Heading Estimation Of Humanoid Robot's Electronic Compass

Electronic compass is influenced by external magnetic field, for example, power noise, external magnetic's influence, iron material's influence and so on. In addition, while the humanoid robot is walking, the body of the humanoid robot is swing the left and right. As a result, errors exist in electronic compass's measure data. According to the error's nature and trait, measuring errror is divided into random error, system error and thick error. In accordance with the random error, particle fliter method was used to decrease the random error in order to improve the measure precision of hunmanoid robot's electronic compass.Bayesian filter theorem, Kalman filter and particle filter were expounded firstly in this paper, and the strong points and the shortcomings of Bayesian filter and Kalman filter were analyzed. Bayesian importance sampling and sequential importance sampling were introduced in detail, and the reason of particle filter's degradation phenomenon was analyzed. The solving strategies of particle filter's degradation problem were given. Secondly, in view of costs, schedule index and using range synthetically considered, two-dimensional electronic compass and two-dimensional tilt sensor were chosen as navigation system of humanoid robot. The frame chart of humanoid robot's navigation system was given, and the working principle and function of several kinds of main chips were introduced in detail. The main program flow chart is given, and each subprogram module's function were introduced. Finally, the reasons of error of electronic compass were analyzed. In accordance with the random error, particle filter was put forward to estimate the heading angle, and Kalman filter was put forward to estimate the heading angle. Effect was compared through the experiment, and the result proved that particle filter algorithm was better and the measure precision of electronic compass can reach plus and minus one degree or so.