An evolutionary method to vision-based self-localization for soccer robots

In this paper, a method using an evolutionary algorithm to automatically set-up the color-feature model of an omnidirectional vision system will be introduced. The mentioned method, in addition to avoiding the issue of over-reliance on lighting conditions when the soccer robot is performing image processing, can also very e ectively speed up the parameter setup procedure of the robot vision system. Hence, when the robot is moving in the soccer eld, it can nish target object detection and self-localization in real time. In order to verify the e ectiveness of the mentioned method, tests have been conducted under di erent bad lighting conditions, and the experimental results show that the soccer robot can always set up the parameters of the vision system. It can also set up the color-feature model that is applicable to the operational environment at that moment and detect target objects such as goals and the eld. Meanwhile, through relative location between detected target objects and the robot, self-localization and path planning can be nished.