基于自适应UKF算法的小球位置的预测

针对足球机器人比赛过程中,小球位置预测精度不高的问题,提出自适应无迹卡尔曼滤波(AUKF,adaptive unscented Kalman filter)算法。该方法利用新息的协方差匹配技术设计模糊控制器,实时调整量测噪声方差,抑制滤波器的发散;根据滤波值与真实值之间的误差来调节采样点到中心点的位置,修正采样策略,分析初始状态对UKF算法的影响,给出UKF算法初始状态的选择公式。最后通过摄像机获取球在一般滚动和连续碰撞情况下的几组图像验证AUKF算法对球的位置预测的精确性和可行性,提高小型足球机器人比赛决策的合理性。

Forecast Tracking of Ball Based on Adaptive UKF Algorithm

Because the prediction accuracy of ball position is not high in robot soccer competition,the adaptive unscented Kalman filter (AUKF) algorithm is proposed.This algorithm adopts the covariance matching technique of information to design fuzzy controller and achieve real-time adjustment of measurement noise covariance,which can inhibit the divergence of filter.According to the error between filter value and true value,the sample point relative to center point is adjusted and the sample strategy can be amended.By analyzing the influence of initial state on AUKF algorithm,the initial state selection formula of AUKF algorithm is given.Finally,on both conditions of ordinary rolling and continuous collision of ball,the accuracy and feasibility of AUKF algorithm are verified based on groups of images from camera.The proposed algorithm can improve the rationality of decision in robot soccer competition.