Omnidirectional vision-based robot localization on soccer field by particle filter

An omnidirectional vision-based localization method based on the particle Alter is proposed to achieve the location of robot on a soccer field in this paper. Two kinds of sensor information are considered in the method to let the robot on the soccer field can estimate its location and then to decide an appropriate strategy. One is the robot action sensor information obtained by the motor´s feedback and the other is the observation sensor information obtained by the image captured by an omnidirectional vision system. The action sensor information is used to expect the robot location distribution. The location distribution is represented by particles. The omnidirectional image is used to observe the environment information. The differences between these particles location´s environment information and the robot observation environment information are considered to calculate the belief values of particles. Then the posture of the particle with the highest belief is used to be the estimated posture of the robot. Some experimental results are presented to illustrate the effectiveness of the proposed method.