Manuela Veloso, Elly Winner, Scott Lenser, James Bruce, and Tucker Balch
Planning actions for real robots in dynamic and uncertain environments is a challenging problem. Using a complete model of the world is not viable and an integration of deliberation and behavior-based reactive planning is most appropriate for goal achievement and uncertainty handling. This paper reports on our successful development of a system integrating perception, planning, and action for the Sony quadruped legged robots. We consider the robotic soccer task, as Sony provided the robots to us specifically for the RoboCup robotic soccer competitions. The quadruped legged robots are fully autonomous and thus must have onboard vision, localization and agent behavior. We briefly present our perception algorithm that does automated color classification and tracks colored blobs in real time. We then brie y introduce our Sensor Resetting Localization (SRL) algorithm which is an extension of Monte Carlo Localization. Vision and localization provide the state input for action selection. In addressing this planning challenge, we created a robust and sensible behavior scheme for the robot that effectively handles dynamic changes in the accuracy of the perceived information. We developed a utility-based system for using localization information. Finally, we have devised several special built-in plans to deal with times when urgent action is needed and the robot cannot afford collecting accurate perception information. We present results using the real robots demonstrating the success of the algorithms. Our team of Sony quadruped legged robots, CMTrio-99, won all but one of its games in RoboCup-99, and was awarded third place in the competition.