Distributed Model Predictive Control for Multi-Agent Systems
Controle prédictif distribuée pour les systèmes multi-agents
Résumé
This work proposes a navigation scheme that addresses the problem of collision-free navigation of multiple non-holonomic mobile robots in the presence of non-coupled and coupled state constraints, and input constraints. The scheme named a two-stage distributed model predictive control (DMPC), in the first stage, each agent computes in parallel, a warm-start trajectory as an initialization for the second stage using a less-constrained DMPC. Then, the second stage solves the full-constrained DMPC scheme with local information exchange between neighboring robots. This approach offers a simpler and more travel-time-efficient alternative to previous DMPC techniques. The performance was validated through simulations and experiments
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |