The Runaway Robot final project for Udacity’s online course, Artificial Intelligence for Robotics.
The project describes a scenario where a robot is lost in the desert, and we are going to use localization, search, path planning and PID controls to track the robot and take it back. We assume that the robot moves in a circular motion on a (x, y) coordinate plane, where it moves for a certain distance, localizes itself, turns a certain angle, and then moves again. Whenever the robot localizes itself, we would get measurements of its position.
The robot is simulated with the program provided by Udacity in robot.py. Udacity also provides a matrix.py to perform basic
matrix computations.
My implementation parts of the project are divided into five guided parts, which will be described next.
In this first part, we assume that the robot’s motion is exact, and its measurements of its positions are also accurate. Based on these assumptions, we would be able to deduce its motion patterns, including the distance between each measurements and the turning angle, after three measurements.
In this part, we assume that the robot’s measurements of its positions have a zero-mean Gaussian noise, while its movements are still exact. I decide to use a Kalman filter to estimate the heading direction, the turning angle and distance traveled between each measurements.
In this part, we administer a robot to start from a location some distance away from the robot that ran away. This pickup robot is faster than the ran away robot, so it can chase the ran away robot easily.
In this part, our pickup robot is the same speed as the ran away robot. As a result, the robot would need some pre-planning and even better localization method to reach the ran away robot. It will use the Extended Kalman Filtering algorithm to reduce the noise from the ran away robot’s measurements, and accurately localize the robot for a fast chase.