The main contents of the thesis are presented in three papers.
In the first paper the focus is path planning and navigation control for a robot equipped with sensors. For a robot trying to reach places in at least partially unknown environments there is often a need to replan paths on-line based on information extracted from the surroundings. In the paper it is assumed that the sensing range of the robot is short compared to the length of the paths it plans and the environment is modeled as a graph consisting of nodes and arcs. The replanning problem is first modeled as a minimum cost flow problem and then solved using the network simplex method. The applicability of the planner is demonstrated by integrating it with a navigation control strategy. Simulation results show that both the planner and the integrated approach works well.
The second paper deals with nonlinear observers. In the paper state observers for control systems with nonlinear outputs are studied. For such systems, the observability does not only depend on the initial conditions, but also on the exciting control used. The paper includes a deeper analysis for a special class of systems and some sufficient conditions are given for the convergence of an observer. It is also discussed, via a vision example, how to actively excite a system in order to improve the observability.
The third paper is closely related to the first and has a small connection to the second. Here path planning and sensor modeling are the main ingredients. As previously mentioned, for vehicles operating in uncharted environments there is a need to adjust plans on-line based on new discoveries. Simple obstacles can be dealt with using local maneuvers, but more complex problems need a more global treatment. For such cases the first paper describes the design of an on-line path planner. In the third paper the older path planner is used as a foundation when designing a more advanced alternative. The main advantage of the new planner is that it adapts well to obstacles of different shapes. Modeling of environments and sensors are used for evaluating the performance of the new method. Test results show a significant improvement compared to the previous planner.