Dynamic Estimation of Model Parameters for In-Pipe Walking Robots

In this paper, the problem of the model parameter estimation for in-pipe walking robots has been studied. Three estimators (regressor function-based method and the least squares-based methods with and without damping) are studied and analyzed. The study is focused on the influence the sensor resolution has on the accuracy of the estimation. The influence the accuracy of the initial estimation has on the methods' performance was also discussed. Two performance functions are proposed in order to simplify the analysis. It was shown that the least squares method has a significant advantage when the exact information about the robot's state is known; however, when the sensors have finite resolution, the performance of this method changes significantly while the regressor-based method is affected much less.