RL-MPC Integration Taxonomy for Linear Systems
A systematic literature review categorizes integrations of Reinforcement Learning (RL) and Model Predictive Control (MPC) for linear and linearized systems. The study covers peer-reviewed works published up to 2025, organizing them via a multi-dimensional taxonomy that includes RL functional roles, RL algorithm classes, MPC formulations, and cost-function structures. The integration aims to combine MPC's structured optimization and constraint handling with RL's data-driven adaptation under uncertainty. The paper addresses fragmentation in the rapidly growing field.
Key facts
- Systematic Literature Review (SLR) of RL-MPC integrations for linear/linearized systems
- Covers peer-reviewed and indexed studies published until 2025
- Multi-dimensional taxonomy includes RL functional roles, RL algorithm classes, MPC formulations, cost-function structures
- MPC offers structured optimization, explicit constraint handling, stability tools
- RL provides data-driven adaptation and performance improvement under uncertainty and model mismatch
- Integration paradigm targets constrained decision-making and adaptive control
- Literature remains fragmented despite rapid growth
- Paper published on arXiv with ID 2604.21030
Entities
Institutions
- arXiv