Predictive Vehicle Velocity Control using Dynamic Traffic Information 2016-01-0121

Predictive velocity control can be used to enable efficient driving regarding fuel efficiency and driving time. Commonly, velocity optimization algorithms only take static information, like road slope and curvature, into account and neglect dynamic information, like traffic lights and other traffic participants, although the information is available through sensors or could be made available by vehicle-tovehicle or vehicle-to-infrastructure communication. Thus, static optimization algorithms do not provide optimal solutions in dynamic environments, caused by driver or assistance systems intervention. Because the incorporation of dynamic information increases the complexity of the problem to find an optimal control policy, its use in real-time applications is often prohibited.

An algorithm is presented which allows a fast computation of all optimal speed profiles with regard to time and fuel consumption. The algorithm iteratively computes possible velocity profiles of the vehicle from which all suboptimal profiles are filtered out. The filtering highly reduces the amount of possible solutions that are tracked, so that the algorithm can be used in real-time control for certain prediction horizons.

Among the optimal solutions the most favorable velocity profile can then be chosen according to driver preferences and traffic situation. The evaluation of many optimal profiles allows to deviate from a fixed weighting between time reduction and fuel efficiency in situations where a small increase in overall fuel consumption can result in a highly reduced time or vice versa.

In a simulation it is shown that the obtained velocity profiles reduce time and fuel consumption, compared to a driver model for different traffic scenarios.