Doctorant sous la direction de

**H. Mounier****Titre de la thèse:**Constrained Optimal Navigation System for Hybrid Vehicles

**Résumé de la thèse:**The main problems in the optimal design of real time optimal navigation system are related to the varying character of system’s parameters, the quality of information collected from the traffic network, the decision taken at the driver level as well as the routing algorithm complexity. The first aspect is essentially related to the characteristics of the involved sub-systems models. In some special cases, such as optimal energy navigation system, the energy consumed also depends on the state of the on board energy sources as well as on the traffic state. This fact not only impacts the complexity of the optimisation problem but also that of the resulting algorithms. The second problem involves the quantity of information gathered by the traffic information system in real time from the traffic network. Currently, the best information consists in constant velocity road segments periodically updated each 3 to 5 minutes. The third one concerns the decision system at driver level and the sub-optimality induced or otherwise said “price of anarchy” highlighting the importance and necessity of coordination to reduce it. In the framework of this PhD, we propose to proceed through different stages. In the first one, we will try to enhance the information on traffic state by sharing it with other drivers in the “vicinity” (term to be defined consistently with respect to technology under use …), as well as using simple and generic models in order to identify the traffic flow. This will allow us to smooth or filter the constant speed information and gain a better knowledge of the real energy consumption function. A Reformulation of the optimal navigation problem, followed by a resolution of the latter consistently with respect to the traffic state dynamic as well as the induced complexity reduction will be one of our main concern. The second problem, closely related to the first one, concerns the traffic state information consistency with respect to the real value of its state. We will propose new classes of algorithms handling more efficiently the available traffic state information in order to increase the predictability of optimal routes and reducing the probability to be stuck in future congestion. The third class of algorithms we have to consider, at the routing level, is the influence of the on board energy sources state as well as the reduction of their feasibility domain or search space, through a more precise computation of the evaluation function used in speed-up technique (A* like ones). From a practical viewpoint, we will also treat the problem of constrained optimal real time routing going through a finite number of way points, in the single vehicle case as well as the multiple one. As an application, we propose to treat the problems related to constrained energy optimal navigation of a fleet of vehicles, which may be one to many or many to many navigation/routing problem. This problem is consistent with a taxi company vehicle fleet management equipped with Hybrid Vehicles as well as to a delivery company.