| dc.description.abstract | In the past decades, autonomy has gone from a technology of the future to an inevitability. Driver-less cars are already on the roads, and autonomous surface vessels (ASVs) are well within range. With this increase in autonomy, follows a demand for robust path planning and collision avoidance methods. Further, ASVs have to share their working environment with other manned vessels. This calls for an additional element of the motion planning problem, in that the ASVs should follow the traffic rules outlined by the International Regulations for Preventing Collisions at Sea (COLREGS).
This thesis investigates how a stochastic approach, in the form of the \acrfull{POP} algorithm, can be used within a motion planner to avoid multiple static and dynamic obstacles. Moreover, a feasibility design by the means of a simulator is completed, in which the simulator consists of a vessel model that is fitted with a guidance and control system. With these systems in place, a range of scenarios are used to assess the motion planner.
It was found that merging the \acrshort{POP} algorithm with an A* search was needed to obtain decisive results, as the \acrshort{POP} algorithm struggled in the encounter with multiple, static obstacles. For collision avoidance, the motion planning algorithm performed better. Much of this improvement is credited to the addition of virtual target points, which ensured that the \acrshort{ASV} adhered to COLREGS in three separate collision scenarios.
Even though the \acrshort{POP} algorithm endures flaws, its stochastic approach is still considered as one of its strengths. This, as the working environment for all ASVs is stochastic by nature. It is therefore vital to incorporate these uncertainties into a motion planner. | |
