Research On Multi-sensor Scheduling Method Of Flying Robot For Power Line Inspection

Overhead transmission line with its expansive coverage is exposure to the complex outdoor environment, vulnerable to severe weather, forest fires and human and animal illegal activities. Transmission line inspection is an effective way of prevention and investigation of power line fault, it is considered to be a practical significant way to guarantee safe and stable operation of power system. Through the inspection, line state can be mastered and anomaly or fault devices can be located so as to be repaired or replaced timely.Power line inspection system using helicopter and unmanned aerial vehicle as carrier employs the joint detect method with all the onboard sensors enabled. Although it can ensure maximum detection capability, it also creates unnecessary costs and data redundancy. With the sensor type increased and function becomes more complex, sensor scheduling work cannot be completed manually. In this paper, a multi-sensor scheduling method is proposed to solve the problem that sensors can’t be scheduled autonomously in the inspection system. Firstly, a top-down multi-sensor scheduling architecture is proposed to divide the scheduling process into three layers which are action planning layer, sensor selection layer and sensor control layer. Layers of the architecture achieve the function of multi-sensor scheduling from different levels, providing general research framework for sensor scheduling. Secondly, a sensor selection model is established using Bayesian networks (BNs), to represent the uncertainty relationship between sensor detection result and the hypothesis of the inspected target, providing basis for the probabilistic inference and sensor selection. Thirdly, a kind of sensor reliability analysis method for inspection task is proposed based on the basic of mutual information, a sensor selection method based on optimization selection criteria within limited cost is put forward and the control of parameters and working mode for the common used airborne sensors is discussed. At last we verify the proposed method with an example the simulation experimental result demonstrates that the proposed method is feasible.