CN110062345B

CN110062345B - Unmanned aerial vehicle-Internet of things data acquisition method and system

Info

Publication number: CN110062345B
Application number: CN201910280068.2A
Authority: CN
Inventors: 林晓辉; 代明军; 毕宿志; 王晖; 苏恭超; 陈彬
Original assignee: Shenzhen University
Current assignee: Shenzhen University
Priority date: 2019-04-09
Filing date: 2019-04-09
Publication date: 2022-03-22
Anticipated expiration: 2039-04-09
Also published as: CN110062345A

Abstract

The invention relates to an unmanned aerial vehicle-Internet of things data acquisition method, which comprises the following steps: s1, constructing an unmanned aerial vehicle-Internet of things data acquisition system, wherein the unmanned aerial vehicle-Internet of things data acquisition system comprises a sensor which is arranged on the ground and used for data acquisition and an unmanned aerial vehicle which is communicated with the sensor, an RFID label is attached to the sensor to store acquired data, and an RFID reader is loaded on the unmanned aerial vehicle; s2, when the unmanned aerial vehicle flies over a deployment area, acquiring and receiving data stored on the RFID label on the sensor through an ATG channel based on a frame time slot algorithm. By implementing the unmanned aerial vehicle-Internet of things data acquisition method and system, not only can sensor data in remote areas or dangerous areas or other areas which are not suitable for the traditional multi-hop relay method in the mobile ad hoc network be acquired, but also the optimal energy efficiency can be obtained.

Description

Unmanned aerial vehicle-Internet of things data acquisition method and system

Technical Field

The invention relates to the field of unmanned aerial vehicles and Internet of things, in particular to an unmanned aerial vehicle-Internet of things data acquisition method and system.

Background

In the last two decades, the proliferation of the internet of things has been driven by the tremendous advances in wireless technology, microsensors, RFID and embedded systems, coupled with the enormous demands of industrial automation and smart home networks. In the internet of things system, sensors or RFID labels are attached to data collection targets, and a large number of physical and virtual 'things' are seamlessly integrated into the internet, so that the related targets of remote monitoring and intelligent control are realized.

While the application of the internet of things seems attractive, data must first be collected at the network edge side for further analysis and processing at the cloud center. Data collection is easy in areas with rich infrastructure support. However, data collection is very difficult or impossible to achieve in remote areas where the cost of deploying telecommunication and power infrastructure is very high, or where the operating environment in these areas is hostile and inaccessible. For example, where a large number of internet of things sensors are placed in harsh terrain for environmental monitoring or wildlife tracking, data collected from internet of things cannot be transmitted to the outside world because these remote areas are not within the coverage area of a cellular network; as another example, in a pasture, where an RFID tag is attached to the ear of each cow, and the physiological and positional data is collected continuously, it is not feasible to collect the data manually due to the highly random mobility of the herd; also for example, in unmanned automated terminals deploying the internet of things, personnel are not allowed to access hazardous cargo handling areas, and automation is required for collecting information from sensors/tags installed in containers urgently needed by the terminal operator.

In the application scenario described above, the traditional multihop relay approach in mobile ad hoc networks is not feasible, since the sensors are powered by miniature batteries and have poor communication and computation capabilities, so the large amount of IoT data will quickly overwhelm the network. Therefore, the deployment and application of the internet of things are greatly restricted by the inherent physical limitation and the severe environmental limitation.

Therefore, there is a need for a method and system that can collect sensor data located in remote or hazardous areas or other areas where conventional multihop relay methods in mobile ad hoc networks are not suitable.

Disclosure of Invention

The technical problem to be solved by the present invention is to provide a method and a system for acquiring data of an unmanned aerial vehicle-internet of things, which can not only acquire sensor data in remote areas or dangerous areas or other areas not suitable for the conventional multi-hop relay method in a mobile ad hoc network, but also obtain optimal energy efficiency, aiming at the above-mentioned defects in the prior art.

The technical scheme adopted by the invention for solving the technical problems is as follows: an unmanned aerial vehicle-Internet of things data acquisition method is constructed, and the method comprises the following steps:

s1, constructing an unmanned aerial vehicle-Internet of things data acquisition system, wherein the unmanned aerial vehicle-Internet of things data acquisition system comprises a sensor which is arranged on the ground and used for data acquisition and an unmanned aerial vehicle which is communicated with the sensor, an RFID label is attached to the sensor to store acquired data, and an RFID reader is loaded on the unmanned aerial vehicle;

s2, when the unmanned aerial vehicle flies over a deployment area, receiving collected data stored on the RFID label on the sensor through an ATG channel based on a frame time slot algorithm.

In the data acquisition method for the unmanned aerial vehicle-internet of things, step S2 further includes:

s21, calculating an optimal frame length coefficient of the unmanned aerial vehicle for receiving the acquired data;

and S22, controlling the unmanned aerial vehicle to receive the acquired data based on the optimal frame length coefficient.

In the data collecting method for the unmanned aerial vehicle-internet of things, the step S21 further includes:

s211, calculating an optimal energy efficiency frame length coefficient based on the optimal energy efficiency of the unmanned aerial vehicle-Internet of things data acquisition system; and/or

S212, calculating the optimal system efficiency frame length coefficient based on the optimal system throughput of the unmanned aerial vehicle-Internet of things data acquisition system.

In the data acquisition method for the unmanned aerial vehicle-internet of things, the step S211 includes:

s2111, respectively calculating system efficiency and energy efficiency according to the following formulas (7) and (10);

wherein eta_ERepresents energy efficiency; eta_sRepresenting the efficiency of the system, E_OIndicating the energy consumption of the sensor by listening to QueryRep at the beginning of each time slot, E_sRepresenting the energy consumption of the sensor in the successful time slot, E_cRepresenting the energy consumption of the sensors in the conflict time slot, L representing the frame length, N representing the number of sensors, alpha_tIndicating the normalized empty slot period, beta_tRepresenting a normalized collision slot period;

s2112, rewriting equation (7) and (10) to equation (13) based on the frame length and the number of sensors being far greater than 1

S2113, obtaining the optimal energy efficiency according to the formula (13) to obtain the optimal energy efficiency frame length L_{E_opt}；

S2114, solving the optimal energy efficiency frame length coefficient according to the formula (13) by maximizing the system efficiency

Wherein alpha is_eRepresenting the normalized energy consumption of the synchronous monitoring in the time slot timing of the sensor; beta is a_eRepresenting the normalized energy consumption of the sensor in the collision time slot.

In the data collecting method for the unmanned aerial vehicle-internet of things, the step S212 includes:

s2121, respectively calculating system efficiency and energy efficiency according to the following formulas (7) and (10);

s2122, rewriting equations (7) and (10) to equation (13) based on the frame length and the number of sensors far greater than 1

S2123, based on the preset parameter alpha_t、β_t、α_e、β_eCalculating optimal system efficiency frame length coefficient beta_{S_opt}Is a constant value.

In the data collecting method for the unmanned aerial vehicle-internet of things, the step S2 further includes:

s2a, calculating the flight speed, the flight height and the frame length coefficient of the unmanned aerial vehicle;

s2b, controlling the unmanned aerial vehicle to fly over the deployment area based on the flying speed and the flying height, and controlling the unmanned aerial vehicle to receive the acquired data based on the frame length coefficient.

In the data acquisition method for the unmanned aerial vehicle-internet of things, the step S2a further includes:

s2a1, constructing constraint equations (18.a) - (18.c) from optimal energy efficiency and performance constraints:

T_f＝dlα[T₀βe^-1/β+T_Se^-1/β+T_C(β-e^-1/β-βe^-1/β)](18.c)；

where l denotes the unmanned aerial vehicle-sensor effective communication area on the ground, v denotes the flying speed of the unmanned aerial vehicle, d denotes the density of the sensors arranged on the ground, T_fRepresents the time length of each reading cycle, epsilon represents the data collection rate limit, beta represents the frame length coefficient, T₀，T_cAnd T_sRespectively indicating the time lengths of null, conflict and successful time slots;

s2a2, selecting the frame length coefficient, the flying speed and the flying height by adopting PSO positioning according to the constraint equations (18.a) - (18. c).

In the data collecting method for the unmanned aerial vehicle-internet of things, the step S2a2 further includes:

s2a21, limiting the flight height to the lowest flight height;

s2a22, selecting the flying speed according to actual needs;

and S2a23, respectively solving an optimal system efficiency frame length coefficient and an optimal energy efficiency frame length coefficient which meet the optimal system efficiency and the optimal energy efficiency according to the flight speed.

s2a24, switching the data acquisition working state of the unmanned aerial vehicle-Internet of things according to the optimal system efficiency frame length coefficient and the optimal energy efficiency frame length coefficient.

The technical scheme adopted for solving the technical problems is that an unmanned aerial vehicle-Internet of things data acquisition system is constructed, and the unmanned aerial vehicle-Internet of things data acquisition system comprises a sensor which is arranged on the ground and used for data acquisition and an unmanned aerial vehicle which is communicated with the sensor, wherein an RFID label is attached to the sensor to store acquired data, and an RFID reader is loaded on the unmanned aerial vehicle; when the unmanned aerial vehicle flies over a deployment area, receiving collected data stored on an RFID label on the sensor through an ATG channel based on a frame time slot algorithm; wherein the drone comprises a processor and a computer program stored on the processor, the computer program when executed by the processor implementing the steps of:

s22, controlling the unmanned aerial vehicle to receive the acquired data based on the optimal frame length coefficient; and/or

By implementing the unmanned aerial vehicle-Internet of things data acquisition method and system, not only can sensor data in remote areas or dangerous areas or other areas which are not suitable for the traditional multi-hop relay method in the mobile ad hoc network be acquired, but also the optimal energy efficiency can be obtained. Further, by controlling the frame length of data acquisition performed by the unmanned aerial vehicle, the optimal energy efficiency or the optimal system throughput can be obtained. Further, by adjusting the flight speed, height, and frame length of the drone, optimal energy efficiency may be achieved.

Drawings

The invention will be further described with reference to the accompanying drawings and examples, in which:

fig. 1 is a flowchart of a first embodiment of a data collection method for an unmanned aerial vehicle-internet of things of the present invention;

fig. 2 shows the range of variation of the drone height h and coverage radius R;

FIG. 3 illustrates the optimal system efficiency and energy efficiency as the frame length changes;

FIG. 4 illustrates the optimal system efficiency and optimal frame length coefficient for optimal energy efficiency as a function of the number of sensors;

FIG. 5 illustrates a trade-off between system optimal system efficiency and optimal energy efficiency; a

FIGS. 6A-D show calculated PHY-MAC parameters, respectively;

FIG. 7 illustrates a one-dimensional search algorithm for an optimal energy-efficient frame length coefficient;

FIG. 8 shows a search line comparison diagram for two search algorithms;

FIGS. 9A-B show preferred PHY-MAC parameters, respectively;

FIG. 10 shows the efficiency in "system efficiency mode" and "energy efficiency mode", respectively;

FIG. 11 shows the number of competing sensors per round of interrogation in "system efficiency mode" and "energy efficiency mode";

FIG. 12 shows the number of sensors associated with airspeed in "system efficiency mode" and "energy efficiency mode";

FIG. 13 illustrates a one-dimensional search algorithm for an optimal frame length;

FIGS. 14 and 15 show the optimal frame length variation and efficiency variation, respectively, for different sensor densities;

fig. 16 shows the energy efficiency gain obtained.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Fig. 1 is a flowchart of a data acquisition method for an unmanned aerial vehicle-internet of things according to a first embodiment of the present invention. As shown in fig. 1, in step S1, an unmanned aerial vehicle-internet of things data acquisition system is constructed. In a preferred implementation of the invention, the unmanned aerial vehicle-internet of things data acquisition system comprises a sensor arranged on the ground for data acquisition and an unmanned aerial vehicle communicating with the sensor, wherein an RFID tag is attached to the sensor to store acquired data, and an RFID reader is loaded on the unmanned aerial vehicle. In step S2, when the drone flies over the deployment area, the acquired data stored on the RFID tag on the sensor is received through the ATG channel based on a frame time slot algorithm. In a preferred embodiment of the present invention, in step S2, an optimal frame length coefficient for the drone to receive the collected data may be first calculated; and then controlling the unmanned aerial vehicle to receive the acquired data based on the optimal frame length coefficient. In a further preferred embodiment of the present invention, the optimal frame length coefficient may be an optimal energy efficiency frame length coefficient, or may be an optimal system efficiency frame length coefficient. For example, the optimal energy efficiency frame length coefficient may be calculated based on the optimal energy efficiency of the drone-internet of things data acquisition system; the optimal system efficiency frame length coefficient may also be calculated based on the optimal system throughput of the drone-internet of things data acquisition system. In another preferred embodiment of the present invention, in step S2, the flight speed, flight altitude and frame length coefficient of the drone may be calculated; and then controlling the unmanned aerial vehicle to fly over the deployment area based on the flying speed and the flying height, and controlling the unmanned aerial vehicle to receive the acquired data based on the frame length coefficient.

By implementing the unmanned aerial vehicle-Internet of things data acquisition method, not only can sensor data in remote areas or dangerous areas or other areas which are not suitable for the traditional multi-hop relay method in the mobile ad hoc network be acquired, but also the optimal energy efficiency can be obtained. Further, by controlling the frame length of data acquisition performed by the unmanned aerial vehicle, the optimal energy efficiency or the optimal system throughput can be obtained. Further, by adjusting the flight speed, height, and frame length of the drone, optimal energy efficiency may be achieved.

The data acquisition method of the unmanned aerial vehicle-internet of things of the invention is further described below with reference to specific embodiments.

In the unmanned aerial vehicle-internet of things data acquisition method, the unmanned aerial vehicle downloads data stored in ground internet of things equipment, namely a sensor, through the ATG channel, and compared with the ground channel, the ATG channel is easy to meet the LoS sight line transmission condition due to the high flying height of the unmanned aerial vehicle. The path loss of the ATG channel depends on the distance D between the transmitter and the receiver, the wave frequency f and statistical parameters of the terrestrial environment. In this context, we use the known LoS probabilistic model, which is a curve fit based on measured data, that can be approximated by a simplified Sigmoid function

Wherein Prb_LosIs the LoS probability of the ATG channel, h is the altitude of the drone, r is the radius of the coverage area of the ATG channel, and a and b are fitting parameters closely related to the ground environment. Thus, the probability that the ATG channel does not satisfy the LoS condition is Prb_NLos＝1-Prb_Los。

The path loss of the ATG channel for LoS and NLoS can be expressed as:

in which ξ_LoSAnd xi_NLoSThe extra path loss associated with the LoS and NLoS channels respectively,

the distance between the unmanned aerial vehicle and the sensor, and c is the speed of light; the ATG channel average path loss can be written as:

to effectively carry out the operation between the unmanned aerial vehicle and the sensorCommunication, we assume that the maximum path loss that an ATG channel can tolerate is PL_maxIt corresponds to a maximum R-R, i.e. the path loss of only ground equipment located within the drone coverage circle of radius R is less than PL_maxSo only the nodes within the circle can send their data to the drone. Let

And R ═ R, then by substituting equations (1) and (2) into (3), we obtain:

in (4) we can see that the radius of coverage is an implicit function of the altitude h of the drone, so we derive h with respect to R such that the derivative is 0, obtaining

We then get the optimal height to maximize the coverage of the drone. Fig. 2 depicts the variation range of R with respect to height h, wherein the parameters a-1, b-0.65, ξ_LoS＝0.1，ξ _NLoS21, and 915 MHz. As can be seen from the figure, the rise in height increases the LoS probability, resulting in a larger radio coverage. However, this effect is offset by a further increase in the altitude of the drone, since an increase in altitude also leads to an increase in the transmission distance, and thus to a higher attenuation of the radio signal.

We assume that an RFID tag is attached to each sensor on the ground. Data collected by the sensors is stored on the tag and transmitted to the drone on which the RFID reader is installed. The tag is powered by a battery on the sensor and is an active tag. Because the ATG channel is beneficial to LoS transmission and the label is active, the effective communication range can reach hundreds of meters, and the method is suitable for long-distance transmission in suburb and field environments. We discuss RFID sensor air data acquisition under the EPC global standard framework, which is the most popular standard in widespread use in various industries around the world. The standard is based on a frame slotted aloha (fsa) in which the communication time is divided into successive frames consisting of a number of slots, thereby randomizing channel access between tags and reducing the probability of transmission collisions. The mechanism is summarized as follows:

at the beginning of each frame, the reader broadcasts a "Query" command, which includes the frame length (number of slots) to initiate a polling cycle (inventoryround). Upon receiving the command, nearby tags randomly and independently select slots in the frame to transmit their stored data. The position of the slot in the frame is used as a transmission counter. If the counter is zero, the label can immediately send a data packet; the reader initiates the slot using the "QueryRep" command. Upon hearing this command, each tag decrements its counter by 1. When the counter reaches zero, the tag contends for the slot by sending a 16-bit long "RN 16" packet to the reader containing the tag temporary ID; if multiple tags send "RN 16" packets to the reader at the same time, a collision occurs and the time slot is wasted. If no tag is sent for the "RN 16" packet, the slot is empty. In either case, the reader will start the next slot by broadcasting a "QueryRep" command, and each waiting tag will again decrement its counter by 1; if only one tag sends "RN 16" to the reader, the reader acknowledges receipt by sending an "Ack" packet back to the tag, acknowledging successful reservation of the slot. The tag then sends its stored data (including the PC, EPC, stored data and CRC-16) to the reader, completing the data collection.

From the above, it can be seen that one frame is composed of a plurality of different types of slots, and T is arranged in ascending order according to the time sequence length₀<T_c<T_sWherein T is₀，T_cAnd T_sRespectively, empty, collision, time length of successful slot. The average number of different time slots in a frame depends on the number of sensors N and the frame length L. Mathematically, they can be expressed as:

wherein alpha is₀，α_sAnd α_cRespectively, indicate null in the frameThe average number of successful and conflicting slots. Channel utilization efficiency is defined as a percentage of the total length of successful slots in a frame and can be written as:

we pass the setting of alpha_t＝T₀/T_sAnd beta_t＝T_c/T_sWill T₀And T_cRelative to T_sAnd (6) carrying out normalization. Substituting (5) into (6), we have

(7)；

For a given number of sensors N, the optimum system efficiency is a function of the frame length L. Thus to η_sTaking the derivative with respect to L and letting the derivative

To obtain

1-N/L＝(1-α_t/β_t)(1-1/L)^N (8)

By solving equation (8), we obtain the optimal frame length L that maximizes throughput_{S_opt}I.e., the reader can collect data from the surface sensors at the highest throughput. The following table lists the relevant symbolic representations used in this section.

We now analyze the energy consumption of the surface sensor. In each data collection round, the sensor randomly selects a time slot in the frame for transmission. Each transmission may be successful (single slot) or unsuccessful (conflicting slots) resulting in E, respectively_sAnd E_cThe energy consumption of (2). In addition, at the beginning of each slot, the sensor needs to listen for "QueryRep" commands for slot synchronization, resulting in energy consumption E_O. Thus, on average, the energy consumed by each sensor in a frame due to listeningIs that

Similarly, we set α_e＝E_O/E_sAnd beta_e＝E_c/E_sTo E_OAnd E_cRelative to E_sAnd (6) carrying out normalization. We assume that there are N sensors contending for the L slots in the frame. Thus, energy efficiency (defined as the percentage of energy used by all sensors to successfully transmit data in each frame) can be written as:

substituting (5) into (9) to obtain

Let β be L/N, provided that L > 1 and N > 1, then (1-1/L)^N-1≈e^-1/βEnergy efficiency is thus reduced to

When N > L, the number of sensors competing with each other far exceeds the frame length, and beta → 0, eta is obtained_E→ 0, this means that all slots in a frame collide and there is no energy available for successful data transfer. When L > N, 1/β → 0 is obtained and

this means that there is no slot collision and energy is used for packet transmission and synchronous listening of slots. If we further assume E_s＞＞E_OThis means that the energy consumption for slot-synchronous listening is negligible, resulting in η_E→1。

Given the number of sensors N, the maximum energy efficiency in (10) is equal to the minimization function

The first and second derivatives of f (L) are:

obviously, f' (L)>0, so f (L) is a convex function of L, meaning that there is a unique L when f (L) is minimized_{E_opt}. Let f' (L) be 0, solve equation E_OL²-E_ONL-2E_cN ═ 0, yielding two roots:

due to L_{E_opt}>0, the frame length for optimum energy efficiency can be expressed as

We take the physical parameters (40 Kbps channel data rate based on BPSK and 96 bit ID) and set N to 1000. Using equations (7) and (9), we can get the system efficiency and energy efficiency as the frame length changes, i.e. fig. 3. As can be seen in FIG. 3, L_{E_opt}And L_{S_opt}Is different, this also means that system efficiency and energy efficiency cannot be optimized simultaneously. And there is a mutual constraint relationship between the two. In UAV-IOT data collection systems, we assume N > 1 and L > 1, so equations (7) and (10) can be rewritten as:

will eta_sWriting

Where g (β) ═ α_t-β_t)β+β_tβe^1/β. The first and second derivatives of g (β) are written as g' (β) ═ α, respectively_t-β_t)+β_te^1/β-β_tβ^-1e^1/βAnd g' (β) ═ β_tβ^-3e^1/β. It can be seen that g' (beta)>0, and thus g (β) is a convex function of β, which represents the system efficiency η_sIs concave. Let g' (β) be 0 and solve the equation to obtain_sMaximized unique beta_{S_opt}. In equations (11) and (12), we obtain η_EMaximized unique solution

We therefore set the optimal energy efficiency frame length coefficient to

In the above discussion we see β_{S_opt}Is the root of the equation g' (β) ═ 0, so β_{S_opt}Is constant, however

The number of sensors N is closely related. According to the set parameters: alpha is alpha_t＝0.01，β_t＝0.65，α_e＝0.025，β_eWhen equation g' (β) is solved to 0.12, root β is obtained_{S_opt}5.998, which is a constant value independent of the number N of sensors.

We also found in solving equation (7) that the root is very close to 5.998 (see fig. 4) unless the number of sensors N is less than 10. When N > 1, the optimal length coefficient in the "system efficiency mode" is the constant value and can be approximated to be β_{S_opt}＝5.998。

In fig. 3 and 4, it is observed that the drone-internet of things data acquisition system cannot work in both the optimal "system efficiency mode" and the optimal "energy efficiency mode". We can only achieve higher system throughput at the cost of more energy consumption and vice versa. Thus, there is a tradeoff between the two metrics. To explain this fact more clearlyLike, let the frame length coefficient β be in the interval [ β [ ]_{E_opt}，β_{S_opt}]And calculates η using equation (13)_sAnd η_E. The results are shown in FIG. 5. In the figure, we can observe the mutual constraint relation between the two efficiencies, so that we cannot optimize the two efficiencies simultaneously. Furthermore, we observe that the energy consumption of the sensor and the system throughput can be adjusted by adjusting the frame length. Therefore, by appropriately adjusting the frame length according to the application scenario, we can choose to let the system operate in the best "system efficiency mode" or the best "energy efficiency mode".

In the unmanned aerial vehicle-internet of things system, the frame length setting of the MAC layer can affect the system throughput and energy consumption. On the other hand, the parameters of the PHY layer, drone altitude and speed, may actually affect both of these metrics as well. An increase in flight speed may allow more data to be collected per unit time, but this may also make the channel contention more aggressive, thereby increasing energy consumption; lower flight speeds may mitigate channel contention, but may result in inefficient utilization of channel resources. Also, variations in flying altitude can lead to fluctuations in radio coverage, system throughput, number of competing sensors, and associated ground node energy and system throughput. In this section, we will discuss how to adjust the three parameters of the PHY-MAC layer to avoid unnecessary ground node energy and channel resource waste. Specifically, in the data acquisition system of the unmanned aerial vehicle-internet of things, the optimal PHY-MAC parameters are found, aiming at saving energy as much as possible and meeting the required system performance.

When the drone flies over the deployment area at speed v, the sensor transmits the internet of things data to the drone. And l represents the effective reading range of the onboard reader. I.e. only sensors within this range can reliably send stored data to the drone. The radius of the coverage area is R, which is determined by the previously analyzed flying height h of the drone, where l is 2R. Note that the time limit for the sensor to transmit its data is l/v, and after this time limit, the drone will fly out of the effective communication range. The density of deployed sensors is d, i.e. on average, there are d sensors per meter along the flight trajectory on the ground. The following table gives the meanings of the symbols used in this section

In each round of reading, the sensor contends for channel access by randomly selecting a time slot in a frame to transmit a data packet. Thus, the sensor may go through multiple read cycles until its stored data is eventually sent to the drone. We assume that at the beginning of each round, the percentage of sensors to be read is α, and therefore the number of sensors waiting to be transmitted in the effective communication area is dl α, and therefore the frame length should be set to dl α. According to equation (5), the percentage of sensors that successfully send their data to the drone in the waiting dl α sensors after each round of reading is

Let T_fIs the average length of the frame length and can be written as T_f＝α₀T₀+α_sT_s+α_cT_c. Assuming that L ═ dl α β > 1 and N ═ dl α > 1, by using equation (5), we can simplify T by the following approximate formula_f：

Assuming that the effective read range is l and the drone flight speed is v, the maximum number of read wheels that the sensor may experience is

The number of newly added sensors is dl/k for each reading round. The number of sensors awaiting transmission in the communication area is therefore

Thus, we obtained

To provide quality assurance in data acquisition, it is specified that after k rounds of reads, the percentage of unread sensors does not exceed e, i.e. we have a performance constraint (1-P)_I)^k≦ e, which may be further written as

For data acquisition, we need to reduce the energy consumption of the ground sensors as much as possible while satisfying the performance constraints in (17). Therefore, according to the definition of energy efficiency in (10), the optimization problem can be expressed as:

it should be noted that both the frame length L dl α β and the effective drone-sensor communication zone L on the ground are affected by the flight height h of the drone. In the optimization problem of equation (18), it can be observed that the minimization of the objective function requires the proper setting of the PHY-MAC variables v, h and β. In addition, subject to the constraints of (18.a) - (18. c). Changes in each variable will result in the adjustment of the other two variables, which increases the complexity of the problem.

We can see that the objective function and constraints in the problem (18) are non-convex, so we use a heuristic method-Particle Swarm Optimization (PSO) to find the best solution. The PSO solves the optimization problem by the following method: for a given quality criterion, the candidate solution (also called a particle) is iteratively improved. In particular, it will move the particle group in the search space based on the position and velocity of the individual of the particle group. The movement of a particle is determined by its own local best known position and the best known positions found by other particles in the whole population. Both positions are updated whenever a better position is found, so the group will move to the best solution when the PSO eventually converges.

Before solving the problem using PSO, we need to decide the search space for the set of parameters v, h, β. For unmanned aircraft flying heights, there is a minimum safe altitude limit on the unmanned aircraft to avoid collisions caused by ground obstacles. In this context, we set h_minIs 10 m. At the same time, the maximum flying height is set to h_maxAt this height, the drone has maximum ground coverage. By applying the method of equation (4)

The maximum flying height can be calculated. Thus, we obtain the height of the drone h_min，h_max]The search space of (2). For the frame length coefficient beta, the optimal system efficiency mode in the figure is beta_{S_opt}Since 5.998, it can be set to an upper limit value of β, β_max＝β_{S_opt}To ensure that the search space contains the best values, we choose a relatively low lower limit value β_min0.2. The search space range for β is therefore [ β ]_min，β_max]＝[0.2，5.998]. Next, we determine the search space [ v ] for the flight speed of the drone_min，v_max]Obtained by equations (18.b) and (18.c)

To simplify equation (19), let a₁＝T₀βe^-1/β+T_Se^-1/β+T_C(β-e^-1/β- βe^-1/β)，a₂＝dve^1/β，a₃＝1-e^-1/βThus, equation (19) is written

Thus, the maximum number of read rounds is:

when k is>0 and 0<a₃<1, obtaining a₁a₂>1, this means v>(da₁e^1/β)^-1. Thus let v_min＝(da₁e^1/β)^-1. We now analyze the upper limit of the maximum airspeed. Constraints (18.a) impose a limit on the flight speed, which can be rewritten as:

at the same time, the boundary conditions of the constraint (18.a) are written

We can observe that (18.b), (18.c) and (22) are independent equations with three variables { v, l, β } (note that there is a one-to-one mapping between h and l). By setting beta ═ beta_{S_opt}We can obtain the boundary velocity v by solving equations (18.b), (18.c) and (22) using numerical methods₂. EdgeVelocity v₂Corresponding to the flight speed of the drone in the optimal system efficiency mode. Therefore we set v_max＝min(v₁，v₂) So we obtain the search space for v [ v ]_min，v_max]。

In the PSO algorithm, to obtain a balance between solution quality and computation time, we set the population size and the number of iterations to 200 and 300, respectively. The sensor parameters are listed in table 3. The calculated PHY-MAC parameters { v, h, β } are shown in FIGS. 6A-6D. In the figure we observe that the optimal flying height is located at the lowest boundary point of the search space. This also means that in order to save energy of the ground sensors, we should limit the number of competing sensors by setting the radio coverage to a minimum. The mechanism of this parameter setting will be analyzed in detail in the latter section. At the same time, we also observed that the optimum flying speed decreases with increasing sensor density. This is because the increase in sensor density results in an increase in data acquisition load and to meet performance constraints, the flight speed must be slowed to avoid system saturation due to the ever increasing number of sensors. From the figure we also observe that the optimum flight speed is at the lowest boundary point v of the search space_min＝(da₁e^1/β)^-1To (3).

To illustrate this more clearly, we compute β in FIG. 6B_{E_opt}Substitution v_min＝ (da₁e^1/β)^-1. By v_minBy calculating v_opt/v_minNormalization v_opt. As can be seen in the result shown in fig. 6C, the normalized value fluctuates slightly around 1, which means that the calculation result converges on the boundary point v_min. The following table shows the sensor parameters used

We rewrite the energy efficiency in equation (10)

It can be seen that for a given frame length, an increase in the number of sensors N may result in η_EAnd decreases. In particular, when the system operates in the optimal "energy efficiency mode", this is obtained

Therein, the

At the same time, the user can select the desired position,

when increasing the number of sensors N, L_{E_opt}And

will increase so that eta_EAnd decreases. This means that in order to save as much energy as possible on the ground sensors, we should set the flying height h to a minimum value to reduce the radio coverage area and thus limit the number of competing sensors.

With the proviso (18.c), obtaining

Substituting it into (18.b) to obtain

When the frame length coefficient beta is fixed, a₁The value is also fixed (see a)₁So that a depends only on the flight speed v. Because v is>v_min＝(da₁e^1/β)^-1And therefore 0<(dva₁e^1/β)^-1<1 and ln [1- (dva)₁e¹ ^/β)^-1]<0. And because of beta>0，ln[1- e^1/β]<0, percentage of competing sensors is

Let f (v) ═ vln [1- (dva)₁e^1/β)^-1]Therefore, f' (v) ═ ln [1- (dva) ]₁e^1/β)^-1]- (dva₁e^1/β-1)^-1. To further simplify f' (v), let x ═ dva₁e^1/β)^-1Thus, therefore, it is

Because of 0<x<1, by using Taylor expansion, obtaining

And

thus f' (v) can be expressed as

Due to f (v) ═ vln [1- (dva)₁e^1/β)^-1]>0 and f' (v)<0, therefore

Will increase with the flying speed v. In other words, when we increase v, the number of competing sensors N_wDl α will also increase, so the energy efficiency η according to the analysis of the previous paragraph_EWill be reduced. Therefore, to save energy from ground sensors, we should let the drone as much as possible at the lowest speed v_minAnd (5) flying.

From the above analysis, we know that to maximize the energy efficiency of the ground sensors, we can set the drone flight altitude and speed to the lowest allowable values. Thus, the optimization of PHY-MAC parameters v, h, β can be simplified to find β_{E_opt}I.e. minimizing energy consumption. This will greatly reduce the computation time of the heuristic search in the PSO. One-dimensional optimization is given by:

here, h is set to h_minL can be calculated using equations (1-4). At the same time, by setting v ═ v_minAnd alpha can be calculated using equations (18.b) and (18. c). Equation (23) thus reduces to a function with only one variable β. Note that the function (23) is not convex or concave, and to find the extreme points of the function, we can set its derivative value to zero and then choose the best β to minimize the function (23)_{E_opt}And (4) point. At the same time, to ensure a₁a₂>1 and function (23) at v_minNearby differentiable, we set v ═ v_min+ σ, where σ is a sufficiently small positive value (e.g., 10 ═ σ)^-5). The one-dimensional search algorithm is shown in fig. 7. To verify the feasibility of the one-dimensional search algorithm shown in fig. 7, we compared the computed results obtained from this algorithm with those obtained from the PSO heuristic search. The result is shown in FIG. 8, where we can see that the computed result matches the PSO search line very well. This means that we can use a simplified one-dimensional search instead of a time-consuming PSO search scheme.

Using the simplified algorithm presented in fig. 7, we can compute the joint PHY-MAC optimum parameters, the results of which are shown in fig. 9A-B. Note that the optimum flying height was 10 meters in the above analysis. By adopting the parameters, the unmanned aerial vehicle-internet of things system can work in an energy efficiency mode, and in the energy efficiency mode, the energy consumption of the ground sensor can be reduced to the greatest extent, and meanwhile, the performance constraint is met. In fig. 9A, the optimal frame length coefficient β can be seen_{E_opt}Decreases as the sensor density increases, due to

An increase in sensor density results in an increase in the number of competing sensors N, resulting in a number β_{E_opt}And (4) reducing. It is observed in fig. 9B that as sensor density increases, in order to meet performance constraints, the drone must slow down the flight speed in order for the system to accommodate more sensors.

By setting the frame length coefficient beta as beta_{S_opt}We can let the drone-internet of things system work in "system efficiency mode" so that the system can collect data from ground sensors with maximum throughput. In other words, under the same performance constraints, the drone, in the "system efficiency mode", may allow higher flight speeds to be employed, at the expense of more energy consumed by the ground sensors, than in the "energy efficiency mode".

To calculate the maximum drone flight speed in "system efficiency mode", we rewrite the constraint of the function (18) to the bottom. We assume that the system should meet the boundary conditions and therefore the inequality signs in the constraint (18.a) are replaced with the equal signs in (18.a 1).

T_f＝dlα[T₀βe^-1/β+T_Se^-1/β+T_C(β-e^-1/β-βe^-1/β)]＝dlαa₁(18.c)

To calculate the maximum flying speed, in equation (18.a1), let β be β_{S_opt}5.998, then we can calculate

By substituting this calculated value into (18.b), we can also calculate α. Then using (18.a1) and (18.c),

it can adopt beta ═ beta_{S_opt}And alpha calculation. Thus, the maximum speed in the "system efficiency mode" can be achieved by simply changing β to β_{S_opt}And it is independent of the flying height h (or l).

In the "system efficiency mode" of the system,with beta ═ beta_{S_opt}5.998, system efficiency in this mode

Is a constant value and is therefore independent of the flying height h and the flying speed v. To save energy of the ground sensors, we should reduce the number of competing sensors by setting the flying height h (or l) to the lowest allowable value.

We calculated the system efficiency and energy efficiency in the "system efficiency mode" and the "energy efficiency mode", and the results are shown in fig. 10. The subscripts "E _ opt" and "S _ opt" in the figures indicate that the system operates in "energy efficiency mode" and "system efficiency mode", respectively. To reduce sensor power consumption in a "system efficiency mode". We set the drone flight altitude to the minimum allowed value, i.e. h ═ h_min10 m. As shown, we can see that the "system efficiency mode" can provide the highest system efficiency (about 90%), which is 1.2-1.3 times that in the "energy efficiency mode". This also means that in the "system efficiency mode" the ground data throughput of the drone is maximized. Thus, the allowed flying speed of the drone is higher than in the "energy efficient mode" considering the sensor density d on the ground and the performance constraint e. However, the cost is that the energy efficiency is very low (10% or less). Since more sensors will be involved in channel contention as the flight speed increases (see fig. 12). As expected, the energy efficiency in the "energy efficiency mode" is much higher (about 4-6 times) than in the "system efficiency mode" in the figure. Again, we can see in the graph that energy efficiency decreases with increasing sensor density in both modes of operation, since an increase in the number of competing sensors decreases energy efficiency.

Fig. 11 shows the number of competing sensors per round of interrogation for both modes of operation. As shown, the number of competing sensors increases linearly with surface sensor density. Furthermore, higher flight speeds in the "system efficiency mode" also mean heavier data collection loads, and therefore the number of competing sensors is greater than the "energy efficiency mode" (about 2.5 times), making the energy efficiency of the "system efficiency mode" much lower than the energy efficiency of the "energy efficiency mode".

For a given performance constraint e, the maximum allowable airspeed is shown in FIG. 12, where v_{E_opt}、v_{S_opt}Representing the maximum speed in the "energy efficiency mode" and the "system efficiency mode", respectively. We can observe in the figure that as the sensor density increases, the flight speed must be slowed down in order to meet the constraints. As expected, v_{S_opt}Higher than v_{E_opt}(about 1.3-1.4 times), which corresponds to a higher data throughput of the "system efficiency mode". However, as shown in FIG. 10, the cost is about 4-6 times the energy consumption of the "energy efficient mode".

From the above analysis, we observe an interrelationship between two indicators-higher system efficiency, but lower energy efficiency; or higher energy efficiency, but less efficient systems. In other words, we cannot maximize both system efficiency and energy efficiency. The key to the problem is how to balance the two. Clearly, it is not reasonable to have the system always operate in a "system efficiency mode" with high power consumption. The design philosophy should be to consume just enough energy to meet the required system performance requirements. Next, we will discuss how to adjust the system parameters to balance these two conflicting indicators according to the changing flight speed.

In some application scenarios, it is very important to obtain remote environment information in a timely manner, especially in situations where the environmental conditions change dramatically. For example, when solar radiation is strongest at noon during the day, the temperature and humidity of the vineyard may vary greatly during the short hours of noon. Therefore, drones need to fly faster and collect and transmit back sensed environmental data in time. Therefore, according to specific conditions, the unmanned aerial vehicle has a minimum speed constraint v related to specific application_app. Thus, the drone-internet of things system contains two performance constraints-minimum airspeed v_appAnd a maximum allowed data loss rate e. The PHY-MAC parameters for the two modes of operation are listed in the following table (h 10 meters)

	d＝5sen/m	d＝8sen/m	d＝12sen/m	d＝18sen/m
					(β_{E_opt}，v_{E_opt})	(1.041,46.17m/s)	(0.972,28.01m/s)	(0.930,18.30m/s)	(0.901,12.02m/s)
(β_{S_opt}，v_{S_opt})	(5.998,59.94m/s)	(5.998,37.46m/s)	(5.998,24.97m/s)	(5.998,16.65m/s)

In the upper mark, the flight velocities v of the "system efficiency mode" and the "energy efficiency mode" are listed_{S_opt}And v_{E_opt}And related optimum beta_{S_opt}And beta_{E_opt}. In particular. v. of_app>v_{S_opt}Is an infeasible area, in which the performance constraint epsilon cannot be satisfied; if v is_app<v_{E_opt}Setting v ═ v_{E_opt}And β ═ β_{E_opt}To reduce the energy consumption of the ground sensor; if v is_{E_opt}<v_app<v_{S_opt}Setting v ═ v_appAnd beta is properly adjusted to meet the performance constraint e while minimizing energy consumption. Now we discuss how at v_{E_opt}<v_app<v_{S_opt}Beta is adjusted. Let v equal v_appAnd substituted into constraints (18.b) and (18.c), we rewrite the optimization problem to:

T_f＝dlαa₁ (18.c)

with (21), through rearrangement (18.a2), (18.b2) and (18.c), the result is obtained

We then calculate the feasible region B of β that satisfies this inequality₁. At the same time, we have also previously demonstrated v>(da₁e^1/β)^-1We then calculate the feasible region B₂In this interval let v_app>(da₁e^1/β)^-1. Then (18.b2) and (18.c) are substituted into (23), now only one variable β remains in the function. In this way, the minimization of E (β) in (23) becomes again a one-dimensional search solution, which can be solved by changing B to B in the feasible region B₁∩B₂Of'Beta of (beta) ═ 0_appTo solve. The calculation of the optimum β is shown in fig. 13_appThe search algorithm of (1).

In the above table, the flight velocities v in the "system efficiency mode" and the "energy efficiency mode" are listed_{S_opt}And v_{E_opt}And related optimum beta_{S_opt}And beta_{E_opt}. We proceed from the algorithm given in FIG. 13 at v_{E_opt}，v_{S_opt}]Varying in-range application-dependent airspeed v_appAnd calculating the optimum beta_app. The results are shown in FIG. 14, where we observed when v is_appFrom v_{E_opt}Change to v_{S_opt}To satisfy performance constraints, the algorithm can adaptively adjust β_appLet the system automatically go from "energy efficient mode" (β)_app＝β_{E_opt}) Switch to "System efficiency mode" (β)_app＝β_{S_opt}See β for two modes in the table above, 5.998_optValue). The associated changes in system efficiency and energy efficiency corresponding to the changes in FIG. 14 are shown in FIG. 15, where "η_E"and [ ] eta_S"represents energy efficiency and system efficiency, respectively. Also, the speed of flight v as associated with the application, under the constraint of the performance requirement ∈_appWhen added, the system will automatically increase system throughput by sacrificing energy efficiency of the surface sensors according to the calculations in fig. 14. Therefore, v is constrained according to performance_appAnd e, we can use the MAC parameter β_appEnergy consumption and system throughput are adaptively adjusted to maintain a balance between energy efficiency and system efficiency.

According to the optimal beta given in FIG. 14_appIn the "system efficiency mode" (η in FIG. 10), the system efficiency can be adjusted_E，_{S_opt}) The energy efficiency of (b) was normalized, and the result is shown in fig. 16. We have observed that at optimum β_appUnder the regulation of (2), the energy efficiency can be obviously improved. After optimization, the energy efficiency in "energy efficiency mode" is 4-5 times higher than in "system efficiency mode". When v is further increased_appTo meet the performance constraints, the system adapts as shown in FIG. 14Will beta_appFrom beta_{E_opt}Increase to beta_{S_opt}To beta, pair_appAnd adjusting so as to gradually switch the unmanned aerial vehicle-Internet of things platform to a 'system efficiency mode' (beta)_app＝β_{S_opt}And the gain is 1).

From the above analysis, it can be seen that the calculation of the optimal PHY-MAC parameter, i.e., frame length, flight speed of the unmanned aerial vehicle, and altitude, needs to comprehensively consider the relevant information of the MAC control layer and the physical layer. Specifically, by synthesizing the perceived load, performance constraints, channel information from the application and physical layers, the parameter optimizer performs calculations and then outputs the resulting parameters to the MAC layer (frame length) and the physical layer (drone speed and altitude). Under the optimized parameter environments, the unmanned aerial vehicle-Internet of things system can acquire ground data with high energy efficiency and meet system performance constraints. Therefore, inter-layer joint design of the system is necessary because we can better achieve the balance of system performance and energy efficiency.

In this document, we discuss the problem of energy-efficient based data acquisition in drone-to-internet-of-things systems. In particular, we reveal a constraint between system throughput and energy efficiency — we can improve system efficiency at the expense of more energy consumption by the surface sensors, or we can save sensor energy by reducing system throughput. Based on this observation, to strike a balance between system efficiency and energy efficiency, we describe drone-internet of things data acquisition as a non-convex problem. Through PSO heuristic search, the optimal value can be found on the boundary point, so that the complexity of the problem is greatly simplified, and the original problem can be simplified into one-dimensional optimization. By solving the problem, PHY-MAC parameters of optimal energy efficiency, namely frame length, flying speed and height of the unmanned aerial vehicle can be obtained. In the system environment of these optimal parameters, we can achieve approximately 4-5 times gain in energy efficiency while meeting performance constraints through cross-layer design. In addition, we have also found that the frame length of the MAC layer can be used as a "control knob" to adaptively adjust the energy consumption and system throughput according to performance requirements. The discovery has certain reference significance to the algorithm design and the product implementation of the network engineer in relevant aspects.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims

1. An unmanned aerial vehicle-Internet of things data acquisition method is characterized by comprising the following steps:

s2, when the unmanned aerial vehicle flies over a deployment area, receiving collected data stored on the RFID label on the sensor through an ATG channel based on a frame time slot algorithm;

the step S2 further includes:

s2b, controlling the unmanned aerial vehicle to fly over the deployment area based on the flying speed and the flying height, and controlling the unmanned aerial vehicle to receive the acquired data based on the frame length coefficient;

the step S2a further includes:

T_f＝dlα[T₀βe^-1/β+T_Se^-1/β+T_C(β-e^-1/β-βe^-1/β)] (18.c)；

where l denotes the unmanned aerial vehicle-sensor effective communication area on the ground, v denotes the flying speed of the unmanned aerial vehicle, d denotes the density of the sensors arranged on the ground, T_fRepresents the time length of each reading cycle, epsilon represents the data collection rate limit, beta represents the frame length coefficient, T₀，T_cAnd T_sRespectively representing null, conflict, and time length representation of a successful time slot; α represents the percentage of sensors awaiting transmission in the communication zone;

s2a2, according to the constraint equations (18.a) - (18.c) and equation (18)

Selecting the frame length coefficient, the flight speed and the flight height by adopting PSO positioning; e_OIndicating the energy consumption of the sensor by listening to QueryRep at the beginning of each time slot, E_cRepresenting the energy consumption of the sensor in the conflicting time slot.

2. The UAV-IOT data collection method according to claim 1, wherein the step S2a2 further comprises:

s2a21, limiting the flight height to the lowest flight height;

s2a22, selecting the flying speed according to actual needs;

3. The UAV-IOT data collection method according to claim 2, wherein the step S2a2 further comprises:

4. An unmanned aerial vehicle-Internet of things data acquisition system comprises a sensor and an unmanned aerial vehicle, wherein the sensor is arranged on the ground and used for data acquisition, the unmanned aerial vehicle is communicated with the sensor, an RFID tag is attached to the sensor to store acquired data, and an RFID reader is loaded on the unmanned aerial vehicle; when the unmanned aerial vehicle flies over a deployment area, receiving collected data stored on an RFID label on the sensor through an ATG channel based on a frame time slot algorithm; wherein the drone comprises a processor and a computer program stored on the processor, the computer program when executed by the processor implementing the steps of:

the step S2a further includes:

T_f＝dlα[T₀βe^-1/β+T_Se^-1/β+T_C(β-e^-1/β-βe^-1/β)] (18.c)；

where l denotes the unmanned aerial vehicle-sensor effective communication area on the ground, v denotes the flying speed of the unmanned aerial vehicle, d denotes the density of the sensors arranged on the ground, T_fRepresents the time length of each reading cycle, epsilon represents the data collection rate limit, beta represents the frame length coefficient, T₀，T_cAnd T_sRespectively indicate null, conflict and successA time length representation of a power slot; α represents the percentage of sensors awaiting transmission in the communication zone;

s2a2, according to the constraint equations (18.a) - (18.c) and equation (18)