TargetDetection and Identification th a Heterogeneous SensorNetwork by Strategic ResourceAllocation and Coordination
Minghui Li, Yilong Lu · 2006
In thispaper, anapproach fortarget acquisition and identification witha heterogeneous sensornetworkthrough strategic resource allocation andcoordination ispresented. Based onsensor management technologies, lowperformance sensors are deployed toguideandcuescarce highperformance sensors inthe networktobestimprove thedataquality. We focuson the problemofdesigning suchanetwork system inwhichissues of resourceselection and allocation, systembehaviorand performance, and multiple constraints must be addressed simultaneously. Simulation results offer significant insight into sensorselection andnetworkoperation, anddemonstrate the benefits introduced byguided search inanapplication oftarget localization andidentification onthebattiefield. I. INTRODLCTION Withtheadvaneements intechnologies, itisbecoming increasingly feasible toconceive anddeploy large-scale sensor netxvorks forawiderange ofapplications includinLg elderly assistance, traffic control, homelandsecurity, military surveillance, envronmental monitoring, and etc.(1). Significant workhasgLone intodevelopment ofalgonrthms to perform avariety oftasks including detection, localization, classification, identification andtracking ofoneormoretargets inthesensorfield, andnumerousapproaches basedon collaborative signial andinformation processing (CSIP) have beeni proposed intheliteratures, see,e.g., (2), (3). Inthe scenarios involving multiple targets, dataassociation of measurements from multiple sensors isknowntobeNP-hard. Multiple hypothesis tracking (NMT)(4)antd Markovchain MonteCarlo dataassociation (MCMCDA)(5)arepossible solutions tothis tough problem atcosts oflonglatenc and extensive computation. Inthis paper, weaddress theproblem oftarget identification andlocalization ina heterogeneous sensor network from another aspect. Through sensor management technologies and strategic resource allocation andcoordination, i.e. placing the right sensor attheright location ontheright time, thedata quality isimproved tothemaximumextent. We takeasa sample tasktheproblem ofdetecting andidentifying enemy vehicles onthebattlefield', withasensor network consisting of unattended ground sensors (UGS)antd unmannedaerial vehicles (UAV). A UGSmaybeonesensor nodeequipped withcollocated multi-muodality sensors ora cluster of distributed sensors, butassume tobesimple infunction, cheap, static butrapidly deployable. Thelocation ofeachUGS is assumed tobektnown inadvance obtained froim on-board GPS receivers ora self-localizatio n