基于置信度的自适应Kalman滤波定位方法

Self-Adaptive Kalman Filtering Algorithm Based on Confidence Level

针对室外自主移动机器人因感知信息缺失或异常波动造成的定位失败或偏差过大的问题,设计了一种基于置信度的自适应Kalman滤波定位方法。该方法根据置信距离和置信度函数计算局部滤波器的置信度,并将其进行加权运算后作为全局滤波器的自适应因子,以得到更为准确的位姿估计。采用D-S证据理论对全局位姿进行评价,并给出单一传感器失效时的组合方案。试验结果验证了方法的有效性和鲁棒性,能满足实际定位的需要。

The problem of positioning failure or excessive deviation of outdoor autonomous mobile robot often appears due to missed perceptional information or abnormal fluctuations. A self-adaptive federated Kalman filtering method based on confidence level was given in this paper. This method firstly defines confidence distance and the confidence function of the sensor data consistency, according to which confidence level of local filter can be calculated. Finally, we carry out weighted computing for confidence level which can distribute an appropriate adaptive factor for global filter , therefore, a more accurate pose estimation is obtained. In order to ensure the safely operation of robot, D-S theory is introduced to evaluate the global pose and give some kind of portfolio programs when a sensor is failure. Experimental results prove that this method is effective and robust and it also can meet the needs of the actual location.