Development of a multisensor-based residual power prediction system for mobile robots

Based on the measurements of multiple sensors, this paper presents a residual power detection system for a mobile robot. We use four current sensors to measure the current variety of the mobile robot, and use multilevel multisensor fusion method to detect and diagnosis current sensor and voltage signals status. Moreover, a two level method is used to isolate faulty measured value such that more exact current status to be obtained. In this method, a redundant management method and a statistical prediction method are used in levels one and two, respectively. We use the same method to measure voltage of power system for mobile robots, and isolate faulty measured values. We design the power detection and isolation module using HOLTEK microchip according to the redundant management method. This module can transmit measured value and decision output to main controller (IPC) using series interface (RS232). However, it is possible that this method is faulty. In this case, the IPC can decide an exact power measured value according the statistical signal prediction method. Then we can predict the residual power of mobile robots using polynomial regression algorithm. Finally, we implement the proposed method on the experiment scenario of can set a power threshold value to calculate the critical time for the mobile robot. Meanwhile, experimental results are given to show the feasibility of the proposed method.