基于自适应交互多模型算法的三维空间水下动态目标跟踪

Three-dimensional Underwater Dynamic Target Tracking Based on Adaptive Interactive Multi-model Algorithm

自主水下机器人(AUV)动态目标跟踪技术是实现目标探测、目标侦察等任务的核心技术之一。为了跟踪机动目标,通常采用基于交互多模型(IMM)算法结合恒定速度(CV)模型和协同转弯(CT)模型;而IMM中的转移概率和CT模型中转弯速率通常根据先验信息固定,可能会导致状态估计不准确。为此,文章基于现有的自适应IMM算法,提出了一种可以自适应调整转移概率的并行IMM算法(APIMM)并结合无迹卡尔曼滤波算法(UKF)对水下三维空间中的机动目标进行状态预测,改进算法基于的模型集选择了CV模型,自适应转弯速率的三维固定中心恒定速率和转向速率(CSCTR)模型和当前统计(CS)模型。仿真结果表明,该算法能更大程度地利用后验信息,拥有更快的模型切换速度,能够对三维空间水下动态目标的状态进行预测,并且预测精度提升了约15%。

Dynamic target tracking is a crucial technique for autonomous underwater vehicles(AUV),enabling key operations such as target detection and reconnaissance.Standard approaches for tracking maneuvering targets often involve the interacting multiple model(IMM)algorithm which integrates the constant velocity(CV)model and the coordinated turn(CT)model.However,these models traditionally utilize fixed transition probabilities and turn rates based on prior information,potentially leading to imprecise state estimations.In response to this,the paper introduces an adaptive parallel IMM(APIMM)based on current adaptive IMM algorithms.This method adaptively adjusts transition probabilities and pairs with the unscented Kalman filter(UKF)algorithm for state prediction of maneuvering targets in a 3D underwater environment.The enhanced algorithm chooses from a model set that encompasses the CV model,the 3D fixed center constant speed and turning rate model(CSCTR)with an adaptive turn rate,and the current statistical(CS)model.Simulation outcomes have demonstrated that this algorithm utilizes posterior information more effectively,possesses an accelerated model switching speed,and improves the prediction accuracy of the underwater dynamic target's state in three-dimensional space by approximately 15%.