Model predictive fuzzy control:
A hierarchical multi-agent control architecture for outdoor search-and-rescue robots
Multi-agent systems are mainly controlled via one of the following architectures: centralized, decentralized (when there is no interaction among the agents that allows them to directly influence the decisions of other agents), and distributed (when agents can exchange information and can thus influence the decisions of each other).
Each of these architectures faces advantages and drawbacks: While centralized control can provide global optimality for the performance, it is computationally very expensive (especially for large-scale systems) and the control system collapses, whenever the centralized controller fails during the mission.
Besides, decentralized control architectures are usually computationally less expensive than the other two architectures, but if non-negligible influences and inter-dynamics exist among the agents, the performance of this architecture is worse due to lack of coordination between the agents.
Finally, distributed control may be used to break the centralized control problem into smaller local control problems with links and inter-dynamics. Distributed control, however, requires exchange of information among the agents (which may be very costly or even impossible for SaR robots) and computationally complicated algorithms for coordination of the agents.
Our proposed architecture, MPFC, integrates the advantages of all these architectures through its bi-layer structure.
The main contributions of this project include:
A novel control architecture, called MPFC, which integrates and combines the advantages of MPC and FLC to include global optimality and predictive decision making of MPC and time-efficient, human-inspired decision making of FLC in one control system
An event-triggered tuning and control architecture that performs via a decentralized architecture in real time, but still incorporates desired levels of coordination and performance improvement at the global level within the multi-agent decision making system
Application of MPFC for autonomous multi-robot path planning in SaR missions, proposing a grid-based 2D model of a dynamic SaR environment that includes fire spread dynamics, as well as performance analysis and assessment of different configurations of the MPFC system compared to different control systems for multi-agent SaR robotics