Fast Multi-Robot Motion Planning via Imitation Learning of Mixed-Integer Programs

Srinivasan, Mohit; Chakrabarty, Ankush; Quirynen, Rien; Yoshikawa, Nobuyuki; Mariyama, Toshisada; Cairano, Stefano Di

doi:10.1016/j.ifacol.2021.11.237

Cited by 6 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However this approach is best suited for mp-MIPs of moderate size because the complexity of the online look-up and offline storage of partitions, increases rapidly with scale [8]. The second approach for real-time mixed-integer MPC relies on predicting warm-starts for the mp-MIP by training Machine Learning (ML) models on large offline datasets [9], [10], [11], [12], [13]. The authors of [9], [10], [11] use various supervised learning frameworks to predict the optimal integer variables for the mp-MIP at a given parameter so that the online computation is reduced to solving a convex program.…”

Section: Introductionmentioning

confidence: 99%

“…The second approach for real-time mixed-integer MPC relies on predicting warm-starts for the mp-MIP by training Machine Learning (ML) models on large offline datasets [9], [10], [11], [12], [13]. The authors of [9], [10], [11] use various supervised learning frameworks to predict the optimal integer variables for the mp-MIP at a given parameter so that the online computation is reduced to solving a convex program. In [12], [13], the authors define the notion of an optimal strategy for a mp-MIP as a mapping from parameters to the complete information required to efficiently recover an optimal solution.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learning for Online Mixed-Integer Model Predictive Control with Parametric Optimality Certificates

Russo¹,

Nair²,

Glielmo³

et al. 2023

Preprint

View full text Add to dashboard Cite

We propose a supervised learning framework for computing solutions of multi-parametric Mixed Integer Linear Programs (MILPs) that arise in Model Predictive Control. Our approach also quantifies sub-optimality for the computed solutions. Inspired by Branch-and-Bound techniques, the key idea is to train a Neural Network/Random Forest, which for a given parameter, predicts a strategy consisting of (1) a set of Linear Programs (LPs) such that their feasible sets form a partition of the feasible set of the MILP and (2) a candidate integer solution. For control computation and sub-optimality quantification, we solve a set of LPs online in parallel. We demonstrate our approach for a motion planning example and compare against various commercial and open-source mixedinteger programming solvers.* denotes equal contribution. SHN and FB are with the Model Predictive Control lab, University of California, Berkeley. LR and LG are with University of Sannio.

show abstract