Development of Velocity planner for a Racing driver model

Sammanfattning: Testing a vehicle’s limit handling capabilities can provide a valuable insight into analysing and hence developing new vehicle designs. A driver model is a convenient tool used particularly in full vehicle simulations and if it is capable of simulating driving on the limit with minimum user input then it becomes all the more effective. There is inadequate information found in literature concerning the simulation of limit handling capability of a vehicle using a driver model.There was a requirement to improve an existing driver model in Modelon AB’s Vehicle dynamics library (VDL), so that it would enable the user find the limits of handling of a vehicle. This also had to be accomplished with minimum user input and without a priori information of the entire track. Furthermore, the vehicle model should be driven on the pre-defined path on the road accurately. Hence the speed calculation has to be accurate, capable of triggering braking, whenever required to successfully maintain the trajectory of the vehicle on the given path. The targets of lateral and longitudinal acceleration/deceleration of the vehicle model were the only input specified by the user. The complete driver model was then to be used in closed loop driving manoeuvre simulations on a 2-D road. The velocity planning block, developed in this thesis, calculates the vehicle speed based on user inputs. It also utilises a limited look-ahead preview of the track and ensures that the vehicle model is pushed to the boundaries of these specified targets while at the same time maintaining the pre-defined trajectory. The tests are conducted on three different tracks; A circular track with constant radius, J-turn and an extended chicaneTwo strategies were employed to solve the problem. First, a single-point look-ahead preview, where information about one point of the upcoming road curvature is known, is incorporated into the driver model. Later, in order to make the solution more robust, a multi-point preview is implemented. The velocity profile calculated was found to push the car to the specified targets and the velocity tracking was performed to a satisfactory level. The vehicle model accurately followed the pre-defined vehicle trajectory without a priori information of the track, and in the case of single-point preview strategy, the look-ahead distance required almost no fine tuning of the preview distance for the three tracks the model was tested on.