The quadruped robot K3lso

Abstract

Abstract Robots have long been sought after to assist humans, and quadruped robots have an advantage in movement over other types. K3lso is a quadruped robot that has been lent to Chalmers University of Technology. This project is a continuation of previous bachelor projects working on K3lso. The main goal of this project is to develop an image processing system for one of K3lso’s cameras and a path planning system for a simulated version of K3lso. The obstacle detection was able to successfully detect objects as long as their heights were not below a decided margin of error. Some problems also arose with shadows cast by the obstacles, creating artifacts. Optimal conditions were found to be a uni formly illuminated room, with matte walls and floors, and no transparent surfaces. Path planning was done using a potential field by creating a vector field with the goal of exerting an attractive force while the detected obstacles exert a repelling force. This allows K3lso to traverse the field to the goal while being repelled from obstacles. From the start to the goal, this calculates a path that the path follower instructs K3lso to follow. Preliminary tests showed a working path that avoided he registered obstacles, but when integrated with the path follower the simulated camera algorithm detects an obstacle where there should not be one. Two controllers were constructed for path following that can walk to points in a curved or straight path. After comparing the two different path-following controllers the path controller was faster at longer distances and had higher accuracy than the curve controller. In conclusion, obstacle detection can detect obstacles that affect the determination of the path, that K3lso then follows.