Research map / Reinforcement learning / Visual and representation RL
Differentiable MPC for End-to-end Planning and Control
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 |
|---|---|---|---|
| Use differentiable MPC instead of a generic LSTM policy | Imitation learning on Pendulum and Cartpole · Pendulum and Cartpole | imitation loss | Helped |
| Optimize dynamics using imitation loss from controls rather than next-state transitions | MPC.dx · Non-realizable Pendulum | imitation performance | Helped |
| Use vanilla system identification instead of optimizing imitation loss | Dynamics learning for MPC imitation · Non-realizable Pendulum | next-state transition fit | Helped |
| Differentiate at the solver's fixed point instead of unrolling iLQR iterations | Differentiable MPC · MPC solver performance | backward-pass efficiency | Helped |
| Differentiate at the solver's fixed point instead of unrolling iLQR iterations | Differentiable MPC · MPC solver performance | forward-pass memory use | Helped |
| Learn dynamics from action-only imitation instead of next-state data | MPC.dx · Pendulum and Cartpole | imitation performance | No clear effect |
| Learn controller parameters by action imitation from randomly initialized dynamics | LQR imitation learner · Linear-dynamics quadratic-cost imitation | model recovery | Mixed |
| Directly optimize the task's imitation loss rather than using system identification as a surrogate | MPC.dx · Non-realizable Pendulum | imitation performance | Helped |
Closest papers by meaning
- Infinite-Horizon Differentiable Model Predictive Control (ICLR 2020 · similarity 0.89)
- Blending MPC & Value Function Approximation for Efficient Reinforcement Learning (ICLR 2021 · similarity 0.89)
- Bootstrapped Model Predictive Control (ICLR 2025 · similarity 0.88)
- Differentiable Information Enhanced Model-Based Reinforcement Learning (AAAI 2025 · similarity 0.88)
- Dream-MPC: Gradient-Based Model Predictive Control with Latent Imagination (ICML 2026 · similarity 0.88)
- TD-MPC2: Scalable, Robust World Models for Continuous Control (ICLR 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.