CN109709535B - A beam dwell scheduling method for cooperative distributed systems - Google Patents

Abstract

The invention belongs to the field of radar system resource management, and particularly relates to a beam resident scheduling method for a cooperative distributed system. The method comprehensively considers the influence of the priority and the deadline of a task working mode on the task priority, utilizes an HPEDF method to calculate the comprehensive priority, establishes a beam resident scheduling optimization model according to the characteristic of the multi-station cooperative radar task scheduling in a cooperative distributed system, and solves the problem by using a heuristic method based on time pointer analysis, thereby forming the beam resident scheduling method for the cooperative distributed system. The invention solves the problem of beam resident scheduling of a cooperative distributed system in which the sub-radars can work independently and can work cooperatively when needed.

Description

A beam dwell scheduling method for cooperative distributed systems

technical field

The invention belongs to the field of radar system resource management, in particular to a method for implementing multi-station cooperative adaptive beam dwell scheduling in a distributed system.

Background technique

Due to the threats of stealth targets, anti-radiation missiles, and low-altitude targets, the air defense guidance radar with single-station radar as the core is difficult to deal with the complex battlefield environment and is encountering a bottleneck in development. A distributed system composed of multiple single-station radars can utilize the diversity of spatial distribution, thereby improving the combat system's anti-strike capability. Therefore, the concept of distributed systems is constantly being studied. The distributed system is to collect and transmit the information of various radars in the system in a "mesh" form by properly deploying multiple radars with different systems, different frequency bands and different polarization modes, and comprehensively process, control and transmit the information from the central station. management to form a unified organic whole. The distributed system firmly grasps the key of data fusion, and fuses the information transmitted by each radar, so that many information that cannot be obtained by a single radar can be obtained. According to different data fusion methods, common centralized processing and distributed processing systems are formed.

The centralized processing system sends the data of each radar node to the central processing unit for fusion processing. This method can realize real-time fusion. Its data processing accuracy is high, and the solution is flexible. The disadvantage is that it requires high processor requirements and has low reliability. The amount of data is large, which is difficult to achieve; in the distributed processing system, each sub-radar uses its own measurement to track the target independently, and sends the estimation result to the headquarters. The communication bandwidth requirement is low, the calculation speed is fast, and the reliability and continuity are good, but the tracking accuracy is not as high as the centralized one.

The collaborative distributed system adopts the data fusion method of distributed processing. Multiple sub-radars can not only independently perform target detection and task tracking, but also work together when needed to jointly complete a certain target detection or task tracking. Improve tracking accuracy. Since the collaborative distributed system includes multiple radar nodes, each radar node may need to perform multiple tasks at the same time. However, the time and energy resources of each radar node and the entire system are limited. In order to avoid conflicts between the radar nodes during the execution of tasks , to give full play to the performance advantages of the collaborative distributed system and improve the utilization of radar resources, it is necessary to implement effective beam dwell scheduling. The beam-dwelling scheduling of the cooperative distributed system involves both the beam-dwelling scheduling of a single radar and the coordination and scheduling of multiple radars.

At present, relatively mature results have been achieved in the research on beam dwell scheduling of phased array radars at home and abroad, which has become the basis for studying the problem of beam dwell scheduling in distributed systems. In phased array radar beam residency scheduling, the main factors affecting its performance include: assignment of task priority and selection of scheduling strategy. In the task priority allocation, the traditional priority allocation method and the comprehensive priority allocation method are formed. The traditional priority method only considers a single attribute of the task, mainly including: work method priority method and deadline priority method. In the comprehensive priority allocation method, two factors of work mode priority and deadline are usually considered comprehensively, such as: Modified Highest Priority First (MHPF), which maps the work mode priority and deadline to the same level and Comprehensive priority calculated using linear weighting.

In terms of scheduling strategies, there are currently two typical scheduling strategies: template-based and adaptive scheduling. Theoretical research and practice have proved that in the multi-task environment, the adaptive scheduling method is the most effective scheduling strategy, which can give full play to the performance of the phased array radar. The adaptive scheduling strategy mainly includes heuristic scheduling method and intelligent optimal scheduling method. The heuristic scheduling method can be divided into scheduling analysis based on scheduling interval (scheduling analysis) and scheduling analysis based on time pointer. Intelligent optimization algorithm is a scheduling method with global optimization ability and can find the best scheduling sequence. Among them, based on genetic algorithm is the most widely used. In addition, the introduction of pulse interleaving technology in the scheduling strategy can further improve the utilization of time resources.

Regarding the research on distributed system beam residency scheduling, there are still relatively few research results at home and abroad. Some foreign countries have proposed a hybrid Bayesian-based scheduling method and power allocation method for multi-target tracking cognitive radar networks; and a variable-parameter task scheduling method for distributed systems, but this method is only equivalent to The simple superposition of single-station radar cannot reflect the synergy between radar nodes in the system. In China, Guo Wenzhong et al. proposed a heuristic discrete particle swarm algorithm to solve the task scheduling problem of distributed sensors, and experiments proved that its performance was better than traditional genetic algorithm; Wan Kaifang et al. proposed a method based on partial observability. Markov decision process method, the simulation results show that the proposed method can realize the efficient management and scheduling of multiple passive radars; Wang Qiang from the University of Electronic Science and Technology of China also planned the deterministic tasks and uncertain tasks in the beam dwell scheduling of multi-functional networked radars method was studied. However, none of the existing technologies can solve the problem of beam dwell scheduling in a cooperative distributed system in which sub-radars can work independently and cooperate when needed. The research on beam residency scheduling of cooperative distributed systems is still in a blank state at home and abroad.

SUMMARY OF THE INVENTION

In view of the above-mentioned problems or deficiencies, in order to solve the problem of beam dwell scheduling in a collaborative distributed system in which sub-radars can work both individually and cooperatively when needed, the present invention provides a beam dwell scheduling for a collaborative distributed system. method.

The technical scheme of the present invention is specifically as follows:

Step 1: Suppose there are M sub-radars in the system, and there are N request tasks in the scheduling interval [t ₀ , t ₀ +SI], expressed as T={T ₁ , T ₂ ,...,T _N }, these tasks The earliest execution time is less than t ₀ +SI, and the latest execution time is greater than or equal to t ₀ ; among them, t ₀ is the starting time of the current scheduling interval, and SI is the duration of a scheduling interval; it is known that the ith task model is T _i ={ _{pi ,rt i ,d i , li ,} τ _Bi ,id}, from which we can directly obtain the working mode priority _{pi of the task, the expected execution time rt i , the deadline d i , the} time window _li , The dwell time τ _Bi , id is the target identifier, and id={1,2,...,M,M+1}, where id=γ indicates that the target is detected or tracked by the γth radar alone, γ=1,2, ...,M, and id=M+1 indicates that the target is jointly detected or tracked by M radars, that is, the headquarters task has id=M+1, and the non-headquarters task has id=γ.

其中γ＝1,2,…,M；Initialization operation: initialize the time stamps tpγ of the M radars respectively and tpγ≥t ₀ , set tpγ_initial=tpγ, and initialize the registers of the scheduling time and the task completion time:

where γ=1,2,...,M;

Step 2: Check whether there is a headquarters task in the request queue, if so, go to Step 3, otherwise go to Step 4;

Step 3: Schedule and analyze the headquarters tasks:

Step 3.1: Let tpγ=max(tp1,tp2,…,tpM),γ=1,2,…,M respectively;

Step 3.2: Check the latest execution time of the headquarters task, send the tasks less than tp1 to the deletion queue, and delete these tasks from the request queue;

Step 3.3: Investigate the earliest executable time of the remaining headquarters tasks, and select the tasks that are not greater than tp1, assuming that there are N1 in total;

If N1=0, update parameters: tpγ=tpγ+Δt, γ=1,2,...,M, where Δt is the minimum sliding step size of the time stamp, go to step 3.5;

If N1≠0, calculate the comprehensive priority of these tasks according to formula (1), select the task with the highest comprehensive priority, and denote it as T _j :

ps _i =[η·Npi +(N1+2−η)·Nd _i ]/ ₍ N1+1) (1)

where i=1,2,...N1; the N1 request tasks are sorted according to the working mode priority p from low to high and the deadline d from far to near, _{Npi and Nd i are the tasks T i in these} two The position in the sequence; η=(N1+1)/2; ps _i is the comprehensive priority of the task Ti _, the larger the value, the higher the comprehensive priority;

Step 3.4: Send T _j to the execution queue, schedule and execute the task at the antenna front end of the γth radar at time tpγ, delete it from the task request queue, and store the scheduling time in the time storage of each radar: time_tpγ=[time_tpγ,tpγ], then update the time stamp of each radar: tpγ=tpγ+τ _Bj , and store the task completion time in the time storage of each radar: time_tpγ=[time_tpγ,tpγ], where γ=1, 2,…,M;

Step 3.5: If tp1>t ₀ +SI, go to step 4, otherwise return to step 3.2;

Step 4: Let γ=1;

Step 5: Check whether there is a task performed only by the γ-th radar in the request queue, if so, go to Step 6, otherwise go to Step 7;

Step 6: Scheduling analysis for tasks performed only by the γth radar:

Step 6.1: Let tpγ=max(tpγ_initial,t ₀ );

Step 6.2: Investigate the latest executable time of the task performed by only the γth radar, send the tasks less than tpγ to the deletion queue, and delete these tasks from the request queue;

Step 6.3: Investigate the earliest executable time of the remaining tasks performed by the γ-th radar, and select the tasks whose task is not greater than tpγ, assuming that there are N(γ+1) in total;

If N(γ+1)=0, update parameters: tpγ=tpγ+Δt, go to step 6.5;

If N(γ+1)≠0, calculate the comprehensive priority of these tasks according to (2), select the task with the highest comprehensive priority, and denote it as T _k :

ps _i ={η·Npi +[N(γ+1)+2−η]·Nd _{i }} /[N(γ+1)+1] (2)

where i=1,2,...N(γ+1); Np _i and Nd _i are respectively the priority p of N(γ+1) request tasks from low to high and the deadline d from large to large according to the working mode. Small sorting, the position of task _Ti in these two sequences; η=[N(γ+1)+1]/2;

将T_k送入执行队列，在第γ部雷达天线前端于tpγ时刻调度执行任务T_k，并将其从任务请求队列中删除，更新参数：tpγ＝tpγ+τ_Bk。若

当tpγ+τ_Bk≤time_tpγ(1)时，在第γ部雷达天线前端于tpγ时刻调度执行任务T_k，并将其从任务请求队列中删除，更新参数：tpγ＝tpγ+τ_Bk；当tpγ+τ_Bk＞time_tpγ(1)时，更新参数：tpγ＝time_tpγ(2)，并将time_tpγ(1)和time_tpγ(2)从time_tpγ中删除；Step 6.4: If

Send T _k to the execution queue, schedule and execute task T _k at the time tpγ at the front end of the γ-th radar antenna, and delete it from the task request queue, and update the parameter: tpγ=tpγ+τ _Bk . like

When tpγ+τ _Bk ≤time_tpγ(1), the task Tk is scheduled to be executed at the front end of the γ-th radar antenna at time _tpγ , and it is deleted from the task request queue, and the parameters are updated: tpγ=tpγ+τ _Bk ; when tpγ When +τ _Bk >time_tpγ(1), update the parameter: tpγ=time_tpγ(2), and delete time_tpγ(1) and time_tpγ(2) from time_tpγ;

Step 6.5: If tpγ>t ₀ +SI, go to step 7, otherwise return to step 6.2;

Step 7: Let γ=γ+1, if γ≤M, go back to step 5, otherwise go to step 8;

Step 8: Analyze the remaining tasks in the request queue, delete tasks whose latest executable time is less than t ₀ +SI, and delay tasks whose latest executable time is greater than or equal to t ₀ +SI to the next scheduling interval for analysis;

Step 9: The analysis of the current scheduling interval is completed, and the respective time stamps tpγ of M radars are obtained, where γ=1, 2, .

It can be seen from the above technical solutions that the beam dwell scheduling of single-station radars is the basis for the coordinated distributed system beam dwell scheduling, because the flexibility of a single radar resource is the necessary basis for multi-station coordination; Station coordinated beam residency scheduling is the key link for the distributed system to exert its advantages; with distributed coordinated tracking as the main line, adaptive scheduling subsystems and headquarters tasks are the core of effectively exerting the comprehensive performance of the distributed system; high-priority headquarters The combination of task reservation mechanism and low-priority subsystem task adaptive execution mechanism is an effective way to enrich the flexibility of collaborative distributed system work.

In the present invention, the influence of task working mode priority and deadline on task priority is comprehensively considered, and the HPEDF method is used to calculate the comprehensive priority. The optimization model of beam dwell scheduling is solved by a heuristic method based on time pointer analysis, and a beam dwell scheduling method for cooperative distributed systems is formed. Therefore, the principle of the above method is mainly divided into two aspects, namely, the establishment of a comprehensive priority algorithm and the establishment of an optimization model for beam dwell scheduling for a collaborative distributed system. Among them, the comprehensive priority algorithm adopts the HPEDF method, and the specific principles of establishing the optimization model for the beam dwell scheduling of the cooperative distributed system are as follows:

Assuming that there are N task requests at the current moment, denoted as T={T ₁ , T ₂ , . . . , T _N }, the executable moment ranges of these tasks all include the current moment. Sort these N task requests twice according to the priority of the working mode from low to high and the deadline from large to small, and record the serial numbers of the task T _i in the two sequences as Npi and Nd _i , _i =1,2,...,N, define the comprehensive priority ps _i of each task request as a function of Npi and Nd _{i ,} in the form of a simple linear function:

ps _i =[η·Npi +(N+2−η)·Nd _i ]/ ₍ N+1) (3)

The larger the value of ps _i , the higher the comprehensive priority of the task. η is a controllable parameter, η∈[1, N1+1], when η=1, it is MEDF (Modified Earliest Deadline First, modified deadline priority), when η=N1+1, it is MHPF (Modified Earliest Deadline First) Highest Priority First, correction working mode priority), when η=(N1+1)/2, it is HPEDF (Highest Priority Earliest Deadline First, comprehensive priority).

During the scheduling process, it should be ensured that all executed tasks are executed within their executable time window, and there will be no time conflicts during the execution of each task. The following scheduling optimization model is constructed for this:

表示第j部雷达执行的任务个数，j＝1,2,…,M,M为系统中子雷达的总部数；st_i1表示第i1个任务的实际执行时刻，i1＝1,2,…,N₁；

表示第j部雷达调度的第i1'个任务的实际执行时刻，

模型中目标函数为最大化系统所有执行任务的综合优先级之和。第一个约束条件说明被执行的任务必须在其可执行时间窗内执行，反映了任务执行的及时性要求；第二个约束条件说明针对单部子雷达，被调度任务的驻留时间在执行过程中是非抢占的，是任务成功调度的必要条件；最后两个约束条件分别说明了任务被延迟和删除的条件。Among them, N ₁ , N ₂ and N ₃ respectively represent the number of tasks that the system executes, delays and deletes within the scheduling interval [t ₀ , t ₀ +SI], t ₀ is the starting time of the scheduling interval, and t ₀ +SI is the end time of the scheduling interval;

Represents the number of tasks performed by the jth radar, j=1,2,…,M,M is the number of headquarters of the system neutron radar; st _i1 represents the actual execution time of the i1th task, i1=1,2,… ,N ₁ ;

represents the actual execution time of the i1'th task scheduled by the jth radar,

The objective function in the model is to maximize the sum of the comprehensive priorities of all execution tasks of the system. The first constraint states that the task to be executed must be executed within its executable time window, reflecting the timeliness requirements of task execution; the second constraint states that for a single sub-radar, the residence time of the scheduled task is within the execution time The process is non-preemptive, which is a necessary condition for the successful scheduling of tasks; the last two constraints describe the conditions for tasks to be delayed and deleted, respectively.

The above optimization problem is a nonlinear optimization problem, so a heuristic algorithm is used to solve it, and according to the characteristics of multi-station cooperative implementation of radar task scheduling in a collaborative distributed system, the scheduling time is given priority to the headquarters task, and then the rest of the idle time is non-linear. The headquarters task allocation scheduling time, thus ensuring the priority of the headquarters task scheduling.

In the radar task scheduling process of the collaborative distributed system, the tasks initiated by the headquarters that need to be completed by multiple radars in the system have the highest priority. If the headquarters task appears, all radars are in an idle state, and they will jointly complete the radar task; if not all radars participating in the task are in an idle state at this time, the "busy" radar must stop the non-active radar it is performing. Headquarters tasks, participate in the execution of current headquarters tasks. In this regard, the present invention designs a multi-station coordinated beam dwell scheduling method, so that the multiple sub-radars in the coordinated distributed system can not only independently perform target detection and tracking tasks, but also cooperate when needed. Complete a certain target detection or task tracking together.

To sum up, the present invention solves the problem of beam dwell scheduling in a cooperative distributed system in which sub-radars can work independently and cooperate when needed.

Description of drawings

1 is a flowchart of an embodiment algorithm;

Fig. 2 is the algorithm flow chart of module 1 in the embodiment;

Fig. 3 is the algorithm flow chart of module 2 in the embodiment;

Fig. 4 is the algorithm flow chart of module 3 in the embodiment;

Fig. 5 is the simulation result diagram that embodiment headquarters task tracking frequency is 2Hz;

FIG. 6 is a simulation result diagram of the headquarters task tracking frequency of 5 Hz in the embodiment.

Detailed ways

The present invention will be further described in detail below with reference to the embodiments and the accompanying drawings.

Using the scheduling method proposed by the present invention, the performance evaluations of task drop ratio (Task Drop ratio, TDR), realization value ratio (Hit Value ratio, HVR) and time utilization ratio (Time Utilization Ratio, TUR) are respectively performed. The definitions of TDR, HVR, and TUR are respectively as (6)-(8):

TDR=N _lose /N _total (5)

Among them, N _lose represents the number of lost tasks, and N _total represents the total number of tasks;

This indicator represents the ratio of the sum of the value of all successfully scheduled tasks to the sum of the value of all requested tasks, and reflects the proportion of high-priority tasks that are successfully scheduled, where N _suc represents the total number of successfully scheduled tasks;

This indicator represents the ratio of the sum of the residence time of all successfully scheduled tasks to the total simulation time, and reflects the system's utilization performance of time resources, where T _total represents the total simulation time.

Taking two sub-radars as an example, id={1,2,3}, where id=1 indicates that the target is only detected or tracked by the first radar (hereinafter referred to as radar 1), and id=2 indicates that the target is only detected or tracked by the second radar The external radar (hereinafter referred to as radar 2) detects or tracks, and id=3 indicates that the target is detected or tracked by radar 1 and radar 2 jointly, that is, the headquarters task has id=3, and the non-headquarters task has id=1 or id=2.

In this embodiment, three tasks of headquarters tracking, precise tracking and general tracking are considered. The headquarters tracking task is performed by Radar 1 and Radar 2 at the same time, and the other two types of tasks are performed by Radar 1 or Radar 2 alone. The ratio of the number of targets tracked by the headquarters, the targets individually tracked by radar 1 and the targets tracked by radar 2 alone is 1:1:1, and the ratio of the number of targets between ordinary tracking and precision tracking is 4:1. From the start of simulation to the first sampling period of tracking, a time is randomly selected as the capture time of the target, that is, the tracking task is set to arrive randomly. The total simulation duration T _total = 12s, the scheduling interval duration SI = 50ms, the pointer sliding step size Δt = 50μs, the task's transmit pulse and receive pulse width are both 1ms, and the number of radar tracking targets is 10-500 batches. Monte Carlo simulations were performed 100 times for the headquarters mission tracking frequency of 2Hz and 5Hz respectively. The resident task parameters of the headquarters task tracking frequency of 2Hz and 5Hz are shown in Table 1 and Table 2, respectively.

Table 1. Parameter table of the radar station mission with the mission tracking frequency of 2Hz in the headquarters

Table 2 Parameter list of the radar stationed mission with the mission tracking frequency of 5Hz at the headquarters

task type work style priority Period (ms) Dwell time (ms) Time window (ms) Headquarters tracking 5 200 5 10 Precision tracking 4 250 4 15 normal tracking 3 500 5 20

Figure 5 is the simulation result of the headquarters mission tracking frequency of 2Hz, in which Figure 5-(a) is the system mission loss rate curve, Figure 5-(b) is the radar 1 mission loss rate curve, and Figure 5-(c) is the radar. 2 The task loss rate curve, Figure 5-(d) is the realized value rate curve, Figure 5-(e) is the time utilization curve, and Figure 5-(f) is the system's performance within 0-200ms when the number of targets is 100. Task scheduling sequence diagram.

When the tracking frequency of the headquarters mission is 2 Hz, Figure 5-(a) is the system loss rate curve, which counts the respective loss rates of the three types of tasks of the headquarters mission, precision tracking and ordinary tracking in the system composed of radar 1 and radar 2. . It can be seen from the figure that when the number of targets reaches 50, ordinary tracking tasks begin to be lost; when the number of targets reaches 60, precision tracking tasks begin to be lost; when the number of targets reaches 150, headquarters tasks begin to be lost significantly. The number of targets is between 10 and 500 batches, and the system's headquarters task loss rate is significantly lower than that of the other two types of tracking tasks. When the number of targets is greater than 400, the loss rate of precision tracking and ordinary tracking tasks is almost 1, which indicates that almost all tasks scheduled by the system are headquarters tasks. This is because the proposed algorithm prioritizes headquarters tasks. When the number of targets increases and the radar is overloaded, almost all the scheduled tasks are headquarters tasks, and almost all time resources are occupied by headquarters tasks, resulting in almost all non-headquarters tasks being lost.

Figure 5-(b) and Figure 5-(c) are the loss rate curves of Radar 1 and Radar 2, respectively, and the curve trends are similar to those corresponding to the system in Figure 5-(a). Figure 5-(d) is the realized value rate curve, which respectively counts the realized value rate of the headquarters task and the single radar tracking task. As can be seen from Figure 5-(d), when the number of targets is less than 50, the realized value rate of radar 1 and radar 2 are both 1, indicating that the applied tasks are all scheduled for execution; when the number of targets is greater than 50, the realized value rate is 1 The value rate curve begins to decline. The scheduling situation of Radar 1 and Radar 2 is similar, and the realized value rate curve basically overlaps. However, the realized value rate curve of the headquarters mission only decreases significantly when the number of targets is greater than 150, and when the number of targets reaches 500, the realized value rate is about 0.8, which is still higher than that of a single radar. This is due to the improvement of the algorithm. The priority of the headquarters task is reduced, the loss rate of the headquarters task is reduced, and the realized value rate is increased.

Figure 5-(e) is the time utilization curve, which counts the time utilization of radar 1 and radar 2 respectively. From the analysis of Figure 5-(a) to Figure 5-(c), it can be seen that the number of headquarters tasks scheduled by Radar 1 and Radar 2 is the same, and the number of scheduled non-headquarters tracking tasks and the residence time are not much different, so The time utilization curves of the two radars basically overlap. When the number of targets is 10-100, the time utilization curve basically increases linearly; when the number of targets is between 100-150, the growth rate decreases; when the number of targets is between 150-400, the growth rate is slower. Combined with Figure 5 From the analysis of -(a) to Figure 5-(c), it can be seen that when the number of targets is greater than 150, the headquarters tasks are obviously lost, indicating that the system is in an overloaded state. The number of added headquarters tasks is slightly more than the number of lost non-headquarters tasks, so the time utilization rate increases slowly; when the number of targets is greater than 400, the time utilization rate basically reaches the maximum value of 1. Combined with Figure 5-(a) to Figure 5 -(c) Analysis can be obtained, this is because when the number of targets is greater than 400, the tasks scheduled and executed by radar 1 and radar 2 are almost all headquarters tasks, and the time axis is occupied, so the time utilization rate is 1. Figure 5-(f) shows the task scheduling sequence diagram when the number of targets is 100 and the simulation time is within 0-200ms, where Ti represents the task sequence number.

Figure 6 is the simulation result of the headquarters mission tracking frequency of 5Hz, in which Figure 6-(a) is the system mission loss rate curve, Figure 6-(b) is the radar 1 mission loss rate curve, and Figure 6-(c) is the radar. 2 The task loss rate curve, Figure 6-(d) is the realized value rate curve, Figure 6-(e) is the time utilization curve, and Figure 6-(f) is the system's performance within 0-200ms when the number of targets is 100. Task scheduling sequence diagram.

In the case where the headquarters task tracking frequency is 5Hz, Figure 6-(a) is the task loss rate curve of the system when the headquarters task tracking frequency is 5Hz. It can be seen that when the number of targets reaches 40, ordinary and precision tracking tasks begin to be lost. When the number of targets is between 50 and 130, the loss rate curve of these two types of tasks increases almost linearly. When the number of targets is between 130 and 170 When the number of targets is greater than 170, the loss rate of the two types of tasks tends to stabilize at the maximum value of 1; when the number of targets reaches 70, the headquarters tasks begin to be lost. Compared with the simulation results when the headquarters task tracking frequency is 2Hz, when the headquarters task tracking frequency is 5Hz, the number of targets lost at the beginning of the three types of tasks are all less than 2Hz, especially the headquarters task, the loss rate is when the number of targets is greater than 70. higher than the loss rate of 2 Hz under the same number of targets. This is due to the increase in the tracking frequency of headquarters tasks, which increases the number of headquarters tasks, which makes the system and individual radars in an overloaded state faster, that is, when the number of targets is small, the mission begins to be lost. However, when the number of targets is between 10 and 500 batches, the system's headquarters task loss rate is still not higher than that of the other two types of tracking tasks, which indicates that the proposed algorithm still has priority scheduling when the headquarters task tracking frequency is 5Hz. Effectiveness of headquarters missions.

Figure 6-(b) and Figure 6-(c) are the loss rate curves of Radar 1 and Radar 2, respectively, and the trend of the curve is similar to that of the corresponding task loss rate curve of the system. Figure 6-(d) is the realized value rate curve when the headquarters task tracking frequency is 5 Hz, and the realized value rate of the headquarters task and the single radar tracking task are calculated respectively. It can be seen from this that when the number of targets is less than 40, the mission realization value rate of the headquarters and individual radars are both 1, indicating that the applied tasks are all scheduled for execution. When the number of targets is greater than 40, the realized value rate curve of a single radar begins to decline. The realized value rate curve of the headquarters mission began to decline when the number of targets reached 70, and it was still higher than the realized value rate of a single radar when the number of targets reached 500. This further shows that when the headquarters task tracking frequency is 5Hz, the algorithm can still prioritize the headquarters tasks, so that the realized value rate is higher than that of the other two types of tracking tasks.

Figure 6-(e) is the time utilization curve of radar 1 and radar 2 when the headquarters mission tracking frequency is 5 Hz. It can be seen from this that the time utilization curves of radar 1 and radar 2 basically overlap. When the target number is 10-70, the time utilization curve basically increases linearly; when the target number is 70-150, the growth rate gradually decreases; when the target number is greater than 150, the time utilization rate reaches and stabilizes at the maximum value of 1 , indicating that when the number of targets is greater than 150, the system schedules almost all headquarters tasks, and the time axis is completely occupied. Figure 6-(f) shows the task scheduling sequence diagram of the simulation time within 0-200ms when the number of targets is 100 in the case of 5Hz, where Ti represents the task sequence number.

Comparing Figures 5-(f) and 6-(f), it can be seen that when the headquarters task tracking frequency is 5 Hz, the density of executed tasks within the same scheduling interval is greater than that in the 2 Hz case, and most of the tasks are scheduled for the headquarters. This is because under the same number of targets, with the increase in the tracking frequency of headquarters tasks, the system will be in an overloaded state earlier, and the algorithm is designed to prioritize headquarters tasks, so the density of scheduled tasks within the scheduling interval increased, and more of them were headquarters missions. Then the loss rate of non-headquarters tasks will increase accordingly. Combining Figure 5-(a) and Figure 6-(a), it can be seen that when the number of targets is 100, the loss rate of non-headquarters tasks in the case of 5Hz is higher than 2Hz case.

To sum up, the beam dwell scheduling method for the collaborative distributed system proposed by the present invention firstly arranges the execution time for the headquarters task, so that the sub-radars simultaneously schedule the headquarters task preferentially, and then arranges the non-headquarters tasks from the remaining time segments. At the execution time, the sub-radars can schedule their respective non-headquarters tasks. The simulation proves that the proposed method can ensure the priority scheduling of headquarters tasks and make the loss rate lower than that of non-headquarters tasks when the headquarters task frequencies are 2Hz and 5Hz respectively.

Claims (1)

1. A beam residence scheduling method for a cooperative distributed system comprises the following specific steps:

step 1: the system has M sub-radars in total, and the sub-radars are arranged in the scheduling interval t₀,t₀+SI]There are N requesting tasks, denoted T ═ T₁,T₂,…,T_NThe earliest execution time of these tasks is less than t₀+ SI, latest execution time t or more₀(ii) a Wherein, t₀For the starting time of the current scheduling interval, SI is the duration of one scheduling interval; the ith task model is known as T_i＝{p_i,rt_i,d_i,l_i,τ_BiId, the working mode priority p of the task can be directly obtained from the task_iDesired execution time rt_iEnd period d_iTime window l_iDuration of residence time τ_BiId is a target identifier, and id ═ {1,2, …, M +1}, where id ═ γ denotes that the target is detected or tracked by the γ th radar alone, γ ═ 1,2, …, M, and id ═ M +1 denotes that the target is detected or tracked by the M radars together, that is, the headquarter task has id ═ M +1, and the non-headquarter task has id ═ γ;

initialization operation: respectively initializing time marks tp gamma of M radars, wherein tp gamma is more than or equal to t₀And initializing a storage device between the scheduling time and the task completion time by setting tp gamma _ initial to tp gamma:

wherein γ is 1,2, …, M;

step 2: whether a headquarter task exists in the request queue is inspected, if yes, the step 3 is executed, and if not, the step 4 is executed;

and step 3: and (3) scheduling and analyzing headquarter tasks:

step 3.1: let tp γ be max (tp1, tp2, …, tpM), γ be 1,2, …, M;

step 3.2: considering the latest executable time of the headquarter task, sending the tasks smaller than tp1 into a deletion queue, and deleting the tasks from the request queue;

step 3.3: considering the earliest executable time of the tasks of the rest headquarters, selecting the tasks which are not more than tp1, and assuming that N1 tasks are total;

if N1 is equal to 0, update the parameters: tp γ ═ tp γ +/Δ t, γ ═ 1,2, …, M, where Δ t is the minimum sliding step of the time stamp, step 3.5;

if N1 ≠ 0, the comprehensive priority of the tasks is calculated according to the formula (1), and the task with the highest comprehensive priority is selected from the comprehensive priorities and is recorded as T_j：

ps_i＝[η·Np_i+(N1+2-η)·Nd_i]/(N1+1) (1)

Wherein i is 1,2, … N1; the N1 request tasks are respectively sorted according to the working mode priority p from low to high and the deadline d from far to near, Np_iAnd Nd_iAre respectively task T_iThe position in the two sequences; η ═ N1+ 1)/2; ps is_iFor task T_iThe integrated priority of (2);

step 3.4: will T_jAnd sending the task into an execution queue, scheduling and executing the task at the time tp gamma of the front end of the antenna of the gamma part radar, deleting the task from a task request queue, and storing the scheduling time into a time storage device of each radar: time _ tp γ ═ time _ tp γ, tp γ]And then updating the time identification of each radar: tp γ ═ tp γ + τ_BjAnd storing the task completion time into a time storage device of each radar: time _ tp γ ═ time _ tp γ, tp γ]Wherein γ is 1,2, …, M;

step 3.5: if tp1>t₀+ SI, go to step 4, otherwise return to step 3.2;

and 4, step 4: let γ equal to 1;

and 5: whether a task which is only executed by a gamma radar exists in the request queue is inspected, if yes, step 6 is executed, and if not, step 7 is executed;

step 6: scheduling analysis is performed on tasks performed by the gamma radar alone:

step 6.1: let tp γ be max(tpγ_initial,t₀)；

Step 6.2: examining the latest executable time of the tasks executed only by the gamma part radar, sending the tasks smaller than tp gamma into a deletion queue, and deleting the tasks from the request queue;

step 6.3: examining the earliest executable time of the rest tasks executed only by the Gamma radar, and selecting the tasks which are not more than tp Gamma, and assuming that N (Gamma +1) tasks are total;

if N (γ +1) is equal to 0, update the parameter: tp gamma is tp gamma plus delta t, and step 6.5 is carried out;

if N (gamma +1) ≠ 0, the comprehensive priorities of the tasks are calculated according to the (2), and the task with the highest comprehensive priority is selected from the comprehensive priorities and is recorded as T_k：

ps_i＝{η·Np_i+[N(γ+1)+2-η]·Nd_i}/[N(γ+1)+1] (2)

Wherein i ═ 1,2, … N (γ + 1); np_iAnd Nd_iRespectively sorting N (gamma +1) request tasks according to the working mode priority p from low to high and the deadline d from large to small, and respectively sorting the tasks T_iThe position in the two sequences; η ═ N (γ +1) +1]/2；

Step 6.4: if it is

Will T_kSending the data into an execution queue, and scheduling and executing a task T at the front end of the gamma radar antenna at the time of tp gamma_kAnd deleting the task request from the task request queue, and updating parameters: tp γ ═ tp γ + τ_Bk；

If it is

When tp gamma + tau_BkWhen the time _ tp gamma (1) is not more than time, scheduling and executing the task T at the time tp gamma of the front end of the gamma radar antenna_kAnd deleting the task request from the task request queue, and updating parameters: tp γ ═ tp γ + τ_Bk(ii) a When tp gamma + tau_Bk>time _ tp γ (1), update parameter: tp γ is time _ tp γ (2), and time _ tp γ (1) and time _ tp γ (2) are deleted from time _ tp γ;

step 6.5:if tp gamma>t₀+ SI, go to step 7, otherwise return to step 6.2;

and 7: if gamma is equal to or less than M, returning to the step 5, otherwise, turning to the step 8;

and 8: analyzing the residual tasks in the request queue and enabling the latest executable time to be less than t₀+ SI task deletion, with the latest executable time greater than or equal to t₀The task of the + SI is delayed to the next scheduling interval for analysis;

and step 9: when the current scheduling interval analysis is finished, the time identifier tp gamma of each of the M radars is obtained, wherein gamma is 1,2, …, M, and the deletion queue, the delay queue, the execution queue and the actual execution time.

Images