Modelling and Trajectory Planning for a Small-Scale Surface Ship

Abstract: Autonomous ships are one way to increase safety at sea and to decrease environmental impact of marine traveling and shipping. For this application, a good representation of the environment and a physical model of the ship are vital components. By optimizing the trajectory of the ship, a good trade-off between the time duration and energy consumption can be found. In this thesis, a three degree of freedom model that describes the dynamics of a small-scale surface ship is estimated. By using optimal control theory and a grey-box model, the parameters are estimated by defining an optimal control problem (OCP). The optimal solution is found by transcribing the problem into a nonlinear program and solving it using an interior point algorithm. The identification method is tested and validated using simulated data as well as using data from real world experiments. The performance of the estimated models is validated using cross validation. In a second track of this thesis, a trajectory is created in two steps. The first is path planning to find a shortest geometric path between two points. In the second step, the path is converted to a trajectory and is optimized to become dynamically feasible. For this purpose, a roadmap is generated from a modified version of the generalized Voronoi diagram. To find an initial path in the roadmap, the A-star algorithm is utilized and to connect start and goal position to the map a few different methods are examined. An initial trajectory is created by mapping a straight-line trajectory to the initial path, thus connecting time, position and velocity. The final trajectory is found by solving a discrete OCP initialized with the initial trajectory. The OCP contains spatial constraints that ensures that the vessel does not collide with static obstacles. The suggested estimation method resulted in models that could be used for trajectory planning to generate a dynamically feasible trajectory for both simulated and real data. The trajectory generated by the trajectory planner resulted in a collision-free trajectory, satisfying the dynamics of the estimated model, such that the trade-off between time duration and energy consumption is well balanced. Future work consists of implementation of a controller to see if the planned trajectory can be followed by the small-scale ship.