc4rl.
Using blendmpc in a paper?¶
Please cite via the repository's CITATION.cff, and open a PR to add your
work here — a growing list of papers using the library is the best signal for
new users.