We develop a multi-agent model predictive control method for a class of hybrid systems governed by discrete inputs and subject to global hard constraints. We assume that for each subsystem the local objective function is convex and the local constraint function is strictly increasing with respect to the local control variable. The proposed multi-agent control method is based on a distributed resource allocation coordination algorithm and it only requires limited information sharing among the local agents of the subsystems. Thanks to primal decomposition of the global constraints, the distributed algorithm can always guarantee global feasibility of the local control decisions, even in the case of premature termination. Moreover, since the control variables are discrete, a mechanism is developed to branch the overall solution space based on the outcome of the resource allocation coordination algorithm at each node of the search tree. Finally, the proposed multi-agent control method is applied to the charging control problem of electric vehicles under constrained grid conditions. This case study highlights the effectiveness of the proposed method.

Due to the lack of efficient control methods for a fleet of vehicles throughout a road network, the large-scale application of cybercars, which are fully automatic road vehicles providing on-demand and door-to-door transportation service, is
still hindered. Although the fleet control problem for cybercars can be straightforwardly addressed in a centralized control setting, for reasons of scalability and fast computation, a centralized control method will not be tractable for the large-scale use of cybercars in the future. In this paper, we focus on the dynamic routing of a fleet of cybercars considering minimization of the combined system cost including the total time spent and the total energy consumption by all cybercars. We first propose a model of the dynamics and the energy consumption of a fleet of cybercars based on a description of the dynamics of every single cybercar and the states of the road network. After that, we propose several tractable and scalable multi-agent control methods including
multi-agent model predictive control and parameterized control for the dynamic routing of cybercars. Finally, experiments by means of numerical simulations illustrate the performance of the proposed control methods.

This thesis aims at developing efficient methods for control of large-scale systems by employing state-of-the-art control methods and optimization techniques. This thesis is divided into two parts. In the first part, we address dynamic traffic routing for urban transportation networks. In the second part, we address multi-agent model predictive control of a class of hybrid systems with limiting information sharing and subject to global hard constraints.

Due to the nonlinearity of the dynamics of vehicles and the discrete nature of the route decision variables, the dynamic routing problem for a large number of vehicles is computationally very hard to solve. In this paper, two efficient parameterized control methods are proposed for the dynamic routing of a fleet of cybercars in a road network only open to cybercars. With the proposed parameterized control methods, the updates of the routes of cybercars are parameterized and then optimized over the parameters with respect to the overall performance of the cybercar system for a representative set of scenarios. After tuning the parameters, the proposed parameterized control methods are implemented online with fixed parameters. Moreover, the two proposed parameterized control methods are well-structured and scalable, and therefore can be applied to road networks with arbitrary topologies. The effectiveness of the proposed parameterized control methods is shown in a numerical simulation study.