Fusion of multi-sensor data collected by military robots

This paper addresses the fusion processing techniques of multi-sensor data perceived through IR sensors of the military robots for surveillance, in which they are positioned in a limited range with a close distance between each of the robots.
Nội dung trích xuất từ tài liệu:
Fusion of multi-sensor data collected by military robotsJournal of Automation and Control Engineering, Vol. 1, No. 2, June 2013Fusion of Multi-Sensor Data Collected byMilitary RobotsSanguk Noh and Kyuseon LeeSchool of Computer Science and Information EngineeringThe Catholic University of Korea, Republic of KoreaEmail: {sunoh, cis}@catholic.ac.krempirically and present the experimental results using oursimulator. In conclusion, we summarize our results anddiscuss further research issues.Abstract—This paper addresses the fusion processingtechniques of multi-sensor data perceived through IRsensors of the military robots for surveillance, in which theyare positioned in a limited range with a close distancebetween each of the robots. To combine multi-sensor datafrom distributed battlefield robots, we propose a set offusion rules to formulate the combined prediction frommulti-source data expressed in degrees of reliability for thetype of a target that has the mathematical properties ofprobabilities. We have implemented three fusion operatorsto compare the capabilities of their fusion processing, andhave experimented them in simulated, uncertain battlefieldenvironments. The experimental results show that the fusionof multi-sensor data from military robots can be successfullytested in randomly generated military scenarios.II.We combine multi-sensor data from distributedbattlefield robots. The battlefield robots estimate thetypes of targets using their sensors in a givenenvironment. After getting the sensor data, the multiplerobots inform the control center of their estimations. Thecontrol center then fuses evidence multi-sensed fromdifferent military robots.A. Combined Prediction Using Fusion RulesThe combined prediction given a specific target for thecommander is defined asIndex Terms—Military surveillance robots, Multi-sensorfusion, Techniques for fusion processingI. tk   itk   tjk for k=1, 2, 3, …INTRODUCTIONBattlefield robots for surveillance equipped with IRsensors keep a close watch on moving targets. Thesemilitary robots are semi-autonomously operated; that is,their actions are mostly decided by themselves, butsometimes controlled by their commanders. The multiplerobots periodically scan regions and, when they spot anypossible threats, inform the control center of theirestimations. The control center then fuses evidencesmulti-sensed from different military robots. Thecommander at the control center [1] provides feedbackson the estimations of the multiple robots based upon theresults of fusion processing.Information fusion from different sensors has becomea crucial component in distributed military surveillanceenvironments [2]. In this paper, we focus on theinformation fusion processing that refines the estimationof types for a specific target and improves the reliabilityof its identification, continuously seeking out its positions.We suggest a set of fusion operators [3] to formulate thecombined prediction from multi-source data expressed indegrees of reliability for the type of a target that has themathematical properties of probabilities.In the following section, we will describe how tocombine multi-sensor data from military robots forsurveillance. In Section III, we validate our frameworkCOMBINING MULTI-SENSOR DATA FROMDISTRIBUTED ROBOTSwhere  itk and  tjk represent the confidence of thepossible type of a specific target, tk , from a robot iand a robot j, respectively;t 0  itk and  jk 1;tt  ik  1 and also   jk  1 .kkWe propose a set of fusion rules to formulate thecombined prediction from multi-source data expressed indegrees of reliability for the type of a target that has themathematical properties of probabilities. Givenconfidence values of  itk and  tjk for k=1, 2, theaggregation operators,   {1, ,n } , in this paper,are as follows: Mean (1): tk Product (2): tt ( ik   jk )/2 ;tktt  ik   jk ; Dempster-Shafer theory [4-6] (3): tk Manuscript received October1, 2012; revised December 22, 2012.©2013 Engineering and Technology Publishingdoi: 10.12720/joace.1.2.95-98(1)95 itk   tjk.tttt1  ((1   ik ) jk   ik (1   jk ))Journal of Automation and Control Engineering, Vol. 1, No. 2, June 2013TABLE I.The combined prediction representing the overalldegrees of belief on the type of a specific target can beobtained by applying aggregation operators to multisource data. The goal of fusion processing is to combinethe estimations from distributed military robots wheneach of them estimates the probability of reliability on thetype of a target, and another goal is to produce a singleprobabilitydistributionthatsummarizestheirprobabilities.Among the aggregation operators, the mean operatorsimply extends a statistic summary and provides an itk = {0.60, 0.10, 0.20, 0.10} tjk = {0.70, 0.20, 0.05, 0.05}ttMean (1)Product (2)Dempster-Shafer (3)Mean (1)Product (2)Dempster-Shafer (3)tand 0.778) of the combined prediction are much biggerthan the other combined values (0.020 and 0.027, 0.010and 0.013, 0.005 and 0.006), compared with the originaldistributions of their estimations. Normalizing thecombined prediction ˆtk , as defined in (2), makes theresulting values of  tk ’s indicate the degrees ofconfidence values on types of a target being comparedwith each other in the range of 0 and 1.agreement on different robots’ probabilities of reliabilityon the type of a target; however, they completely excludethe degrees of disagreement or conflict. The advantage ofusing the Dempster’s rule in our fusion processing is thatno priors and conditionals are needed.The normalization of combined prediction is given astkIII.EXPERIMENTATIONWe have implemented an individual fusion processusing the aggregation operat ...