The Research Of Key Issue For Collision Avoidance Under Unstructured Environment

With increasing complexity,the existence of stochasticity and uncertainty creats new challenges for traditional obstacle state estimation and collision avoidance in unstructured environments.Meanwhile,in the case of system model unknown/partially unknown,introduction of stochasticity and uncertainty into the system model in the form of disturbance and other parameters is inapplicable,which is inconducive to the collision avoidance as well.How to face such challenges not only occupies the research frontier,but also plays an important role in applications of obstacle state estimation and collision avoidance.In this dissertation,we carry out the following researches including the collision avoidance path planning,motion state estimation,route classification and route abnormality detection.1.For the case of motion parameter incompleteness in motion estimation using bearing-only measurement,Kalman filter system is limited by its dependence on prior known motion model.A bearing-only unknown input observer(UIO)is constructed to obtain state estimations of obstacles.The UIO we proposed is suitable for general motions since motion models are not necessary during process of estimating. Simulations in single-obstacle and multiple-obstacle scenes show that the accuracy of the UIO is higher compared with the M-estimator,and the combination of UIO and ADAPF can achieve destination faster with collision avoidance.2.For non-cooperative target route attribute discrimination subject to uncertainty resulting from tracking error,we propose a multi-feature reasoning based track-to-airline association method(MFR).Firstly,according to the requirements of air traffic control system for airway and flight an air target to airline association model is developed.Secondly,the mass functions of distance and direction based on probability-distribution function and of chord length based on linear function are given and then those evidences are combined by combination of weighted belief functions based on evidence distance and conflicting belief.Besides,the meta-class is induced to avoid the misclassification.The proposed method is shown valid via both simulation data and real data of ATC radar system.3.For the anomaly detection of the disguised route and route attributes switching,we propose an alternative anomaly detection method in the framework of transferable belief model(TBM)theory.The novelty of this work is that it can detect both unreliable evidence source and abnormal behavior of the targets within our architecture by using a temporal analysis and a new discounting coefficient through introducing the concept of contribution degrees of features.Detection of abnormal behavior is based on a prediction/observation process and the influence of the faulty sources is weakened by discounting coefficients.The simulations show the feasibility and effectiveness of the proposed algorithm.4.For the collision caused by the artificial field potential algorithm,the sufficient condition is found with revealing the essential limitation of the repulsive force calculation.A new concept called satisfaction factor is proposed.The satisfaction factor that meets the three collision avoidance criteria,theoretically avoid the collision by modifying the repulsive force.Since the value of the satisfaction factor changes real-time with the relative angle among the destination,the obstacle and the motion platform in different scene,the ADAPF method we proposed possesses a stronger self-adaptive ability.Simulations in typical scene show that ADAPF can effectively avoid collisions while collisions take place in the standard artificial potential field method.