Cooperative Multitask Planning Strategies for Integrated RF Systems Aboard UAVs

Abstract

Limited by the load capacity of UAVs, it is difficult for an integrated radio frequency (RF) system aboard a single platform to have both wide-area and comprehensive battlefield sensing capabilities. One possible approach to solve this dilemma is to use multiple UAVs to perceive the scene cooperatively and simultaneously. To this end, this paper mainly discusses the cooperative task planning strategies facing cooperative UAVs with integrated RF systems when performing several tasks simultaneously. First, considering the complexity of the planning problem, the physical model for UAV formation cooperation is discussed. Then, based on the irregular and ad hoc characteristics of cooperative UAV networks, the essential compositions for UAVs cooperation are formulated that includes input information and planning constraints as well as evaluation indicators. Furthermore, to solve the given task planning problem, four new planning strategies are targeted designed for different planning purposes. Finally, a simulated cooperative UAV multitask planning scenario including cooperative detection, cooperative localization, and jamming is designed. Simulation results verify the effectiveness of these strategies as well as their advantages, disadvantages, and the multiscenario adaptability of each strategy.

1. Introduction

UAVs will play an increasingly key role in battlefield situations in future warfare due to their merits of a long range of operation and freedom from the physiological limits of human pilots. Various kinds of UAVs have been used in battlefields to execute detection, communication, reconnaissance, and strike tasks. For RF missions, UAVs are usually equipped with multiple RF systems, such as radar, data links, passive reconnaissance devices, and jammers [1,2,3]. However, UAVs capable of multitasking have their limits. Objective limitations, such as the light platform load and relatively weak radiated power, restrict further improvement of UAV perception capabilities, especially when the equipped RF systems are independent. To overcome these shortcomings, some attempts have been made to remove such restrictions, i.e., developing an integrated RF system with a high degree of aperture sharing or using multiple cooperative UAVs.

An integrated RF system can improve the multitask capabilities of a single UAV platform, while cooperation offers additional degrees of freedom. These features create new possibilities for further improving the perception performance of UAVs. With such flexibility and designability, it is natural to consider how the right groups can be chosen dynamically from a given UAV formation to execute the scheduled tasks. Thus, the problems of multitask planning for cooperative UAVs arise.

To our knowledge, few studies have specifically considered this issue. The related studies are summarized in

Table 1. Development of related studies.

Related Works	Main Contributions	Application
Baptiste [4]	Resource scheduling algorithm for a single platform	Single platform
Barbato A [5]	Resource scheduling algorithm for multitask and multifunction radar	Single platform
Lee C G [6]	Real-time resource scheduling algorithm for radar system	Single platform
Chavali P [7]	Task planning algorithm for multitarget tracking	Multi-platforms and multitargets
Mir H S [8]	Localization algorithm for multi-sensors	multi-sensors for cooperation
Zhang Y Z [9]	Task decision-making algorithm for cooperative reconnaissance	Multi-sensors for cooperation
Zhang Z [10]	Resource scheduling method for netted radar	Multi-platfoms and multi-sensors

. In contrast, more works tend to discuss the resource scheduling problems faced by single-platform radar systems and task planning problems in radar networks [4,5,6,7]. In addition, cooperative reconnaissance and localization for multi-sensors are researched widely [8,9,10]. However, research on resource scheduling of integrated RF system and cooperation for multi-platform is rare. In conjunction with existing research on task planning methods [11,12] and optimization algorithms [13,14,15,16], this paper presents a comprehensive study on this pending multitask planning problem.

Inspired by the studies mentioned above, we have conducted in-depth research on the cooperative task planning for integrated RF systems aboard UAVs. In previous work, we had given the accurate cooperative physical and mathematical model [17]; in this paper, we focus on the effectiveness of different task planning strategies. The main contributions of this paper include:

(a)

Four novel task-planning strategies are designed for different planning purposes based on the constraint conditions and evaluation indicators.

(b)

The results for different task planning strategies are presented by simulation.

(c)

The effectiveness of these strategies, as well as their advantages, disadvantages, and scenarios for which they are most suitable, are analyzed by comparison and analysis.

The remainder of this paper is organized as follows. Section 2 introduces the application scenario and the physical model of the UAV formation. Section 3 presents the essential compositions for UAVs cooperation task planning. Specific task planning strategies and flows are shown in Section 4 and Section 5. Simulated planning scenario results and discussions are detailed in Section 6. Conclusions are given in Section 7.

2. Physical Model for UAV Cooperation

In modern warfare, a UAV formation may contain various kinds of UAVs. Each UAV is assumed to be equipped with an integrated RF system, and each UAV can perform one or more tasks, such as reconnaissance, jamming, radar detection, and communication. For example, Figure 1 shows a typical graphical integrated RF system description capabilities of UAV formation. All UAVs are multifunctional, and all UAVs have communication capabilities.

In brief, the ultimate goal of task planning is to selectively divide UAVs into several groups, and UAVs in the same group execute the same task. For ease of understanding, a detailed cooperative multitask planning example at a certain moment is shown in Figure 2. The number of UAVs in group n is determined by the specific requirements of the corresponding task to be scheduled. To be more specific, a group requires at least two UAVs to execute the cooperative localization task, namely, n must not be less than 2.

3. Essential Compositions for UAVs Cooperation Task Planning

Collaboration among multifunctional UAV formations is conditional. The mathematical analysis of constraint conditions should be based on the physical model of UAV formation to execute multitasks. The mathematical models should be constructed to cover various aspects of physical processes, including the function of UAVs, collaborative conditions of integrated RF systems, and the location and time constraints of UAVs during collaboration. Consider a detection scenario: a UAV formation with $M$ UAVs, and $J$ tasks to be planned. If the mth UAV is allocated to execute the jth task, and is denoted by $TASK=\left\{T_{j,m}\right\}$. Next, the mathematical descriptions of the essential components for task planning are discussed in detail.

3.1. Input Information

For task planning, specific input information is needed to support decision-making. The input information must match the physical scene and model. In this task planning problem for UAV formation, the input information mainly includes the geographic information, task information, and UAV information. The geographic information mainly includes the related information of the areas of interest. The task information mainly includes the coordinates of target, task type, start time, and so on. The UAV information mainly refers to the related parameters of UAVs, for example, the capabilities of UAVs and coordinates of UAVs.

3.2. Constraint Conditions

In practice, many constraints exist in task planning for UAV formation, such as the function and position of UAVs, and the capabilities and time constraints of integrated RF systems. The relevant constraints can be roughly divided into several categories.

(a)

Function constraint

In terms of UAV functionality, only UAVs with a certain function can participate in the collaboration of this type of task. Specifically, only if a UAV has the corresponding RF capability to execute the task to be planned will this UAV node be included in the scheduling queue. Let $Nod{e}_k^{T_j}$ denote the set of UAVs executing task $T_j$ at time $t_k$; then, this constraint can be formulated as $Nod{e}_k^{T_j}\in n_{T_j}$, where $n_{T_j}$ is the available UAV set that can execute the given task $T_j$.

(b)

Position constraint

At a certain collaborative moment, all UAVs collaboratively executing a task must point in the same direction and the beam cannot be obstructed in the transmission direction.

For some of the cooperative tasks, such as passive cooperative positioning, the selected UAV nodes must meet certain geometric constraints. Specifically,

$$d_r\tan \theta >\frac{\lambda }{D}R$$

(1)

where $d_r$ is the baseline length, $\lambda$ is the carrier wavelength, $D$ is the target size, and $R$ is the range from the target to the UAV. A detailed geometric representation is shown in Figure 3.

(c)

Integrated RF resource constraint

As the integrated RF system on the UAVs is integrated and may share a common array, each UAV can only execute, at most, one task at any time, and mutual electromagnetic interference among UAVs should be avoided.

(d)

Time resource constraints

For a planned task $T_{j,m}$, its start time must be greater than the sum of the end time of the previous task and the beam switching time $\eta$.

3.3. Evaluation Indicators

Following the particularities of this cooperative task planning problem, four indicators are introduced to evaluate the candidate strategies [17].

(a) Task planning revenue: the ratio of the number of successfully planned tasks to the total number of tasks to be planned. Considering the priorities of tasks, a weighted revenue is introduced:

$$W={\displaystyle \sum _{j=1}^J{\omega }_j{P}_j}/{\displaystyle \sum _{j=1}^J{P}_j}$$

(2)

where ${\omega }_j$ is a Boolean parameter that represents whether task $j$ is planned successfully and $P_j$ is the priority of the jth task. The more tasks with higher priorities that are successfully planned, the greater $W$ will be. The maximum value of $W$ is 1.

(b) Time utilization rate: the ratio of the sum of the dwell time and the beam switching time for all successfully planned tasks to the total scheduling interval $t^{\mathrm{SI}}$. This metric can be defined as

$$t_{\mathrm{util}}={\displaystyle \sum _{m=1}^M\left({\displaystyle \sum _{j=1}^{J_{m_{\mathrm{suc}}}}(t_{j,m}^{\mathrm{Dwell}}+\eta )}\right)/M{t}^{\mathrm{SI}}}$$

(3)

where $t_{j,m}^{\mathrm{Dwell}}$ is the dwell time of the jth task executed by the mth UAV, $J_{m_{\mathrm{suc}}}$ is the number of tasks that are successfully executed by the mth UAV, $\eta$ is the beam switching time, and $t^{\mathrm{SI}}$ is the scheduling interval. For a given interval, the higher the time utilization is, the greater the effectiveness of the task plan. The highest $t_{\mathrm{util}}$ is 1.

(c) Time-shifting rate: the ratio of the difference between the actual execution time and the expected start time of the task to the time window of the task. Due to various conflicts, such as limited UAV nodes and time resources, the actual execution time of task may lead or lag behind its expected start time. The maximum allowable range of this time shift is the “time window” of a task. To evaluate the time-shifting rates of all tasks, the average time-shifting rate (ATSR) is defined as

$$\delta t_{\mathrm{shift}}=\frac{1}{J_{\mathrm{suc}}}{\displaystyle \sum _{m=1}^M{\displaystyle \sum _{j=1}^{J_{m_{suc}}}\frac{\left|t_{j,m}^{\mathrm{exe}}-t_j^{\mathrm{start}}\right|}{t_j^{\mathrm{window}}}}}$$

(4)

where $t_{j,m}^{\mathrm{exe}}$ is the actual execution time of the jth task executed by the mth UAV, $t_j^{\mathrm{start}}$ is the expected start time, $t_j^{\mathrm{window}}$ is the time window, and $J_{\mathrm{suc}}$ is the number of tasks that are successfully executed. The lower the ATSR is, the better the planning effectiveness will be.

(d) Task execution equality rate: the difference in working time between the “busiest” UAV and the “idlest” UAV in the formation. To avoid the case in which some UAVs are always kept busy while some UAVs are always idle, the task execution equality rate $u$ is introduced which could improve the task planning success rate.

$$\begin{array}{ll}u=& \max \left|t_{m_1}^{\mathrm{busy}}-t_{m_2}^{\mathrm{busy}}\right|\\ & \forall m_1,m_2\in \left\{1,\cdots ,M\right\}\end{array}$$

(5)

where $t_{m_1}^{\mathrm{busy}}$ is the total working time of the $m_1$th UAV. A smaller $u$ implies that the planned tasks are more balanced.

By combining the above evaluation indicators, the optimal strategy can then be selected by solving the following optimization problem:

$$\begin{array}{l}\max \zeta =\frac{W}{u{e}^{\delta t_{\mathrm{shift}}}}\\ \mathrm{s}.\mathrm{t}.\\ 0<W\le 1;\\ 0<\delta t_{\mathrm{shift}}\le 1\end{array}$$

(6)

4. Task Planning Strategies

Task planning for the collaboration of UAVs with integrated RF systems is essentially a question of how to select UAV nodes and task execution time for different cooperative tasks. Which task is planned first? By what time? Which UAV or UAVs are selected? All of these factors affect the planning result. A good task planning strategy should enable a limited number of UAVs to successfully execute more tasks within a certain scheduling interval. To summarize, the main factors involved in the UAV collaboration problem include the task planning sequence, task execution time, and UAV node selection. Based on the above analysis, four strategies for task planning and resource scheduling for integrated RF systems aboard UAVs are proposed from different perspectives.

4.1. Strategy I: Highest Priority of the Task and Shortest Distance between UAVs and the Target

The specific design is as follows: (1) First, all tasks are sorted according to their priorities; the task with the highest priority will be preferentially planned. (2) When there is more than one task with the highest priority, the task with the earliest expected start time is selected. (3) During the process of UAV assignment, when constraint conditions such as function, position, and time constraints are met, the node or combination of nodes closest to the target is preferred.

4.2. Strategy II: Highest Priority of the Task and Best Task Execution Equality Rate of UAVs

The specific design is as follows: (1) First, all tasks are sorted according to their priorities; the task with the highest priority will be preferentially planned. (2) When there is more than one task with the highest priority, the task with the earliest expected start time is selected. (3) During the process of UAV assignment, the node or combination of nodes that has completed the fewest tasks and has the lowest time utilization rate is preferred.

4.3. Strategy III: Earliest Expected Start Time of the Task and Shortest Distance between UAVs and the Target

The specific design is as follows: (1) First, all tasks are sorted according to their expected start times; the task with the earliest expected start time will be preferentially planned. (2) When there is more than one task with the same expected start time, the task with the highest priority is selected. (3) During the process of UAV assignment, when constraint conditions such as function, position and time constraints are met, the node or combination of nodes closest to the target is preferred.

4.4. Strategy IV: Earliest Expected Start Time of the Task and Best Task Execution Equality Rate of UAVs

The specific design is as follows: (1) First, all tasks are sorted according to their expected start times; the task with the earliest expected start time will be preferentially planned. (2) When there is more than one task with the same expected start time, the task with the highest priority is selected. (3) During the process of UAV assignment, the node or combination of nodes that has completed the fewest tasks and has the lowest time utilization rate is preferred.

5. Task Planning Flow

The detailed task planning flow is shown in Figure 4. Firstly, load all the input information, including that which is UAV related and task related; secondly, pull out all tasks in the task planning interval from the Task Scheduling list; thirdly, select the task to be planned according to the task planning strategy; and then, load the preset number of UAVs and select the candidate UAVs that can perform the task according to the constraint conditions. Repeat the above steps until there is no task, and generate the task planning scheme and resource scheduling process for each UAV. The detailed task planning flow was introduced in a previous work [17].

6. Simulation Results and Discussion

6.1. Simulation Parameters

Consider eight UAVs in a formation.

Table 2. Capabilities and positions of UAVs.

UAV	Rec. Capability	Radar Capability	Jam. Capability	Coordinates/m
				x	y	z
UAV #1	1	0	0	200,420	200,356	5434
UAV #2	1	1	1	200,142	199,674	4939
UAV #3	1	1	0	199,830	199,714	4659
UAV #4	1	1	1	200,220	200,253	5300
UAV #5	1	1	0	199,716	200,351	4730
UAV #6	1	1	1	199,818	199,822	4530
UAV #7	1	0	0	199,664	199,940	5240
UAV #8	1	1	1	199,692	200,297	4827

gives the capabilities and coordinates of UAVs in the formation; “1” means that the UAV has the corresponding capability, while “0” means it does not. For example, UAV #1 has only passive reconnaissance capability, so it can execute only reconnaissance-related tasks. If one or more UAVs are malfunctioning, the capabilities of UAVs in Table 2 would be updated. Through task planning strategies, new task planning results can be generated in real-time. The detailed detection scenario is shown in Figure 5a.

Consider 50 tasks to be planned in a scheduling interval. The lower the number of tasks, the higher the task planning success rate for each strategy, possibly reaching 100% for all strategies. The higher the number of tasks, the more failed tasks for all strategies, which may not be sufficient to illustrate the advantages and disadvantages of each strategy. Therefore, based on the dwell time and time window of each task, the number of tasks in this simulation is set as 50. This parameter can be adjusted according to different application needs. The existing targets are randomly distributed in an area of 400 km × 400 km × 20 km, shown in Figure 5b. The parameters of tasks to be planned are listed in

Table 3. Parameters of tasks to be planned.

S/N	Task Type	Nodes	Dwell Time /ms	Time Window	Coordinates/m			Start Time /ms	Deadline /ms
					x	y	x
1	1	1	8	40	390,914	140,398	8393	70	110
2	2	1	12	40	321,448	40,453	508	10	50
3	1	1	8	40	28,170	142,416	4714	72	112
4	7	2	8	30	265,990	265,183	14,982	65	95
5	4	4	10	40	317,821	59,505	17,900	89	129
6	1	1	8	40	298,014	292,073	14,615	37	77
7	8	1	10	30	200,229	198,593	5750	17	47
8	3	3	6	30	304,018	187,893	15,860	18	48
9	2	1	12	40	226,019	228,558	4557	44	84
10	5	2	15	30	309,574	256,529	10,933	34	64
41	2	1	12	40	106,449	217,285	6306	59	99
42	8	1	10	30	106,386	210,546	5161	33	63
43	8	1	10	30	366,505	358,620	962	28	58
44	8	1	10	30	193,621	294,000	5901	36	66
45	4	4	10	40	399,429	238,880	2277	64	104
46	7	2	8	30	11,064	342,277	14,901	45	75
47	7	2	8	30	355,477	40,526	5273	67	97
48	7	2	8	30	188,358	222,482	16,826	49	79
49	4	4	10	40	62,130	121,062	15,196	39	79
50	3	3	6	30	260,324	374,762	5106	38	68

, including task priority, which corresponds to task type, the number of UAVs needed for task execution, time window, dwell time, expected start time, deadline, and coordinates.

6.2. Simulation Results

Given the scenario described above, the proposed multitask planning scheme was implemented. According to the four different task planning strategies, the simulation results are presented and analyzed as follows. The task planning results of the four strategies in this part are all based on the UAV parameters shown in Table 2 and the task parameters shown in Table 3, which makes it easier to compare and analyze the advantages and disadvantages of each planning strategies.

6.2.1. Results for Strategy I

Strategy I is the task planning strategy based on the priority of tasks and the shortest distances between the UAVs and targets. The corresponding results are shown in Figure 6. Figure 6a shows the execution results for all tasks. There are 50 tasks in total. Under Strategy I, 44 tasks are planned successfully, and 6 tasks fail, namely, Tasks 27, 28, 29, 32, 39, and 40. The main reason for failure is that there are many tasks in this period, and the tasks are too concentrated in time to succeed. Figure 6b shows the 3D results of task planning. The color bar on the right corresponds to Tasks 1–50. The horizontal axis is the execution time, and the longitudinal axis shows UAVs #1–8. The same color at the same time point in this figure represents the same task (the same as Figure 6c). Figure 6c shows the 2D results of task planning. The horizontal axis and longitudinal axis are the same as in Figure 6b. The length of an “*” line represents the dwell time of the corresponding task, and the start time and end time of the “*” line represent the actual execution time and completion time of the task, respectively. According to Figure 6b,c, nine tasks are performed by UAV #1, and eight tasks are performed by UAV #8; the UAV with the fewest tasks is UAV #7. UAVs #2 and #4 perform the most tasks. Most obviously, on the right side of Figure 6b, the green area occupies UAVs #2, #3, #4, and #6. According to the task parameters in Table 3, this area may correspond to Task 5 and Task 36. The task type is cooperative detection, which requires a node to have radar capability. Corresponding to Table 2, UAVs #2, #3, #4, and #6 all have radar capability and meet the distance constraint conditions; this is consistent with the simulation results. The task type, priority, and dwell time of Tasks 5, 36, and 39 are the same, but their expected start times are different. Due to the planning strategy, Tasks 5 and 36 are successfully executed, but Task 39 fails due to insufficient UAVs.

Figure 6d shows the time-shifting rates of the tasks and the ATSR. The blue “*” in the figure represents successfully planned tasks, while the red “*” represents failed tasks. The horizontal axis represents task number (No. of tasks), and the longitudinal axis represents the time-shifting rate of each task. The specific definition is given in (4), that is, the ratio of the difference between the actual execution time and expected start time to the time window of the task. The maximum value is 1. As shown in Figure 6d, a longitudinal axis of 0 indicates that the task was executed according to its original expected time and was not delayed. Except for the 6 failed planning tasks, the time-shifting rates of the other 44 tasks are mostly between 0 and 0.8. The ATSR (pink line) is approximately 28%.

Figure 6e shows the time utilization rate of each UAV. UAV #4 has the highest utilization rate, which is greater than 90%, and UAV #7 has the lowest utilization rate, of approximately 47%. The others are between 60 and 90%. One reason for the lower time utilization rate of UAV #7 is that its only RF function is the reconnaissance ability, as shown in Table 2. Similarly, UAV #1 also has only the reconnaissance capability, but its time utilization ratio is higher than that of UAV #7; this is related to the positions of the UAVs and the randomness of the task list. After 80 ms, there are fewer reconnaissance tasks, so the time utilization rates of nodes with only reconnaissance capability are relatively low. As shown in Figure 6e, the average time utilization is (pink line) approximately 70%.

6.2.2. Results for Strategy II

Strategy II is the task planning strategy based on the priority of the tasks and the task execution equality rate of the UAVs. The corresponding results are shown in Figure 7. According to Figure 7a, there are 5 failed tasks and 45 successful tasks. The failed tasks are Tasks 27, 28, 32, 39, and 40. Unlike under Strategy I, Task 29 is successfully planned. Figure 7b,c shows that the UAV with the most tasks is UAV #2, and the one with the fewest is UAV #7. The time utilization of UAVs #2 and #5, which perform most of the tasks, is relatively high. Different from the results for Strategy I, on the right side of Figure 7b, Task 36 is allocated to UAVs #2, #3, #4, and #8, and Task 5 is assigned to UAVs #2, #3, #4, and #6. The main reason is that Strategy II is based on the task execution equality rate. To balance the task execution equality rate of the UAVs, Tasks 36 and 5 are planned separately. As shown in Figure 7d, the ATSR is approximately 23%, which is better than the rate of 28% achieved under Strategy I. Figure 7e shows the time utilization rates of the UAVs. The highest value is for UAV #2, with a time utilization rate greater than 90%, and the lowest value is for UAV #6; most of the values are between 70% and 90%. Generally, the average time utilization rate is approximately 74%, which is better than that under Strategy I.

6.2.3. Results for Strategy III

Strategy III is the task planning strategy based on the expected start times of tasks and the shortest distances between the UAVs and targets, and the task planning results are shown in Figure 8. According to Figure 8a, there are 9 failed tasks and 41 successful tasks, which is worse than the results of Strategies I and II. Figure 8b,c shows that the UAV with the most tasks is still UAV #2, and the UAV with the fewest tasks is UAV #7. The time utilization rates of UAVs #2, #4, and #5 are relatively high. As shown in Figure 8d, the ATSR is nearly 50%, which is also worse than those under Strategies I and II. Figure 8e shows the time utilization rates of the UAVs. UAV #4 has the highest time utilization, more than 90%, and UAV #7 has the lowest time utilization, less than 25%. Therefore, according to (5), the task execution equality rate under Strategy III is not good.

6.2.4. Results for Strategy IV

Strategy IV is the task planning strategy based on the expected start times of tasks and the task execution equality rate of the UAVs, and the results are shown in Figure 9. According to Figure 9a, there are 43 successful tasks and 7 failed tasks. Table 3 shows that the expected start times of the failed tasks are too close. According to Figure 9b,c, UAV #1 successfully performs 8 tasks, and UAV #2 performs 11 tasks. UAV #7 has the fewest tasks. In Figure 9d, the ATSR exceeds 30%, with four items exceeding 70%. In Figure 9e, the UAV with the highest utilization rate is UAV #8, which has a time utilization of just 85%, and the lowest utilization rate is that of UAV #7, which is lower than 50%.

6.3. Comparison and Analysis

For Strategies I–IV, statistics were calculated for 100 Monte Carlo runs, and the corresponding results are shown in Figure 10, Figure 11 and Figure 12. The main process was to generate a group of random tasks and UAVs and then execute task planning under Strategies I–IV (Str. I–IV). As shown in Figure 10, Strategies II and IV result in higher time utilization rates, indicating that a task planning strategy based on the task execution equality rate can improve the task planning success rate to a certain extent. By comparing Strategies II and IV, it can be found that the time utilization rate of each UAV under Strategy II is approximately equal to that under Strategy IV. Figure 11 shows a comparison of the time-shifting rates under the different strategies and indicates that Strategy II results in the lowest time-shifting rate, while Strategy III leads to the highest. This is because Strategies I and II are mainly based on task priorities. Initially, most tasks are planned based on their expected start times, with a time shift of zero. However, Strategies III and IV are mainly based on the tasks’ expected start times. If one task is delayed, it will cause all later tasks to be delayed, which increases the overall time-shifting rate. Therefore, in terms of the time-shifting rate, the overall trend is that strategies based on priority are better than those based on the expected start times, and strategies based on the task execution equality rate are better than those based on the shortest distances between the UAVs and targets. As shown in Figure 12, Strategies II and IV are superior. According to (6), the higher W is, the lower u is and the lower $t_{\mathrm{util}}$ is, and the higher $\zeta$ is, the better the task planning strategy. Figure 12 shows that, because W is directly related to the task planning (TP) success rate, these two curves show parallel development trends. Because Strategies I and II are based on priority, tasks with higher priority have a higher probability of being successfully planned, so the W values of Strategies I and II are higher than those of Strategies III and IV. According to the previous analysis, Strategies III and IV have higher $t_{\mathrm{util}}$. In terms of u, Strategies II and IV, which consider the task execution equality rate when selecting UAVs, achieve higher task planning success rates and time utilization rates.

In summary, for the problem of multitask planning for a UAV formation, four novel planning strategies are proposed based on evaluation indicators such as the task planning revenue, time utilization rate, time-shifting rate, and task execution equality rate. Different strategies have different advantages and disadvantages and are suitable in different scenarios. Through the simulation results, we can obtain the following conclusions.

(1)

A strategy based on the task priority is better than one based on expected start time. Strategies based on priority have higher task planning revenue and lower time-shifting rate, which can ensure that important tasks could be executed in a prioritized and timely manner.

(2)

A strategy based on the task execution equality rate is superior to a strategy based on the shortest distance between UAVs and the target. In the task planning process, the task execution equality rate is relatively important and should be considered. Doing so can improve the task planning success rate by avoiding failure of task planning due to excessive task allocation to certain UAVs. The better the task execution equality rate is, the higher the task-planning success rate, and the higher the weighted total revenue of task planning.

(3)

If the sum of all dwell times of all tasks is far less than the scheduling interval of UAVs, both strategies based on expected start time and task priority could be selected, and both could ensure a certain task planning success rate. However, a strategy based on task priority is more suitable in scenarios with more tasks to be planned, which could obtain higher task planning revenue and lower time-shifting rate.

(4)

If the number of tasks to be executed is large and dense, the number of failed tasks would be increased. In this scenario, a strategy based on task priority and task execution equality should be selected first. Then, if the timeliness requirements of tasks are low, the time window of tasks should be increased appropriately. The larger the time window of the task is, the greater the probability of successful planning. In addition, increasing the number of UAVs or increasing the number of UAVs with certain functions can also increase the probability of successful planning for this type of task.

7. Conclusions

This paper aims to solve the task planning problem for a UAV formation toward multitasks. Because UAVs in the formation have different capabilities, and multitasks require different numbers of UAVs to execute, this planning problem is a complex and exploratory one. For this given problem, four planning strategies are proposed based on evaluation indicators. The proposed indicators correspond to the physical sense of the problem being solved. The simulation results verify the effectiveness of these strategies as well as their advantages, disadvantages, and scenarios for which they are most suitable. The research in the article can be applied to the collaborative application of several or dozens of UAVs. The comprehensive RF tasks can also be changed as other tasks, and only the task parameters need to be modified according to the new tasks. The research is also applicable to situations where the number of UAVs on the battlefield changes. The planning strategy takes dozens of microseconds to achieve real-time task planning. Since cooperation among UAVs with integrated RF systems is a relatively new concept, further research is still needed to clarify the relevant cooperation characteristics, perfect the cooperation model, and improve the effectiveness of task planning.