Abstract
Pose estimation and trajectory tracking of a spherical rolling robot is a complex problem owing to kinematics and dynamics of the system and the constraint of not being able to add range sensors like ultrasonic or infrared distance sensors on the robot. The pose estimate for the robot under study, needs to be derived purely using inertial measurement unit (IMU) and odometry from analog wheel encoders, which in turn include high uncertainties. Adding to this, the system kinematic and dynamic model to accurately predict the behavior is quite complex. In this paper we present a simplified kinematic model, sensor filtering techniques and the control strategy adopted to locate and navigate the robot to a desired waypoint autonomously. A filter block provides clean heading output from the IMU and incremental pulses from an analog wheel encoder; the pose estimator uses heading and incremental pulses to calculate its position according to the system kinematic model. A pure pursuit algorithm generates left & right wheel velocities to keep the robot on a commanded waypoint, using the robot kinematic model and localization data. The validity of our kinematic model and performance of our waypoint tracking are verified with the ground truth using a motion capture system and onboard sensors, where the application domain is bio-inspired, micro (small scale) robotics.