Dermaku, A. (2012). Path recognizing for mobile and humanoid robots through low cost sensors [Dissertation, Technische Universität Wien]. reposiTUm. http://hdl.handle.net/20.500.12708/160540
The aim of this work is the development of the new methods oriented on the creation of new software approaches which will try to calculate an optimal or nearly optimal path from robot to the target object. The new heuristic algorithms should be implemented which considering the data obtained from multiple sensors (for example light sensors, touch sensors, ultrasonic sensors, infrared sensors, visual sensors etc.) decides for the new moving steps of robots on partial segment of the path, as well an optimal behavior of robot when it stay in front of target objects. The gathered data from multiple sensors should make possible also an optimal navigation of robot in terms of avoiding collisions between robot and objects. Using of the Extended Kalman Filters and Neural Network approach should make possible the updating of the robots position during its localization. Especially the applying of Neural Network leads to the reducing of the robots position error, as difference between desired and achieved position. The final goal is finding of local optimal path using different heuristics algorithms which enable a fast and local - best decision. The approaches include also the propagations in consideration. The new implemented heuristic algorithms based on hypertree decomposition and geometric intersection calculates the optimal path between start and target position. The simulation application in C# and Matlab is implemented. The main advantages of such approaches are the time reducing of map-calculating and replacement of expensive high-tech devices with low-cost sensors and better software solutions. For objects detecting the Vision System as part of perception is used. It makes possible the map calculation of robots work environment. Finally , the simulations for real robots HUMI and ARCHIE implemented at the Institute of Handling Robotics and Technology (IHRT) are done. The evaluations are done by comparing the simulation obtained from heuristics algorithms and simulation obtained from exact algorithms on the same test.
