Repeated Path Planning for Mobile Robots in Dynamic Environments
This thesis presents a global navigation strategy for repeated traverse of mobile robots in dynamic environments.
The research is motivated by the fact that most real-world applications of mobile robotics imply repeated traverse between predefined target points in large, uncertain real-world domains.
The goal of this study is to develop a framework for navigation in large-scale dynamic environments that permits the robot to fulfill its assignment by adapting to use of easily traversable paths.
While mobile robots usually tackle the problem of path planning in uncertain environments by implementing obstacle avoidance routines, in this approach the robot also learns to use routes along which obstacles occur less often. The few alternative approaches that consider the problem of global obstacle avoidance function in well-structured environments using an a priori known topological map. The approach reported in this thesis uses a grid-based map for path planning. It is argued that a grid-based map together with a stochastic wavefront planner offers a greater number of alternatives to the path planning problem and that the robot adapts to be able to use the best routes even when very little a priori knowledge of the environment is available.
To adapt to dynamic environments the robot uses case-based reasoning to remember the paths followed and to reason about their traversability. While case-based reasoning is usually applied for planning problems in static environments, this approach also works in dynamic environments and uses a unique similarity measure to reduce the solution space.
This thesis also offers an alternative decision-making strategy for mobile robots in hazardous environments that is based on the concept of irreversible decisions.
The experiments made in simulated and real environments demonstrate that this approach to global navigation permits the robot to increase the predictability of its behavior and to minimize the risk of collision an time delays.