| Popis: |
Probabilistic methods provide a powerful paradigm for modeling of robot motion and its environment; a precursor for autonomous navigation of mobile robots. As a graduate-level engineering course, probabilistic robotics encompasses the techniques used in robot localization and mapping in unstructured environments. This article presents a simulation and animation software program developed mainly as a teaching tool that can help the students visualize different robot localization solutions through both parametric filters (viz., the Extended Kalman Filter, Unscented Kalman Filter) and nonparametric filters (viz., the histogram filter, Rao-Blackwell particle filter). The program is also a powerful tool for performance analysis and tuning of such filters commonly used for robot localization, mapping, and autonomous navigation. The simulation features dead reckoning (e.g., INS), range-only sensing (e.g., rangefinder), bearing-only sensing (e.g., digital camera), or a combination of them. The program includes a simple graphical user interface that allows for changing both filtering and sensing parameters, and monitoring the effects of those changes on the animation of a unicycle in a 2D environment. The program animates the unicycle motion and shows the kinematic results in several graphs simultaneously to evaluate the performance of different methods in finding the robot pose. The program is available as an open-source Matlab script to facilitate future modifications and improvements of the code by the students interested in robotics, mechatronics, and control engineering. This article presents the features of the program and briefly discusses the algorithms implemented in the software. © 2010 Wiley Periodicals, Inc. Comput Appl Eng Educ 20: 721–727, 2012 |
Plný text