Study On The Key Technologies In Behavior Representation For Agent-based Pilot-Aircraft Combination

In virtual battlefield,it is the urgent need and one research hotspot for military forces around the world to simulate various fighter aircrafts based on CGF(Computer Generated Forces)with high fidelity.However,compared with the pilot-in-loop simulation system,current CGF rarely considers the human factor,so that it faces the problem of simpler function,less intelligence,and lacking fidelity and credibility.To solve these problems,agent-based PAC(Pilot-Aircraft Combination)behavior representation method is proposed from the perspective of the pilot,and from deep inspection into functional features and operation process of modern advanced fighter aircrafts.By building the model architecture,deep study on the key technologies in behavior representation for the pilot and the aircraft,the PAC model is developed and demonstrated,along with high fidelity simulation of a typical fighter aircraft.The main research results and innovations are as follows:(1)To solve the problem of low credibility owing to ignore the pilot-aircraft interaction,agent-based PAC model architecture is presented,which decomposed the CGF for a fighter aircraft into two parts: aircraft-platform-agent and pilot-agent.Based on MAS(Multi Agent System)modeling approach,aircraft-platform-agent is further decomposed into behavior representations of multiple sub-system agents.Pilot-agent applies a layered hybrid architecture,which expands it into four layers of behavior model: awareness process,situation assessment,decision plan,and operation control.The PAC model has the advantages of modularity,combinability,high fidelity,and easy extensibility,which lay the foundation for its inner algorithm research and final engineering implementation.(2)For aircraft-platform-agent,pilot-agent in-loop six degree of freedom aircraft motion model and flight control model are built,and the fire control computing model of mid-range intercept air-combat is established based on lead attack aiming approach.On the basis of the fire control model,a new approximate computation method based on aiming error is proposed to acquire the attack area of missile.These agent models for the key sub-systems are created with high fidelity in real-time simulation,and realize the sharing of aircraft-platform-agent between constructive simulation and pilot-in-loop simulation.(3)To improve the human-like fidelity of pilot-agent,fuzzy logic approach is applied to pilot cognitive behavior representation.On the layer of awareness process,a new situation awareness model is presented base on attention-situation awareness model from NASA(National Aeronautics and Space Administration),and the model is established base on the situation awareness model and FLS(Fuzzy Logic System)which uses type-1 and type-2 LFS to process information with different certainty and producing fuzzy cognition conformable to real pilot.On the layer of situation assessment,the model is established by combining DBN(Dynamic Bayesian Network)and the type-1 and type-2 fuzzy inputs from the layer of awareness process,and IFDBN(Interval Fuzzy Dynamic Bayesian Network)based situation assessment method is presented to simulate the assessment behavior of pilot-agent under high uncertainty environment.On the layer of decision plan,a two-layer task-goal-oriented FPN(Fuzzy Petri Network)considering constraint and supporting situations is built for decision model of pilot-agent,and BP(Back Propagation)algorithm based method is used to learn the parameters of this FPN which is difficult to determine artificially.(4)On the layer of decision plan for pilot-agent,the model is built from discrete control and continuous flight control.By taking task goals and action sets from the layer of decision plan as the driver,discrete control model is built based on action priority and workload.By employing proportional guidance and fuzzy controller,continuous flight control model is built.To simulate the predictive control ability of a proficient pilot,MPC(Model Predictive Control)based flight control method is studied,and MPC-FLS approach and MPC-SAPSO(Simulated Annealing Particle Swarm Optimization Algorithm)approach are presented to realize variety of flight control modeling and simulation.(5)To demonstrate the feasibility and validity of PAC model,special test-bed is developed.By embedding pilot-in-loop simulation system into the test-bed,it can test the function of aircraft-platform-agent and support the design and test of pilot-agent.Through modular design in vsTasker software,the CGF entity for PAC is created,and the models for aircraft-platform-agent and pilot-agent are implemented.Finally the simulation scenario of mid-range intercept air-combat is designed and be used to test the aircraft-platform-agent,pilot-agent and PAC model.By comparing the test results between PAC model and pilot-in-loop simulation,the feasibility of modeling and design method for PAC is validated,and it shows that the PAC model has achieved higher fidelity and credibility.