@MASTERSTHESIS\{IMM2013-06710,
    author       = "T. Gruber",
    title        = "Optimization for fast and safe trajectories",
    year         = "2013",
    school       = "Technical University of Denmark, Department of Applied Mathematics and Computer Science",
    address      = "Matematiktorvet, Building 303B, {DK-}2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk",
    type         = "",
    note         = "{DTU} supervisor: Niels Kj{\o}lstad Poulsen, nkpo@dtu.dk, {DTU} Compute",
    url          = "http://www.imm.dtu.dk/English.aspx",
    abstract     = "The subject of this dissertation is optimizing the trajectory of a car on a highway. The goal is to give decisions to the driver about what speed to adopt and when to change lane in order to satisfy their desire for speed but still ensure safety.
A simulation framework has been programmed so that the car of study can evolve in an environment that simulates real situations. Within this simulation model, the goal is to find an optimal trajectory for the car of study. Optimal trajectories can be retrieved offline, where time backtracking is possible: the A* algorithm has been implemented to perform this search. But the aim is also to find an online algorithm to assist the driver in real-time: an agent-oriented algorithm has been programmed for a real-time assistance algorithm. Only safety and time has been taken into account for the optimality criterion; for the offline optimal search, safety is considered as a constraint and time is optimized. Making safety constraints vary for several optimal searches provides a range of optimal solutions that can be used as benchmarks to assess the performance of the online algorithm.
The implemented programs work and are computationally achievable. The optimal searches provide a various range of trajectories which enable to choose the best trade-o between time and safety. The agents have shown to take good decisions that make the trajectory safe and fast for basic tests. But when the scenario gets more complex, they may take irrational decisions. Nevertheless, this thesis shows the feasibility of such a program and provides a first flexible framework to be improved."
}