Vehicle networking V2V2I routing method based on adaptive variable beacon power

Abstract

The invention discloses an Internet of Vehicles V2V2I routing method based on adaptive variable beacon power, in order to improve the link stability of a vehicle connected with an RSU, the vehicle firstly collects neighbor beacons to analyze the current signal coverage state of the vehicle, the current beacon power of the vehicle is adaptively adjusted, and the vehicle is connected with the RSU; and then, analyzing available neighbor vehicle nodes through improved Q-learning, searching an optimal neighbor node and selecting the optimal neighbor node as a next-hop route, thereby completing V2V2I data routing. According to the method, the beacon power can be adaptively adjusted according to different conditions, the vehicle energy consumption is saved, and the method has good technical effects on the vehicle selfish number, the vehicle average connection RSU time, the routing hop count and the link retention time when the beacon power change range is + / -30% of the standard power.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a vehicle networking V2V2I routing method based on adaptive variable beacon power. Background Art

[0002] MEC has been extensively studied in the Internet of Vehicles in recent years because of its advantages of being close to users, having low transmission delay, and fast processing speed. Vehicles communicate with roadside units (RSUs) and unload data to nearby MECs for processing, which is called V2I communication. However, due to the limited signal coverage of vehicles, in scenarios where the vehicle and RSU are far apart, if the vehicle wants to communicate with the RSU, it must rely on the intermediate vehicle routing method to transmit data, which is the V2V2I communication method. In V2V2I communication, the most important thing is the V2V link selection problem. At present, domestic and foreign scholars have conducted a lot of research on V2V routing selection, among which how to select the next-hop vehicle as the routing node has become the main consideration. The main considerations currently include latency, routing hops, bandwidth, distance, vehicle selfishness, security, energy, and mobility.

[0003] Existing technical solutions include: 1. The vehicle beacon power is in a fixed state; 2. All vehicles are assumed to participate in vehicle routing work unconditionally; 3. The routing algorithm adopts the classic Q-learning algorithm.

[0004] However, the existing technical solutions have the following shortcomings: 1. It is not possible to flexibly adjust the beacon power according to the different environments of the vehicles. 2. In reality, it may not meet the situation where some personal reasons do not want to participate in routing. 3. Although the classic Q-learning can improve the link conditions of V2V to a certain extent, it cannot adapt well to the state of fast-moving vehicles. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a vehicle networking V2V2I routing method based on adaptive variable beacon power, which includes the following steps:

[0006] Step 1: Collect the beacon information of neighboring vehicles within one-hop signal range;

[0007] Step 2: Analyze whether there are available vehicles among the neighboring vehicles;

[0008] Step 3: If there is one or more available vehicles, go to step 5; if there is no available vehicle, go to step 4;

[0009] Step 4: Increase the beacon power according to the adaptive formula to collect available vehicles; if the beacon power of the vehicle is at the maximum value and the search for the vehicle times out, trigger event 1 and end the current stage of unloading work in advance.

[0010] Step 5: Analyze the states of the available vehicles and the current vehicle. If the states of the available vehicles and the current vehicle are not in state S4, it is necessary to dynamically adjust to state S4.

[0011] Step 6: Collect the beacon information of available vehicles, consider the selfishness of vehicles, vehicle mobility, and link stability, and determine the next hop route based on improved Q-learning;

[0012] Step 7: Unload data according to the selected route.

[0013] Preferably, the state of the vehicle includes:

[0014] State S1: The vehicle not connected to the RSU can receive the beacon information of the vehicle V1 connected to the RSU, but the vehicle V1 cannot receive the beacon of V2;

[0015] State S2: Vehicle V2 that is not connected to the RSU does not receive any beacons from neighboring vehicles and RSUs, while vehicle V1 that is connected to the RSU can receive the beacon from vehicle V2;

[0016] State S3: vehicle beacon signals do not overlap each other;

[0017] State S4: normal communication state.

[0018] Preferably, when the vehicle is continuously in an environment with few vehicles or a traffic jam, power saving event 1 and event 2 can be triggered, wherein:

[0019] Event 1: On a road with few vehicles, the vehicle is at the maximum beacon power value for a long time and cannot find an available vehicle. That is, when the vehicle detects that the number of neighbor lists is less than the threshold for a long time, Event 1 will be automatically triggered to reduce the vehicle beacon power to the minimum value, reducing unnecessary energy consumption;

[0020] Event 2: When a vehicle is in a congested road, a large number of vehicle broadcasts will cause congestion in the wireless channel. When the vehicle detects that the number of neighbor lists is greater than the threshold for a long time, it will automatically trigger Event 2, that is, reduce the beacon power while ensuring normal connection.

[0021] The present invention has the following beneficial effects:

[0022] In the existing V2V routing algorithms, there are certain solutions for the dynamic changes in topology caused by vehicle movement, but they do not comprehensively consider the vehicle beacon coverage status, the control of variable beacon power, and the selfishness of vehicles. In this regard, the present invention comprehensively considers the above-mentioned influencing factors, analyzes the coverage status of different beacon signals in detail, and realizes adaptive control of beacon power by combining on-demand control and trigger events. Finally, the optimal next hop is selected from the neighbor list through improved Q-learning to complete V2V2I data unloading. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a framework diagram of the V2V2I system of the present invention;

[0024] Figure 2 A schematic diagram of the beacon broadcast and data transmission timeline of the present invention;

[0025] Figure 3 A schematic diagram of the signal coverage status classification of vehicles in the present invention;

[0026] Figure 4 is the overall beacon power adaptive control flow chart of the present invention;

[0027] Figure 5 Graph showing simulation results of the method of the present invention on the HighD dataset. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0029] In the traditional vehicle-to-vehicle-to-road test unit (V2V2I) routing algorithm, the beacon power of the vehicle is in a fixed state and cannot be dynamically adjusted in combination with event-driven and on-demand working modes, resulting in poor dynamic adaptability of the vehicle and a large amount of energy waste. In order to improve the flexibility of beacon power adjustment of vehicles in dynamic environments, while considering realistic factors such as the signal coverage status, selfishness, and vehicle motion characteristics of routing vehicles, a V2V2I routing algorithm ABPR based on adaptive variable beacon power is proposed. Specifically, in order to improve the link stability of the vehicle connecting to the RSU, the vehicle first collects neighbor beacons to analyze the signal coverage status of the current vehicle, adaptively adjusts the current beacon power of the vehicle, and then analyzes the available neighbor vehicle nodes through improved Q-learning, finds the optimal neighbor node and selects it as the next hop route to complete the V2V2I data routing.

[0030] Figure 1 The V2V2I system framework is shown in Figure 1. The V2V2I system framework mainly consists of two parts: V2V and V2I. V2V is responsible for data transmission between vehicles, while V2I is responsible for data transmission between vehicles and RSU.

[0031] The vehicle networking V2V2I routing method based on adaptive variable beacon power of the present invention comprises the following steps:

[0032] Step 1: Collect the beacon information of neighboring vehicles within one-hop signal range;

[0033] Step 2: Analyze whether there are available vehicles among the neighboring vehicles;

[0034] Step 3: If there is one or more available vehicles, go to step 5; if there is no available vehicle, go to step 4;

[0035] Step 4: Increase the beacon power according to the adaptive formula to collect available vehicles; if the beacon power of the vehicle is at the maximum value and the search for the vehicle times out, trigger event 1 and end the current stage of unloading work in advance.

[0036] Step 5: Analyze the states of the available vehicles and the current vehicle. If the states of the available vehicles and the current vehicle are not in state S4, it is necessary to dynamically adjust to state S4.

[0037] Step 6: Collect the beacon information of available vehicles, consider the selfishness of vehicles, vehicle mobility, and link stability, and determine the next hop route based on improved Q-learning;

[0038] Step 7: Unload data according to the selected route.

[0039] The broadcast beacon parameters of the vehicle include not only the motion characteristic information of the current vehicle, but also the connection status of the vehicle and the status of the connected RSU, the neighbor vehicle list, selfishness parameters, etc. The broadcast beacon parameters of the vehicle are shown in Table 1.

[0040] Table 1 Vehicle broadcast beacon parameters

[0041]

[0042]

[0043] In addition to exchanging information with other vehicles through broadcast beacons, vehicles also need to exchange beacon information with RSUs. The broadcast beacon parameters of RSUs are shown in Table 2.

[0044] Table 2 RSU broadcast beacon parameters

[0045]

[0046] Specifically, the definition of variable beacon power is as follows. Compared with traditional fixed beacon power, variable beacon power has greater flexibility in vehicle energy consumption regulation and search range. When the vehicle is in a suburb with low vehicle density, the beacon power of the vehicle can be increased to search for more available vehicles. When the vehicle is on a congested road, there are a large number of available vehicles, and the beacon power can be reduced to reduce channel congestion and save energy. The increase and decrease of vehicle beacon power are both operated on the basic beacon power, and the formula for the beacon power variation range is defined as follows:

[0047] P adap =P base *△ad

[0048] △ad∈{N,1,2-N},(1>N>0)(1)

[0049] Where P adap Represents the range of variable power, P base Represents the basic beacon power, Δad represents the increment multiple, and Δad can only be one of the maximum value 2-N, the minimum value N, and 1. The value of N needs to be based on the hardware conditions of actual communication. When the increment is 1, it means maintaining the standard beacon power.

[0050] Since vehicles need to maintain a stable connection link during the data routing process, and the frequent changes in beacon power will cause unstable link connection, this paper adopts the maximum, minimum and standard beacon power values ​​for step-by-step changes to meet the vehicle beacon power value change requirements while reducing the impact of decreased link stability caused by frequent beacon power changes.

[0051] Specifically, regarding the vehicle signal coverage status under variable beacon power. If vehicles need to communicate with each other, it must be ensured that the signal coverage is within the signal coverage range of each other. Figure 3 As shown in (a), vehicle V2 that is not connected to the RSU can receive the beacon information of vehicle V1, but the beacon power of V2 is small and the signal range cannot cover V1, so normal wireless communication cannot be achieved.

[0052] Figure 3 The diagram is a schematic diagram of the signal coverage status classification of the vehicle in the present invention. Although it can be seen from the overall that vehicle V2 needs to increase the beacon power so that the signal can cover vehicle V1 and achieve normal data unloading. However, in actual vehicle driving, both vehicle V1 and vehicle V2 are independent vehicles, and there is no upper-level system to inform the vehicle of its current status. The vehicle can only infer the status from the received beacon information. There are four main status classifications of the vehicle as follows.

[0053] (1) State S1: Vehicles not connected to the RSU can receive beacon information from vehicle V1 that is connected to the RSU, but vehicle V1 cannot receive beacons from V2.

[0054] After receiving the beacon message sent by the neighboring vehicle, the vehicle will neber All information about neighbor vehicles is recorded in Figure 3 In state S1 shown in (a), for vehicle V2, V2 can receive the beacon from V1, which contains the neighbor list parameter L of V1. neber Therefore, vehicle V2 extracts the L of vehicle V1 neber After analysis, it is found that there is no vehicle V2 in the neighbor list of vehicle V1. Vehicle V2 can judge that it is currently in the S1 state with vehicle V1 and needs to actively increase the beacon power. When the beacon power of vehicle V2 increases to cover vehicle V1, when V2 receives the beacon of vehicle V1 again, vehicle V2 will be in its neighbor list, and finally achieve normal communication.

[0055] For vehicle V1, only the beacon information of RSU is received, and the neighbor vehicle parameter L in the beacon parameter of RSU car Contains V2 vehicles, which can achieve normal communication without adjusting the beacon power.

[0056] (2) State S2: Vehicle V2 that is not connected to the RSU does not receive any beacons from neighboring vehicles and RSUs, while vehicle V1 that is connected to the RSU can receive the beacon from vehicle V2.

[0057] like Figure 3 As shown in (b), for vehicle V1, although vehicle V1 can communicate with RSU normally, when receiving the beacon from vehicle V2, V2 is not connected to RSU and the neighbor list L neber Therefore, the V1 vehicle needs to actively expand the beacon power coverage until the signal range can cover the V2 vehicle and complete normal communication.

[0058] (3) State S3: Vehicle beacon signals do not overlap each other.

[0059] like Figure 3 As shown in (c), when the signals of both vehicles cannot cover each other, the beacon will not change actively because the V1 vehicle has established a connection with the RSU and there are no unconnected vehicles in the neighbor list. When the V2 vehicle has not established a connection with the RSU and the available vehicles in the neighbor list are empty, it will actively increase the beacon power to search for more vehicles. When the beacon power of V2 increases, if the signal can cover the V1 vehicle, the state will be transferred to the S2 state for resolution.

[0060] If the beacon power of vehicle V2 is at the maximum state for a period of time and no neighboring vehicle is detected, it will be in the energy-saving state with the minimum beacon power until the detection is restarted in the next control cycle.

[0061] (4) State S4: Normal communication state.

[0062] For the first three signal coverage states, the final Figure 3 The S4 state shown in (d) can complete normal V2V2I communication.

[0063] Figure 2 The figure is a schematic diagram of the beacon broadcast and data transmission timeline of the present invention. When a vehicle is driving on the road, the vehicle's beacon information acquisition and data transmission routing nodes have a certain timeliness, and the broadcast time of different vehicles is also different, so it is necessary to obtain beacon information and use it in time. Processing beacon information and data transmission are carried out in two independent parts in the vehicle's control unit.

[0064] In the independent task 1 of beacon broadcasting, the broadcast of the vehicle's own information, the beacon reception within one hop signal, and the update of the neighbor vehicle information table are completed within each broadcast cycle T.

[0065] In the independent task 2 of data transmission, the vehicle will perform adaptive beacon power control and route selection based on the information collected by the beacon broadcast in task 1, and finally complete the data transmission according to the next selected route until entering the next transmission cycle.

[0066] Specifically, the selfishness based on energy consumption is defined as follows: selfish nodes in the network tend to pursue individual interests and are unwilling to cooperate or share resources, which has a negative impact on the performance and effectiveness of the network. According to the definition of selfishness of nodes in the network, they can be divided into static and dynamic selfish nodes.

[16] , that is, when the remaining resources of the vehicle are sufficient, it will show general cooperation, but when the remaining resources are insufficient, it will switch to selfishness. As new energy vehicles gradually become the trend of the future, the remaining battery power of the vehicle directly determines the vehicle's range. When the vehicle battery power is too low, you can choose to prohibit the use of other power-consuming functions of the vehicle to increase the range. This paper uses the remaining battery power of the vehicle as the basis for judging the selfishness of the vehicle, as shown in formula (2).

[0067]

[0068] in, represents the remaining energy of the current vehicle at time i, Represents the energy threshold. When the remaining energy of the vehicle is lower than the threshold, the selfishness of the vehicle will be triggered, and the beacon power will be kept at the minimum value. The selfish vehicle will no longer actively participate in the routing work of the vehicle, and only retain the unloading function of its own vehicle until the remaining power of the vehicle is higher than the threshold. It will participate in the vehicle's data routing again.

[0069] On the other hand, the battery power of the vehicle will continue to decline with the driving time and data transmission time. There are many factors that affect the battery power, such as frequent acceleration and deceleration of the vehicle, too many applications turned on in the vehicle, and long-term high-power signal transmission. Some parameters will change with the usage habits of different users. In the Internet of Vehicles, communication is the most energy-consuming operation. This article mainly considers the impact of the beacon transmission power on energy consumption. In actual situations, the relationship between beacon transmission power and energy consumption will be more complicated because it also involves a series of other factors, such as the efficiency of the communication system, the signal propagation environment, and hardware performance. Therefore, in order to simplify the relationship model, this article classifies all other identical influencing factors into basic energy consumption. The simplified formula is defined as follows:

[0070]

[0071] in, is the total energy consumed in the ith cycle; P i is the signal transmission power in the i-th cycle; It is the basic energy consumption required for the vehicle to travel in the i-th cycle.

[0072] Figure 4 This is the overall beacon power adaptive control flow chart of the present invention. For vehicles that are not connected to the RSU, it is necessary to control the beacon power as needed to adjust the three different unconnected signal coverage states S1, S2 and S3, and their final states all need to be changed to the S4 state where the signals cover each other. For vehicles that have been connected to the RSU, in the face of some situations such as congestion or a lack of vehicles in the suburbs, it is necessary to control the beacon power separately according to different events.

[0073] Beacon power on-demand control: For vehicles that are not connected to the RSU and those that are already connected to the RSU, measures to increase the beacon power will be taken according to different situations. The power increase value is defined according to formula (1), and the beacon power is continuously increased until the maximum value. For vehicles that are not connected to the RSU, since there are no available vehicles or RSUs within the current signal range, it is necessary to actively increase the beacon power to search for available vehicles, which corresponds to solving the situation of vehicles that are not connected to the RSU in states S1 and S3. For vehicles that are already connected to the RSU, if it is detected that the neighboring vehicle is not connected to the RSU, it is necessary to passively increase the beacon power, which corresponds to solving the situation of the RSU connected in state S2.

[0074] Event-triggered beacon power control: In order to further reduce the energy consumption of beacon power, when the vehicle is continuously in an environment with few vehicles or a traffic jam, power saving events 1 and 2 can be triggered.

[0075] Event 1: On a road with few vehicles, the vehicle is at the maximum beacon power value for a long time and cannot find an available vehicle. That is, when the vehicle detects that the neighbor list L neber The number of In this state, event 1 will be automatically triggered, reducing the vehicle beacon to the minimum power value and reducing unnecessary energy consumption.

[0076] Event 2: When the vehicle is in a congested section, there may be a large number of available vehicles in the same signal range. At the same time, a large number of vehicle broadcasts will also cause congestion in the wireless channel. Therefore, when the vehicle detects the neighbor list L neber The number of long-term greater than the threshold In this state, event 2 will be automatically triggered, that is, reducing the beacon power while ensuring normal connection.

[0077] The following is a detailed description of the improved Q-learning routing algorithm. Standard Q-learning is a model-free method in reinforcement learning. The agent continuously explores the unknown environment, aims to maximize the cumulative reward, and finds the control strategy in repeated experiments. The learning process of the agent can be represented by a triple {S, A, R}, where S represents the state space, which contains all the states that the agent may transition to. A represents the action space, which contains the set of all actions that the agent may perform. The agent achieves state transition by taking actions. R represents the immediate reward, which is the corresponding positive reward or negative reward obtained by the agent when it transitions from the current state to the next state after performing an action.

[0078] Q-learning environment definition: In the Q-learning based V2V2I routing algorithm, the environment is the entire VANET scenario, where the agent is each unloaded data packet in the vehicle, each vehicle is a routing node, and the behavior of the vehicle forwarding the data packet represents an action, and when the data packet reaches the connected vehicle, it will receive the corresponding reward value.

[0079] Definition of Q-learning reward: When a vehicle performs a forwarding action, the reward value obtained is formally expressed as shown in formula (4).

[0080]

[0081] Among them, R i (m,k) represents the current vehicle V i Select Vehicle V m As the next hop routing node, the reward value obtained by reaching the kth RSU. mIf it is already connected to the nth RSU, it gets a reward value of 1, otherwise it is 0.

[0082] Q-table definition: In Q-learning, the learning process is maintained by a Q-table with state space as rows and action space as columns. In this paper, for routing selection with multiple candidate nodes, the action space is represented by candidate RSUs, and the state space is the neighboring vehicle nodes within the one-hop signal range. The state update formula is shown in Equation (5).

[0083]

[0084] Among them, Q i (m,k) represents the vehicle V i Select Vehicle V m As the Q value corresponding to the kth RSU; α represents the learning rate, α∈[0,1], when α is 0, it means that no new information is learned; β represents the discount rate, β∈[0,1], the discount rate determines whether the current agent pays more attention to previous experience or immediate interests; R i (m,k) represents immediate reward; Indicates vehicle V m The largest Q value of the kth RSU in the neighbor list is Indicates vehicle V m Neighbor list.

[0085] Learning rate based on link evaluation function: In traditional Q-learning, the learning rate α is a fixed value. However, in complex and changeable VANET scenarios, the fixed learning rate leads to poor algorithm adaptability. Therefore, in order to further improve the adaptability of the learning rate in different scenarios, the link dynamic evaluation function is used as the learning rate. The link dynamic evaluation function between vehicle i and vehicle m is defined as follows:

[0086]

[0087] in, is the weight factor, is the influencing factor of the vehicle motion characteristics between vehicle i and vehicle m. The closer the motion characteristics between vehicles are, the greater the possibility that the link will be maintained for a long time. is the existence time of vehicle m in the neighbor list of vehicle i. The longer the existence time, the stronger the reliability; is the remaining power factor of vehicle m. When the battery power is more sufficient, the selfishness of the vehicle is weaker, and the possibility of continued routing forwarding in the future is greater. If the remaining power of the vehicle is about to fall below the selfishness threshold, the learning rate can be reduced in advance to avoid selecting this vehicle. and The definition of is shown in equations (7) to (9):

[0088]

[0089] in, Indicates the absolute value of the speed difference between vehicle i and vehicle m; is the sum of the absolute values ​​of the speed differences of all vehicles in the neighbor vehicle list of vehicle i; represents the neighbor list of vehicle i; represents the remaining power of vehicle m, Represents the sum of the remaining power of all vehicles in the neighbor vehicle list of vehicle i.

[0090] Discount rate based on routing node length: In traditional Q-learning, the discount rate β is also a fixed value. According to the characteristics of V2V2I data routing, all vehicle unloading data needs to be routed to RSU. The more routing nodes there are for vehicle unloading data, the longer it takes and the more unstable it is. Therefore, this paper sets the discount rate as a dynamic parameter of the routing node length. The definition of the discount factor is shown in formula (10).

[0091]

[0092] in, It means that after vehicle i selects vehicle m as the next hop routing node, it reaches the total routing nodes of the RSU. It is the sum of all path nodes reaching RSU in the neighbor vehicle list of vehicle i.

[0093] Based on the improved Q-learning algorithm: In summary, the improved Q-learning algorithm is defined as shown in formula (11).

[0094]

[0095] The improved Q-learning takes into account the status of the vehicle in different situations and can better adapt to the fast-moving environment of VANET, as well as the changes of the vehicle's status over time.

[0096] In order to verify the reliability of the ABPR algorithm in a more realistic VANET scenario, the experimental data set uses the HighD data set of roads near Cologne, Germany, and compares it with the GPSR and RSAR algorithms in terms of the average total energy consumption of vehicle unloading, the number of selfish vehicles, the average proportion of time connected to RSU, the number of routing hops, and the link lifetime. The number of vehicles on the road is selected as 10, 20, 30, 40, 50, 60, 70, 80, and 90. At the same time, ABPR uses ±70%, ±50%, and ±30% of the standard power to verify the algorithm performance of ABPR under different beacon powers.

[0097] It should be noted that the longest single recorded road section in the HighD dataset is 420m, and there is no record of continuous vehicles between multiple sections. However, the experiment in this paper needs to record the continuous connection time of the vehicle. Therefore, in order to match the data in the dataset, the RSU interval is shortened from the conventional 2000 interval to 400m, and the standard beacon power coverage radius of the corresponding vehicle is scaled from 300m to 60m.

[0098] The specific experimental simulation parameters are shown in Table 3.

[0099] Table 3 Simulation parameter settings

[0100]

[0101] Figure 5 Graph showing simulation results of the method of the present invention on the HighD dataset.

[0102] Vehicle beacon power value: For the new energy vehicles that are developing continuously today, the vehicle's endurance has always been the focus of vehicle users. Therefore, under the condition of ensuring normal vehicle communication, the smaller the energy consumption of irrelevant vehicles, the longer the corresponding vehicle's endurance can be. Figure 5 As shown in (a), when the total number of vehicles on the road is greater than 60, the number of vehicles within the beacon range of a single vehicle continues to increase. When the event triggering threshold N is reached, the event 2 energy-saving mode will be triggered, making the beacon power of the current vehicle become the minimum setting value, thus achieving the purpose of saving vehicle energy consumption. The GPSR and RSAR algorithms used as comparative experiments will maintain a fixed beacon throughout the process and be in standard power consumption mode.

[0103] On the other hand, communication capability is also an important foundation for vehicles to realize intelligent driving. In order to improve communication capability, it is necessary to actively increase the beacon power value when there are fewer neighboring vehicles, and find enough available vehicles to help improve data offloading capabilities, such as Figure 5 (a) When the total number of vehicles is 10, the three different comparison algorithms of ABPR all become the maximum values ​​within their respective ranges, which is used to improve the communication capabilities of vehicles.

[0104] Comparison of the number of selfish vehicles: The definition of selfishness is based on realistic considerations. Figure 5As shown in (b), both GPSR and RSAR algorithms use fixed beacon power, which makes the number of selfish vehicles increase as the number of vehicles increases. However, after the number of vehicles exceeds 60, the average beacon power value of the ABPR algorithm decreases, the energy consumption of the vehicles decreases, and the number of selfish vehicles decreases. As the number of vehicles increases, more and more vehicles actively reduce the beacon power, and the number of selfish vehicles decreases. Since the GPSR and RSAR algorithms are always at the standard energy consumption value, and the initial power value of the vehicle is a normally distributed value, as the number of vehicles increases, more and more selfish vehicles are generated.

[0105] When the total number of vehicles is 10, the ABPR algorithm increases the beacon power to find available vehicles, while increasing the energy consumption of the vehicles, and the number of selfish vehicles increases. However, overall, the variable beacon power used by ABPR ranges from ±30%, ±50% to ±70%, and the overall selfish vehicles are reduced by 13.79%, 14.94%, and 17.24% compared to the standard beacon power.

[0106] Comparison of the proportion of time connected to RSU: Reducing the beacon power can save the vehicle's energy consumption to a certain extent, but it increases the risk of the vehicle's normal connection to RSU compared to the standard power. In order to increase the stability of the routing connection while reducing the beacon power, the ABPR algorithm adopts an improved Q-learning algorithm. The dynamic learning rate is based on the vehicle's motion characteristics, so that the vehicle selects a vehicle node with similar motion characteristics to the current vehicle as the next-hop routing node, so that it can maintain a relatively stable connection within the coverage of the lowest beacon power. The variable beacon power used by ABPR has increased by 0.63% and 0.2% from the fixed value of ±30% and ±50%, respectively. However, the ±70% parameter has decreased by 1.75%, which shows that blindly reducing the minimum beacon power value to save energy will have a counterproductive effect. Figure 5 As shown in (c), when the ABPR algorithm is at ±70% beacon power, the average proportion of time that vehicles are connected to the RSU is lower than the standard beacon power, regardless of whether the vehicles are sparse or congested.

[0107] The reason why the average proportion of time that all vehicles are connected to RSU does not reach 100% is that when a vehicle enters a new environment, there is a certain lag in the propagation of beacon information and the establishment of a link connection. Therefore, during the entire driving cycle of the vehicle, a part of the time needs to wait for the connection to be established. When the total number of vehicles is only 10, due to the small number of vehicles, the standard beacon power cannot cover enough vehicles, so the proportion of time connected to RSU is small.

[0108] Comparison of average routing hop count: The average data routing hop count refers to the number of forwarding vehicles that a data packet passes through from the unloading vehicle to the RSU. The larger the average routing hop count, the more unstable it is in the dynamic environment of VANET. Figure 5 As shown in (d), the GPSR routing algorithm uses a greedy solution and only selects the closest vehicle as the next hop route, resulting in the average number of vehicle hops increasing as the number of vehicles increases. The RSAR algorithm uses the traditional Q-learning algorithm to select the next route according to the maximum value in the Q-table and uses a fixed learning rate to continuously learn neighbor nodes. Compared with the GPSR algorithm, the average number of route hops is reduced to a certain extent.

[0109] The number of routes for the GPSR algorithm increases with the number of vehicles, but the number of routes for the RSAR algorithm increases first and then decreases. The reason is that there are fewer vehicles and fewer routing nodes at the beginning. As the total number of vehicles increases, the number of route hops increases and reaches a stable state when the number of vehicles is 40. As the number of vehicles continues to increase, the vehicles on the road become more and more congested, and the vehicle speed decreases. Most of the vehicles within the beacon range begin to become stable. When the RSAR algorithm selects the next hop route, some vehicles traveling in the opposite direction are excluded from consideration, thus reducing the redundant hops and the final average route hops begin to decrease.

[0110] However, compared with ABPR, the fixed learning rate of RSAR does not take into account the dynamic environment of VANET. In the experimental environment, vehicles are constantly moving, and the vehicles choose different RSUs to be unloaded. Therefore, ABPR introduces dynamic learning rate and discount rate, considers the motion characteristics and routing hops of vehicles, and dynamically calculates the learning rate and discount rate each time the Q-table is updated, so that the data packet selects the vehicle with more similar motion characteristics and a shorter routing hop number to reach the RSU as the next node, which reduces the average routing hop number as a whole. On the other hand, when the beacon power decreases, the number of routing hops will be further reduced due to the decrease in the number of optional vehicles. The three different beacon powers used by the ABPR algorithm, from ±30%, ±50% to ±70%, the average routing hop number decreases as the beacon power decreases, and the overall reduction is 55.34%, 56.59%, 61.10% and 21.82%, 24.01%, 31.91% compared with GPSR and RSAR, respectively.

[0111] Routing lifetime comparison: Routing lifetime refers to the link retention time. The longer the link retention time, the more stable the communication. Figure 5 As shown in (e), the average link holding time of the GPSR algorithm is only 24.6%, while that of the RSAR algorithm is 33.1%. The average holding time of the three different powers of ABPR from ±30%, ±50% to ±70% are 38.8%, 37.8% and 35.7% respectively.

[0112] The reason why the GPSR algorithm has a low link holding time is that it selects the next route only according to the greedy forwarding strategy, and the nearest vehicle may be traveling in the opposite direction of the current vehicle or may be accelerating. Therefore, it only selects the next hop node based on the closest distance.

[0113] The RSAR algorithm uses the traditional Q-learning algorithm to select the next-hop route. Since the traditional Q-learning algorithm searches for the next-hop based on the maximum value in the Q-table, and the vehicle's Q-table is stored in each vehicle, according to the Q-table maintenance rules, the longer the neighbor vehicle exists in the neighbor list, the more stable the vehicle is, the larger the value of the Q-table will be, the more stable the link selected as the next-hop vehicle will be, and the longer the link retention time will be. The ABPR algorithm has increased the overall link retention time by 57.65%, 53.71%, 45.1% and 17.18%, 14.25%, 7.85% compared to GPSR and RSAR, respectively, from ±30%, ±50% to ±70%.

[0114] Simulation experiments on the Cologne, Germany dataset show that the ABPR proposed in this paper can not only adaptively adjust the beacon power according to different situations and save vehicle energy consumption compared with the GPSR and RSAR routing algorithms, but also outperform the comparison algorithms in terms of the number of selfish vehicles, the average vehicle connection time to RSU, the number of routing hops and the link holding time when the beacon power variation range is ±30% of the standard power.

[0115] It should be noted that the above specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the protection scope of the present invention. Those skilled in the art should understand that the present invention description and its drawings are illustrative and do not constitute a limitation of the claims. The protection scope of the present invention is defined by the claims and their equivalents.

Claims

1. A vehicle networking V2V2I routing method based on adaptive variable beacon power, characterized in that: The following steps are involved: Step 1: Collect the beacon information of neighboring vehicles within one-hop signal range; Step 2: Analyze whether there are available vehicles among the neighboring vehicles; Step 3: If there is one or more available vehicles, go to step 5; if there is no available vehicle, go to step 4; Step 4: Increase the beacon power according to the adaptive formula to collect available vehicles; If the vehicle's beacon power is at the maximum value and the vehicle search timeout occurs, event 1 is triggered and the current stage of unloading work ends in advance; Step 5: Analyze the states of the available vehicles and the current vehicle; if the states of the available vehicles and the current vehicle are not state S4, dynamically adjust to state S4; Step 6: Collect the beacon information of available vehicles, and decide the next hop route based on the selfishness of the vehicle, the mobility of the vehicle, and the stability of the link and based on the improved Q-learning; Step 7: Unload data according to the selected route.

2. The method according to claim 1, characterized in that The vehicle status includes: State S1: The vehicle not connected to the RSU can receive the beacon information of the vehicle V1 connected to the RSU, but the vehicle V1 cannot receive the beacon of V2; State S2: Vehicle V2 that is not connected to the RSU does not receive any beacons from neighboring vehicles and RSUs, while vehicle V1 that is connected to the RSU can receive the beacon from vehicle V2; State S3: vehicle beacon signals do not overlap each other; State S4: normal communication state.

3. The method according to claim 2, characterized in that The beacon power variation range formula is defined as follows: P adap =P base *△ad △ad∈{N,1,2-N},(1>N>0) (1) Where P adap Represents the range of variable power, P base Represents the basic beacon power, Δad represents the increment multiple, and Δad can only be one of the maximum value 2-N, the minimum value N and 1. The value of N needs to be based on the hardware conditions of actual communication. When the increment is 1, it means maintaining the standard beacon power.

4. The method according to claim 3, characterized in that When the vehicle is continuously in an environment with few vehicles or in a traffic jam, power saving events 1 and 2 can be triggered, where: Event 1: On a road with few vehicles, the vehicle is at the maximum beacon power value for a long time and cannot find an available vehicle. That is, when the vehicle detects that the number of neighbor lists is less than the threshold for a long time, Event 1 will be automatically triggered to reduce the vehicle beacon power to the minimum value, reducing unnecessary energy consumption; Event 2: When a vehicle is in a congested road, a large number of vehicle broadcasts will cause congestion in the wireless channel. When the vehicle detects that the number of neighbor lists is greater than the threshold for a long time, it will automatically trigger Event 2, that is, reduce the beacon power while ensuring normal connection.

Images