摘要：Derived from look-ahead control in magnetic navigation, a new mode called look-back controlis defined. In this mode, a mobile robot is controlled to move forward tracking a magnetic tape with amagnetic sensor mounted on its rear bottom, while the sensor is mounted on the robot 's front bottom inlook-ahead control. To generate control commands, an estimation of a front bias between the robot's pivotand the magnetic tape is required first. We propose a Grey Qualitative Particle Filter (GQPF) combiningthe grey qualitative theory and the particle filter to process involved multi-modal interval uncertaininformation.An existed motion model is applied to get a prior estimation in the particle filter predictionstage. In the updating stage，the proposed approach utilizes qualitative reasoning and quantitativecomputing to obtain a quantified interval observation, which is merged with the prior estimation to obtaina final estimation of the front bias. A proportion controller takes the estimation as an input to generatecontrol commands. Online experimental results have verified the effectiveness of the proposed approach.
