Resource Allocation Methods For Target Tracking Applications In Cognitive Radar Networks

With the development of military technology,the detection environment of radar is be-coming more and more complex,and the tracking task is also more challenging.The complex environment is mainly reflected in the following aspects:（1）There are many types of mod-ern stealth,high-speed,and low-altitude targets,which makes the target environment more complex?（2）To detect long-distance weak targets,the radar needs to increase the transmit power,but the accompanying strong clutter will degrade the detection performance?（3）The emergence of new jammers and various jamming styles makes radar face a complex elec-tromagnetic environment.In a complex detection environment,it is difficult for traditional monostatic radar to meet the continuous detection and stable tracking requirements of non-cooperative targets,due to its simple adaptive signal processing mode.Although cognitive radar can improve the performance of traditional radar,the performance improvement of monostatic radar in terms of coverage and tracking accuracy is limited.It has become a consensus to improve target detection performance through radar networking.However,the detection of each node in the traditional radar network is relatively independent,and the ad-justment of the working mode and parameters is not flexible enough.The fusion center only performs simple comprehensive processing of the track reported by each node（selecting the main station or weighted summation）,and the resources of the system such as multi-signal and multi-processing modes are not fully utilized,so the degree of synergy is low.Therefore,it is urgent to transform the traditional open-loop radar network into a closed-loop cognitive radar network to meet the needs of target detection in complex environments.The performance improvement mechanism of cognitive radar networks mainly includes the following three aspects:（1）Optimizing the deployment of radar node or platform move-ment trajectories to improve the benefits of space diversity,thus comprehensively perceiving the geographical and target environment?（2）Controlling the limited resources of each node to match multiple tasks and environmental changes by using the prior information fed back by the fusion center,thus fully extracting the target information from the echo and providing more useful points of target?（3）Developing efficient data fusion algorithms to estimate and predict the states of targets to guide dynamic resource allocation.Therefore,how to auto-matically plan the motion trajectory of multiple platforms,reasonably allocate the limited resources of each node to achieve the best match of the target environment,and improve the performance of fusion algorithms are the keys to maximizing the detection performance of the system,and it is also the bottleneck of performance improvement in cognitive radar net-works.To improve the target tracking accuracy,target capacity,and resource utilization rate,this paper establishes several closed-loop tracking frameworks with the perception of the tar-get environment online,trajectory planning of platforms,and allocation of system resources.The main contents of this paper are as follows.（1）Aiming at cognitive multiple target tracking（MTT）problems in active radar net-works,two power allocation schemes are studied under different backgrounds in this paper.In the background of low interception,a power allocation model constrained by multi-target tracking with guaranteed performance is constructed,the feasibility and separability of the model are analyzed,and a comprehensive solution framework is designed to save the total power of the system.On this basis,aiming at improving the target capacity of active radar networks under the background of saturation attack,a new optimization objective function is introduced,then a non-convex and non-smooth power optimization model is constructed in this paper.Subsequently,a three-step solution method is proposed according to the analysis of model characteristics,and the physical mechanism of the three-step method to increase the target capacity is analyzed.The three-step method first relaxes the model through the Sigmoid function,then uses the block coordinate descent method combined with the proxi-mal inexact augmented Lagrange multiplier method to solve relaxation problems,and finally compensates the relaxation error based on the power allocation model established in the low interception scenario.Compared with the genetic algorithm and the pattern search method,the target capacity of the three-step method is the largest.（2）Aiming at the channel allocation problem for maneuvering target tracking in pas-sive radar networks,the tracking performance measure,namely the best-fitting Gaussian pre-dicted conditional Cramér-Rao lower bound is derived,and two adaptive channel allocation models are formulated in this paper.The first channel allocation model aims to maximize the maneuvering target tracking accuracy with a small number of available channels,and the second channel allocation model aims to minimize channel consumption while satisfying tracking accuracy constraints.For the first channel allocation model,the l₂-Box alternating direction multiplier method is developed to solve it based on the idea of equivalent substi-tution of Boolean variables.It avoids the error introduced by the linear convex relaxation algorithm and obtains an approximate optimal solution quickly.A multi-start heuristic al-gorithm is proposed to solve the second channel allocation model.This algorithm considers the contradiction between the given tracking accuracy requirements and the available re-sources,thus providing a comprehensive solution scheme.Both models and corresponding algorithms reduce the computational load and communication requirements involved in cen-tralized processing.（3）Aiming at the cognitive MTT problem in the active and passive radar networks,a tracking performance measure based on heterogeneous measurement information is derived,and an optimization model concerning the joint transmit resources of the active radars and the receiving processing resources of the passive radars is constructed in this paper（mixed inte-ger nonlinear programming problem）.This paper analyzes the physical meaning of resource allocation,develops a composed resource optimization scheme,and effectively improves the overall MTT performance.The composed resource optimization scheme first introduces two auxiliary vectors to rewrite the original problem as an equality constraint one,then uses the alternate direction multiplier method to alternately solve several simple sub-problems,in which the sub-problem concerning the resource vector is convex,and the sub-problem with respect to the auxiliary vector are separable.The results show that the performance of the composed resource optimization scheme is better than that provided by the traditional uniform allocation scheme and the convex relaxation method,and is close to the optimal performance provided by the exhaustive method,but its computational complexity is much lower than the exhaustive method.（4）Compared with the ground-based radar network,the airborne radar network has the advantages of strong mobility and a wide field of view.Aiming at the cognitive MTT us-ing multiple unmanned aerial vehicles（UAV）equipped with different types of radars,the path planning of passive airborne radar network in the background of reconnaissance,and the joint path planning and resource allocation of active airborne radar network in the back-ground of detection are carried out in this paper respectively.In the context of reconnais-sance,the timing of target radiation signals is random,so the arrival time of angle mea-surements of different targets is asynchronous concerning the same payload.As a result,this paper first constructs a composite measurement according to maximum likelihood esti-mation on asynchronous measurements,and then the Kalman filter is used to estimate and predict the target state.The fusion center evaluates the impact of UAV positions on tracking performance based on the predicted information,then establishes a trajectory optimization model combining maneuver,altitude,collision avoidance,obstacle avoidance,and threat avoidance constraints.According to the time-varying characteristic of security constraints,a hierarchical solution algorithm is designed,which can quickly obtain the near-optimal MTT performance.On this basis,a joint path planning and resource allocation problem of active airborne radar networks in the detection background is further considered,and the overall MTT performance is improved by optimizing the power,beam,and position of multiple air-borne radars.The effectiveness of joint path planning and resource allocation in airborne radar networks is verified by simulation.