engleski

Motion Planning of Mobile Robots in Indoor Environments

Motion planning is a part of mobile robot navigation that deals with finding a feasible and collision-free path towards a goal for a mobile robot to navigate to. A sequence of motion control commands are given to the robot that must comply to its kinematic and dynamic constraints. Two principal approaches can be identified that use opposite strategies: planning and reactive control. The former has the advantage of using global information about the environment and can thus provide the robot with a path that ensures global-space connectivity. The latter uses sensor data in the local vicinity of the robot and generates motion commands based exclusively on the current state of the robot and the environment, giving the robot a capability to quickly adapt to environment changes. By combining the two approaches the resulting architecture can resolve the motion planning problem in dynamic or partially unknown environments in a robust manner. This work aims at providing a possible architecture to such integration. The global geometric path is generated according to A* and D* graph search algorithm within an occupancy grid map representation of the environment. The path is further explored to find the key representing points by using free-space bubbles. Their central points provide the basis for B-spline interpolation to obtain the final smooth path. Reactive control used for obstacle avoidance purposes is primarly based on dynamic window approach performed direct in velocity space of the robot. Two path following approaches are presented that both use the notion of a reference point on the global path. The first exploits characteristics of the basic geometric path by using an effective path comparison to possible robot trajectories of the obstacle avoidance module and thus represents a direct integration between path planning and reactive control. The second is based on the virtual vehicle approach and exploits the curvature characteristics of the smooth global path on a local scale according to a criterion called curvature effort. It can be viewed as an indirect combination of reactive control and planning since the local smooth path is recalculated on-line based on sensor data whereas the global free-space path is still maintained.

mobile robot navigation; motion planning; obstacle avoidance; path planning; path following; graph search methods

nije evidentirano