Safe Approximate Nonlinear Model Predictive Control Without Expert Demonstrations
Safe Approximate Nonlinear Model Predictive Control Without Expert Demonstrations
This research introduces Sequential Approximate MPC (Seq-AMPC), a label-free recurrent framework for nonlinear model predictive control that generates complete control sequences using shared parameters across the prediction horizon.
The paper proposes Seq-AMPC, a label-free recurrent framework for safe approximate nonlinear model predictive control that achieves closed-loop cost ratios within 0.3% of NMPC across four benchmarks without requiring expert demonstrations.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
mean closed-loop cost ratio difference from NMPC
0.3
View original evidence
itation training. Across four benchmarks, training takes 0.2-1.7 GPU hours per policy without expert control labels. The filtered controller applies 97.2-99.1% of neural proposals and achieves mean closed-loop cost ratios within 0.3% of NMPC on jointly completed episodes. Tests with suboptimal feasible initial plans further demonstrate the benefit of neural replanning.
Open source S5
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
itation training. Across four benchmarks, training takes 0.2-1.7 GPU hours per policy without expert control labels. The filtered controller applies 97.2-99.1% of neural proposals and achieves mean closed-loop cost ratios within 0.3% of NMPC on jointly completed episodes. Tests with suboptimal feasible initial plans further demonstrate the benefit of neural replanning.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org