# Model Predictive Control

Small execution errors in imitation policies accumulate into failure, and earlier fixes relied on additional on-policy learning or costly online planning. A2WM learns local action sensitivity offline from action-perturbed rollouts of a frozen policy, predicting task-relevant geometry as a nominal trajectory plus a linear response to corrections, which turns online refinement into one model evaluation and one convex quadratic program solve. Across four simulated manipulation tasks it reaches the highest mean success among the evaluated baselines, 59.51% versus 27.18% for Diffusion Policy, with 3.8 ms mean refinement overhead per control step excluding base-policy sampling.

Small execution errors in imitation policies accumulate into failure, and earlier fixes relied on additional on-policy learning or costly online planning. A2WM learns local action sensitivity offline from action-perturbed rollouts of a frozen policy, predicting task-relevant geometry as a nominal trajectory plus a linear response to corrections, which turns online refinement into one model evaluation and one convex quadratic program solve. Across four simulated manipulation tasks it reaches the highest mean success among the evaluated baselines, 59.51% versus 27.18% for Diffusion Policy, with 3.8 ms mean refinement overhead per control step excluding base-policy sampling.

Nonlinear time-optimal model predictive control (TOMPC) drives constrained systems to a target in minimum time, but online solving is too slow for fast dynamics. This work imitates a long-horizon, nonconvex TOMPC with hard terminal constraints using a neural network, and introduces self-informed sampling: the network also predicts time-to-go, and the expert solver is queried only where a rollout departs from time-optimal behavior. On a Furuta pendulum, this sampling reaches the performance of standard dataset aggregation in roughly one-third of the training time and less than half the data, while the resulting controller keeps the expert's region of attraction and GPU inference raises the control frequency from about 1.4 Hz to 1 kHz.

Nonlinear time-optimal model predictive control (TOMPC) drives constrained systems to a target in minimum time, but online solving is too slow for fast dynamics. This work imitates a long-horizon, nonconvex TOMPC with hard terminal constraints using a neural network, and introduces self-informed sampling: the network also predicts time-to-go, and the expert solver is queried only where a rollout departs from time-optimal behavior. On a Furuta pendulum, this sampling reaches the performance of standard dataset aggregation in roughly one-third of the training time and less than half the data, while the resulting controller keeps the expert's region of attraction and GPU inference raises the control frequency from about 1.4 Hz to 1 kHz.

Short-horizon model predictive control runs fast but cannot see an opponent beyond its horizon, so it needs a learned, context-dependent cost correction; when passing is possible on either side, both corrections are good and deterministic regression averages them into the opponent. ZipDrift instead trains a one-step drifting generator over cost corrections, tilted by closed-loop cost and learned without differentiating through the optimiser, then generates a few candidates ranked by a learned critic. On a straight-track overtaking benchmark with an eight-dimensional stagewise correction, all 600 test episodes were collision-free, against a 4.0% crash rate for an idealised ZipMPC-style regressor.

Short-horizon model predictive control runs fast but cannot see an opponent beyond its horizon, so it needs a learned, context-dependent cost correction; when passing is possible on either side, both corrections are good and deterministic regression averages them into the opponent. ZipDrift instead trains a one-step drifting generator over cost corrections, tilted by closed-loop cost and learned without differentiating through the optimiser, then generates a few candidates ranked by a learned critic. On a straight-track overtaking benchmark with an eight-dimensional stagewise correction, all 600 test episodes were collision-free, against a 4.0% crash rate for an idealised ZipMPC-style regressor.

Hospital delivery robots must decide both which request to take and in what order to visit stops, while STAT and routine items carry different clinical deadlines. HMART puts that joint decision into a shared graph-attention encoder with two sequential decoders, bounds the tail of weighted tardiness with a conditional value-at-risk constraint, and coordinates the shared elevators through a model predictive control layer. In a discrete-event simulation of a four-floor hospital with three elevators and twelve robots, the engineered-decoder instantiation lowers mean total weighted tardiness from 2359±214 to 29±6 against the strongest dispatching rule and raises the STAT on-time rate from 0.925±0.011 to 0.969±0.007.

Hospital delivery robots must decide both which request to take and in what order to visit stops, while STAT and routine items carry different clinical deadlines. HMART puts that joint decision into a shared graph-attention encoder with two sequential decoders, bounds the tail of weighted tardiness with a conditional value-at-risk constraint, and coordinates the shared elevators through a model predictive control layer. In a discrete-event simulation of a four-floor hospital with three elevators and twelve robots, the engineered-decoder instantiation lowers mean total weighted tardiness from 2359±214 to 29±6 against the strongest dispatching rule and raises the STAT on-time rate from 0.925±0.011 to 0.969±0.007.

The work proposes a dual-process-inspired hybrid framework in which a TransFuser++ neural planner generates human-like trajectories in routine driving while a meta-cognitive component, using risk fields shaped by knowledge-graph reasoning, decides when to switch to a model predictive control planner; tested in CARLA on out-of-distribution scenarios of vehicles (including an ambulance) running a red light, it reduced vehicle-to-vehicle collisions from 48 to 5 out of 160 simulations, an 89% reduction.

The work proposes a dual-process-inspired hybrid framework in which a TransFuser++ neural planner generates human-like trajectories in routine driving while a meta-cognitive component, using risk fields shaped by knowledge-graph reasoning, decides when to switch to a model predictive control planner; tested in CARLA on out-of-distribution scenarios of vehicles (including an ambulance) running a red light, it reduced vehicle-to-vehicle collisions from 48 to 5 out of 160 simulations, an 89% reduction.

The authors propose Ctrl-CWM, a multi-agent controllable crowd world model that first learns and freezes an encoder via trajectory prediction on real pedestrian videos to preserve human motion dynamics, then uses an actor to propose displacements, a critic to rank imagined crowd trajectories, and a planner combining critic scores with user costs to select actions, enabling crowd generation and run-time control without retraining; on ETH–UCY, SDD, and GCS it outperforms CrowdES on most crowd realism and collision metrics and achieves the highest average compliance in avoidance and attraction scenarios.

The authors propose Ctrl-CWM, a multi-agent controllable crowd world model that first learns and freezes an encoder via trajectory prediction on real pedestrian videos to preserve human motion dynamics, then uses an actor to propose displacements, a critic to rank imagined crowd trajectories, and a planner combining critic scores with user costs to select actions, enabling crowd generation and run-time control without retraining; on ETH–UCY, SDD, and GCS it outperforms CrowdES on most crowd realism and collision metrics and achieves the highest average compliance in avoidance and attraction scenarios.

ReCo couples a response-consistent reinforcement-learning locomotion policy with policy-aware model predictive control (MPC): response shaping trains the policy to respond to commands consistently and repeatably across randomized dynamics, and an identified closed-loop response model then lets MPC jointly plan locomotion commands and arm motion, reducing position and orientation root-mean-square error (RMSE) on the simulation benchmark by 28.7% and 27.4% relative to the best baseline for each metric, with real-world experiments demonstrating onboard continuous legged manipulation and coordinated base and arm motion.

ReCo couples a response-consistent reinforcement-learning locomotion policy with policy-aware model predictive control (MPC): response shaping trains the policy to respond to commands consistently and repeatably across randomized dynamics, and an identified closed-loop response model then lets MPC jointly plan locomotion commands and arm motion, reducing position and orientation root-mean-square error (RMSE) on the simulation benchmark by 28.7% and 27.4% relative to the best baseline for each metric, with real-world experiments demonstrating onboard continuous legged manipulation and coordinated base and arm motion.