Chance Constrained Non-Linear MPC
Model Predicitve Control for multi-agent systems using probablistic constraints.
Multi-agent Model Predictive Control is a developing field, and many variations of it are being researched on. In this project we have tested different multi-agent MPC approaches, exploring centralized and decentralized control methods. The decentralized MPC was implemented with probabilistic Chance Constraints to get robust avoidance behavior while navigating with uncertain localization. Chance-constrained MPC was enforced by minimizing area of overlap using a convex function on area. This improvised Chance Constrained MPC was tested on simulation with 2 and 3 agents.
The poster for the project