The Airborne Multi-Sensor Target Threat Assessment Based On Data Fusion

It has been an ever more obvious tendency that the evolution andformulation of modern air combat strategies and tactics are largelydependent on the level of informationization, and as an example, thevarious battlefield application researches, based on the airbornemulti-sensor fusion mechanism, become available due to the accelerationof the operational speed in computers, the improvement of the anti-jamcapabilities as well as the throughput of communication links, theintelligentization of software algorithms, etc. Different from thesingle-sensor methods, the multi-sensor based system can effectivelysynthesize all information sources, handling imprecision along withuncertainty and even analyze contradictory scenarios. This thesis presentsthe study of an effective modeling of the threat assessment of enemytargets in air combat, by applying data fusion measures on the airbornemulti-sensor complex system, and concentrates on the following aspects:(1) A review of the main task of the airborne multi-sensor data fusionsystem, data fusion needs and objectives. The establishment of a generalairborne multi-sensor model of a fighter is proposed for the particularity ofthe fighter avionics, based on D-S theory (evidence theory).(2) The systematic introduction of the basics of D-S theory, especiallythe analysis of one of the three key issues in this area, namely thecombination of evidences under high evidence conflict, which might bederived from the disturbance on sensing devices or the incompleteness ofFOD (frame of discernment). Previous efforts take for granted the integrityof FOD, however, the proposing of GET (generalized evidence theory)suggests a new way of solving the high conflict problem when the FOD isincomplete. To overcome the minor flaw of the first edition of GET, thispaper defines a new combination rule and proposes the risk reducingdiscount factor method to modify the model.(3) The construction of target threat evaluation system by furtherapplying D-S theory, as an extension of the already done airbornemulti-sensor target recognition platform. This paper chooses the top-fivesignificant features (identity, distance, path angle, velocity, electronicjamming) of flying objects as evidence sources, and brings forward the non-homologous BPA (basic probability assignment) generation methods;meanwhile AHP (analytic hierarchy process) is adopted to quantify theweights of the above criteria in accordance with their contributions tothreat, then followed by evidence fusion using the newly presentedweighted average combination rule to obtain the final threat in BPApresentation; and lastly, PPT (pignistic probability transformation) and theweighted sum conversion helps to produce a deterministic threat value ofan flying object.(4) Two simulations devised to validate the effectiveness of proposedmodel. The first experiment runs according to the scenario presented byChengdu Aircraft Design Institution to test100aircrafts, the result satisfiesthe availability and real time requirements of ranking those aircrafts basedon threat degree. The simulation experiment of two airplanes (foe and own)heading opposite is established for the comparison of three evaluationalgorithms. As the result indicates, the proposed algorithm and the generalweighted method are able to effectively assess the threat situation ofbattlefield while the conclusion from DS/AHP contradicts our commonsense due to its wrong adoption of Dempster combination rule. Thissimulation analysis，also as a counter-example, can put into question theDS/AHP’s modification on the classical AHP.