Research On Autonomous Agents Of Human Operator In Supervisory Control Of Reconnaissance UAVs

In supervisory control system of reconnaissance UAVs,the operator has accomplished the target identification tasks with the help of the automation.In the process of performing tasks,the image information that the operator should process is massive.Not all information is useful to the operator.In addition,the operator has observed the image information from the UAVs for a long time which leads to an increased workload of the operator.The autonomous agent of the operator can preprocess the image information from the UAVs and help the operator perform tasks to alleviate the workload of the operator.Therefore,designing an autonomous agent of the operator is very important for the supervisory control of the multi-UAVs.This paper has researched on autonomous agents of human operator in supervisory control of reconnaissance UAVs.The main work and innovation points of this paper are listed below.(1)This paper designs the autonomous agent of the operator in detail This paper has modeled the image information which includes the suspected targets and the human operator,respectively.The model of the image information including the suspected target is not only related with the types and numbers of suspected targets but also related with the target that the operator is paying attention to.The workload of the operator is bigger when the operator works longer,while the situation awareness of the operator is weaker.So the operator should spend more time on performing the tasks.After understanding the input and output of the autonomous agent of the operator,we design the modules of the autonomous agent in detail.The autonomous agent of the operator is divided into four parts.These four parts are information classification and storage module,task scheduling module,task time allocation module and human machine interaction module.(2)This paper classifies and stores the images including the suspected targets.Information classification and storage module classifies the image information including the suspected target and store it in the database.The classification of the image is determined by the types of the suspected targets.The relevant information of the image includes the location of the targets,the coordinates of the target selected frame,the time and the UAV ID.We put these information into the database in order to manage the information effectively.(3)In the case of uncertain processing times,this paper proposes heuristic scheduling algorithm based on the focused targets of the operator.Based on the target which is paid attention to by operator,we allocate dynamic weights to the image including the focused target in the database.When the operator processes the images,the processing times of the tasks are related with the situation awareness of the operatorand the difficulties of the tasks.So the processing times of the tasks are uncertain but we can make sure of the upper and lower of the processing times in advance.Aiming at uncertain processing time of the tasks and dynamic weights,we have proposed a heuristic scheduling algorithm U1 based on the focused target of the operator.This algorithm can recommend the image with bigger weights and shorter processing time to the operator first.(4)This paper proposes eye movement classification algorithm based on Kalman filter.This paper models the eye movement of the operator.The motion of the eyes is the superposition of an ideal basic motion with constant velocity and white noise.The white noise illustrates the acceleration that is highly time varying.We apply the Kalman filter algorithm to propose eye movement classification algorithm based on the measuring equations of the eye tracker and the model of the eye movement.The eye movement can be divided into three types: fixations,saccades and smooth pursuits.The experiment shows that the algorithm can distinguish the types of eye movement of the operator.(5)This paper proposes the optimal attention allocation algorithm based on dynamic programming.This paper applies the dynamic programming algorithm to gain the optimal solution of the problem of processing time allocation of the tasks.This algorithm takes the correctness of the decision of the operator into consideration,then allocates the optimal processing time to the image and rest time after it.The operator performs the tasks and rest with the suggested time duration.So the workload of the operator falls into an ideal interval.(6).In understanding the structure of the supervisory control system of the multi reconnaissance UAVs,we build a simulation environment of supervisory control system of multi reconnaissance UAVs.