The main goal of this PhD is to extend the use of vehicle dynamics computation models by proposing new methodologies inspired from the motion planning area. The vehicle dynamics models which are being considered are industrial grade : they are complex, non linear and most of the time only available as a black box. They are traditionally used with simulation techniques requiring a precise initial state and inputs coming from a driver. The methods we are proposing take into account a more realistic description of the current test : an initial state the vehicle is starting from, a corridor to be followed and an arrival area. Additionally to solving the problem of finding such a solution maneuver, two other questions are being addressed : finding a representative set of the numerous solutions; finding the limit of a parameter (for instance the initial speed) after which no solution is found. Several techniques have been successfully tested : global exploration methods, optimal control and a modified trajectory deformation algorithm. These generic tools are able to painlessly adapt to new vehicles or obstacles. Each one has specific pros and cons. These methods have been applied to the example of standardized drive maneuvers. The physical limits of a vehicle passing through these tests have been robustly discovered.
