Disclosure of Invention
In view of the above, the invention provides a method for optimizing unmanned aerial vehicle task completion time by combining trajectory, power control, device grouping and scheduling of a multiple unmanned aerial vehicle acquisition system based on NOMA, which effectively improves unmanned aerial vehicle acquisition time efficiency, and aims at the problems of grouping scheduling of ground nodes in an unmanned aerial vehicle auxiliary acquisition network and data acquisition requirements of unmanned aerial vehicles.
In order to achieve the purpose, the invention provides the following technical scheme:
a joint optimization method of a multi-unmanned aerial vehicle acquisition system based on NOMA is characterized by comprising the following steps:
s1: constructing a multi-unmanned aerial vehicle-assisted non-orthogonal multiple access emergency acquisition data network model:
(1) Communication network channel model
M unmanned aerial vehicles serve K ground equipment nodes in the area without ground base stations as mobile base stations, and the data volume to be transmitted by each ground node is S
k Definition of u
k Is the location of the kth ground node,
![Figure BDA0003845651990000021](https://patentimages.storage.googleapis.com/10/1c/8a/70e7d59d30ae9a/BDA0003845651990000021.png)
the horizontal projection position of the mth unmanned aerial vehicle in the nth time slot is determined; suppose that the flying height of the unmanned aerial vehicle is constant H and has a fixed starting point and a fixed stopping point. The method comprises the steps that the Time path of an unmanned aerial vehicle executing an acquisition task is discretized into N Time slots, data of ground node equipment are collected in each Time slot by Time Division Multiple Access (TDMA) between M unmanned aerial vehicles, uplink transmission is carried out between each pair of equipment in a NOMA mode, a communication Link (LOS) link between the unmanned aerial vehicle and the ground equipment is assumed to be dominant, a LOS-to-ground channel model is adopted by the channel model, and channel gain only depends on the distance between the unmanned aerial vehicle and the ground equipment. Furthermore, assuming that the doppler effect caused by drone mobility is compensated, the channel gain between drone m and kth SN in the nth slot can be expressed as:
wherein
Defined as the channel gain, β, between drone m and the kth SN in the nth slot
0 Representing the channel gain at a unit distance of 1 meter.
(2) Non-orthogonal multiple access grouping and scheduling model
Defining L as the logarithm of scheduling in each time slot, i.e. scheduling 2L ground nodes, 2L in each time slot under the condition of pairwise pairing<<K; variables of
Is defined to represent the scheduling relationship between drone m and devices i and j in the nth time slot,
indicating that communication is established with device i, j, otherwise not; by definition, there is an expression:
scheduling variable x in each time slot n k,n Comprises the following steps:
according to the SIC receiver rule of NOMA, each common-channel NOMA user is allocated a decoding sequence according to the channel state without loss of generality, and the channel gain of a node i in a time slot n is assumed
Channel gain greater than node j
I.e., the decoding order of node i is higher than j, and it is known that, and nodes i and j can only be scheduled by one drone at most once in one slot, the packet scheduling constraint can be expressed as:
(3) Ground node data uplink transmission model
The method specifically comprises unmanned aerial vehicle flight trajectory constraint, and task resource constraint of ground nodes comprises communication demand constraint, peak power constraint, subtask division constraint, time slot variable constraint and the like of node equipment; scheduling the ith unmanned aerial vehicle in the nth time slot under the unit bandwidth ij The uplink communication rate of user i of the group is
Scheduling the ith unmanned aerial vehicle by the mth unmanned aerial vehicle in the nth time slot ij User j of the group has an uplink communication rate of
Wherein p is k,n K ∈ { i, j } is expressed as the transmit power of the kth node in the nth slot, σ 2 Is additive white gaussian noise; definition of alpha k,n ∈[0,1]A scaling factor, which is expressed as the amount of tasks that the kth node needs to transmit in the nth slot, then has the following communication constraints:
wherein p is min Minimum transmit power, p, in case of scheduling on behalf of a node max Representing the peak power of the node; setting the total time period for completing the tasks of the unmanned aerial vehicle as T and the time length T in the nth time slot n The following constraints should be satisfied:
wherein L is n ={i,j,m\x mn ij =1}; the unmanned aerial vehicle flight path constraint comprises unmanned aerial vehicle steering angle constraint, maximum flight speed constraint and start and stop point constraint, minimum safety distance constraint between the unmanned aerial vehicles, and specifically, the position coordinates, speed and the like of each time slot of the unmanned aerial vehicle meet the following dynamic constraint:
wherein, V
max Representing the maximum instantaneous speed of the drone, cos
min A cosine value representing the maximum steering angle of the drone,
the steering angle, T, of the drone at the nth time slot represented by the ground
n Is the duration of the nth time slot, q
e Is the starting and stopping point of the unmanned aerial vehicle, d
safe Is the minimum safe distance between drones.
S2: the method takes the task execution duration of all the unmanned aerial vehicles to be maximized as an objective function, considers collision constraint, steering angle constraint, flight speed constraint, NOMA group channel gain constraint, peak power constraint of each device, minimum total data collection constraint of each device node, grouping scheduling constraint of each node and the like among the unmanned aerial vehicles, and designs an optimization problem:
s.t.
(13)(14)(15)(16)
wherein (17 a) and (17 b) are constraint conditions of time slot length, and Δ T is an upper value limit of each time slot. Constraints (17 c) to (17 e) are packet scheduling constraints of ground node equipment, each NOMA group can only communicate with one unmanned aerial vehicle at the same time in each time slot, and L NOMA groups provide communication service through a time division mode, and (17 f) is NOMA decoding rule constraint and decoding is carried out first with better channel gain; (17 g) and (17 h) are task requirement constraints; (17i) For peak power constraints of ground nodes, p min Minimum transmit power, p, in case of scheduling on behalf of a node max Representing the peak power of the node.
S3: designing a combined optimization method of the multi-unmanned aerial vehicle acquisition system based on NOMA according to the specific optimization problem in the step S2, decoupling a target problem into three subproblems, converting non-convex subproblems into convex subproblems by using a block coordinate descent algorithm, 0-1 integer programming, a dichotomy, continuous convex approximation and the like, and then iteratively solving: the optimization problem proposed in step S2 is a non-convex optimization problem of mixed integer fraction, and is difficult to directly solve. Therefore, in step S3, the original problem is first decomposed into three sub-problems, that is, the joint problem of the flight trajectory of the unmanned aerial vehicle and the sub-task segmentation coefficient, the ground device power control, and the NOMA device grouping scheduling problem are optimized respectively. Then, by introducing auxiliary variables, applying a power control algorithm based on a dichotomy, a grouping scheduling algorithm based on 0-1 integer programming and a joint algorithm based on continuous Convex Approximation (SCA) track and subtask allocation, converting the three subproblems into Convex problems to solve, and gradually approximating a final solution of an optimization problem by using a CVX tool box and a cyclic iteration algorithm based on a Block Coordinate Descent (BCD) method.
The invention has the beneficial effects that: the invention aims to solve the problem of task collection of ground equipment in an emergency scene, and constructs a system model in which a plurality of unmanned aerial vehicles collect tasks by using an NOMA auxiliary sensor, and a joint optimization method of a NOMA-based multi-unmanned aerial vehicle collection system. Compared with the method for unmanned aerial vehicle uplink acquisition in the orthogonal multiple access network, the method has the advantages that the unmanned aerial vehicle task completion time is shortened efficiently, the system can run more efficiently, and the method has high application value.
Detailed Description
Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1 and 2, the invention provides a joint optimization method of a multiple unmanned aerial vehicle acquisition system based on NOMA, aiming at the problem of task acquisition of ground equipment in an emergency scene. The method comprises the following steps:
s1: constructing a multi-unmanned-aerial-vehicle-assisted non-orthogonal multiple access emergency acquisition data network model, which comprises a network channel model, a data uplink transmission model and a non-orthogonal multiple access grouping and scheduling model; under the condition of a network channel model, carrying out non-orthogonal multiple access grouping and scheduling on nodes, determining dynamic flight tracks of multiple unmanned aerial vehicles and task resource control of ground nodes according to an unmanned aerial vehicle data transmission model, and providing efficient communication service for ground non-orthogonal multiple access;
s2: the task execution duration of all unmanned aerial vehicles is maximized as a target function, and the total time for the unmanned aerial vehicles to acquire data is minimized by considering collision constraint, steering angle constraint, flight speed constraint, NOMA (non-uniform time adjustment) group internal channel gain constraint, peak power constraint of each device, minimum acquisition total data volume constraint of each device node, grouping scheduling constraint of each node and the like:
s.t.(17a)~(17i),(13),(14),(15),(16)
s3: aiming at the mixed integer non-convex problem established in the step S2, the traditional convex optimization method cannot be used for solving directly, and the invention provides an efficient multivariable iterative optimization algorithm based on BCD for solving; specifically, a target problem is decoupled into three sub-problems based on BCD, in the three sub-problems of the invention, a grouping scheduling sub-problem of equipment can be rewritten into an obvious 0-1 integer programming problem, the problem can be solved through an inlingprog solver, a power control problem is a standard convex optimization problem, the problem can be solved through an improved algorithm based on a dichotomy, a combined problem of flight trajectory and sub-task segmentation is non-convex, and the combined problem is firstly converted into an approximate problem through an SCA method to further obtain an optimal solution. Therefore, the original problem is decomposed into three independent sub-problems, when each sub-problem is optimized, the optimization variables of other sub-problems are fixed, and the three sub-problems are iterated and optimized alternately, so that the loop jump-out condition is met finally, and the final solution of the original problem is found.
When the flight trajectory of the unmanned aerial vehicle, the subtask segmentation coefficient and the equipment power control are fixed, the minimum weight expression is as follows:
reformulating the 0-1 integer programming problem, and rewriting the variable ground node grouping scheduling problem into:
s.t.(17b)(17c)(17d)(17e)(17f)(17h)
the optimal solution of the problem (P1) can be obtained through an integer linear programming solver, the obtained solution updates local optimization variables, and then the solution with the r-th sub-optimization can be used as the initial parameters of other local optimization problems for r +1 times. Further, solving a non-convex sub-problem based on fixed ground node transmitting power and a grouping scheduling variable by adopting an SCA method, wherein the method comprises the following steps:
s.t.
first by introducing auxiliary non-negative variables
As an upper bound on the right of the constraint (21 a), then the constraint (21 a) can be expressed as:
the i, j, m e L represents the scheduling grouping number corresponding to the L groups of equipment in each time slot, the generality is not lost, the variable i e L is simplified and defined as the unique group identifier in each time slot, and as the symbol corresponding to the strong equipment variable i in one time slot also belongs to the unique identifier, the group number of the group in each time slot can be uniquely represented by the strong equipment i; defining relaxed non-negative variables for inter-group scheduling with TDMA and intra-group scheduling with NOMA
The constraint (22) can also be re-expressed as:
can give a relaxation variable greater than zero
And auxiliary variables
The constraints are as follows:
wherein
An equivalent expression (P2-1) of (P2) is obtained as follows:
s.t.
(21b),(21d)-(21g),(23a),(24b)-(24e)
due to the non-convex constraints (25 b), (25 c), (25 d), (21 e), (21 g), (21 d), the problem (P2-1) is still non-convex, due to the fact that
And
and x
2 And
is convex, then constraint (25 b) and constraint (25 c) can relax to the following convex constraint:
wherein
Values are taken for corresponding first-order taylor local points in the r-th iteration, and:
furthermore, for non-convex constraints (21 e) (24 d) (25 d), by at a given point
Applying a first order Taylor expansion corresponding to each of the first and second expansion coefficients
And
it is possible to obtain:
for the unmanned aerial vehicle steering angle constraint in (21 g), first define
It is thus possible to obtain:
the constraint (21 g) can be rewritten as:
l T θ-cos min ||l||·||θ||≥0 (30)
since both left terms are non-convex, we find that the concave bound of each term relaxes the constraint:
-cos min ||l||·||v||≥-0.5cos min (ε||l|| 2 +ε -1 ||θ|| 2 ) (32)
wherein
By using the constraints in (31) and (32), the lower bound of (21 g) is:
therefore, we can get the upper bound solution (P2-2) of (P2-1) by solving the following convex problem, as follows:
s.t.
(21b)(21d)(21f)(23a)(24b)(24c)(24e)(24a)(24a)(24b)(33)
the problem (P2-2) can be solved using a standard convex optimization tool, such as CVX, with the local optimization variables updated with the solution found, and then the r-th sub-optimal solution can be used as the initial parameter for the r +1 other local optimization problems.
When the flight trajectory of the unmanned aerial vehicle, the subtask segmentation coefficient and the equipment grouping scheduling variable are fixed, the power control subproblem can be rewritten as follows:
s.t.
(17i)(23a)
wherein
Although the constraint (35 a) is still non-convex, using the relationship between transmit power and transmission time, we can design an algorithm to get the optimal solution. In (35), since the slicing coefficient of the transmission data, the flight path of the drone and the packet scheduling condition of the nodes are fixed, the task completion duration at each time slot is regarded as a function related to only the node transmission power. And, because of the independence of each node's transmit power within each time slot, each term in the right side of equation (35 a) is independent of each other, and because of the uniqueness of the ground node scheduling packet, i.e., each node device is scheduled at most once within a time slot, and also occurs at most once within each group, then within each time slot nMinimization
Is equivalent to (P3), then defined:
it is easy to know the transmission power p of the strong devices in each group
i,n And the time slot length T
n In a negative correlation trend, it is known that under the constraint of (17 i), the transmission power of a strong device must be maximized to make the left side of the constraint (35 a) smaller, i.e., the transmission power of a strong device must be maximized
While the transmit power p for the weak devices within each group
j,n Since it is already clear in (17 a), function f
1 (p
j,n ) At p
j,n Dependent variable p within a constrained range
j,n Monotonically increasing, function f
2 (p
j,n ) At p
j,n Dependent variable p within constraint range
j,n Monotonically decreasing, F (p)
j,n )=max{f
1 (p
j,n ),f
2 (p
j,n ) Due to f
1 ,f
2 The optimal solution can be obtained by designing a search algorithm based on the dichotomy.
Therefore, the original problem is decomposed into three independent sub-problems (P1, P2 and P3), a non-convex sub-problem (P2) in the original problem is converted into a convex problem (P2-2) by using an SCA method, optimization variables of other sub-problems are fixed when each sub-problem is optimized, the three sub-problems are updated through iteration and optimized alternately, the error accuracy requirement is met finally, and a final solution of the original problem is found.
Finally, the above preferred embodiments are intended to illustrate rather than to limit the invention, and although the invention has been described in detail by way of the foregoing preferred examples, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention as defined by the appended claims.