Multi-layer intelligent agent simulation method and system for vehicle-road cooperative computing

Abstract

The invention relates to a multi-layer intelligent agent simulation method and system for vehicle-road cooperative computing. The method comprises the steps that S1, a basic data input module reads information data of a satellite map, road body facilities, road area intelligent equipment and vehicles; s2, a road scene construction module constructs a road scene; s3, the traffic operation management and control module constructs a traffic management and control model and continuously updates the management and control scheme; s4, performing calculation task distribution, processing and feedback by the vehicle-road cooperative calculation module; and S5, outputting traffic and communication evaluation indexes in various traffic scenes by a result visualization module. According to the method, intelligent road simulation scenes under external information interaction strategies such as different road basic environments, traffic flows and road areas-vehicles are constructed, road area intelligent equipment layout, traffic control and vehicle-road cooperative calculation strategies are optimized, and distribution, processing and feedback full-process simulation of calculation tasks between road area communication calculation units and intelligent vehicles is realized. And outputting traffic and communication evaluation indexes in various scenes.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a multi-layer intelligent agent simulation method and system for vehicle-road collaborative computing. Background Art

[0002] With the development of new technologies such as intelligent driving, Internet of Vehicles, and 5G communications, each vehicle needs to have a comprehensive and accurate grasp of information about the vehicle, road, and environment in real time, which will lead to a substantial increase in various computing tasks. Making full use of idle computing power of vehicles and road communication computing units to coordinate computing tasks can effectively improve the overall computing task processing capabilities of the system.

[0003] Existing research mainly focuses on the mobile communication level, focusing on the optimization of the algorithm for allocating computing tasks between vehicles and roads. Based on a roadside unit (RSU) equipped with a mobile edge computing (MEC) server, Guo et al. showed that the intelligent task allocation algorithm based on deep reinforcement learning can significantly reduce the computing burden of vehicles. On this basis, Shahidinejad et al. introduced a context-aware mechanism to improve the allocation effect of computing tasks. Liu et al. further considered the dependency constraints between different computing tasks and proved that the proposed optimization framework has lower system computing latency than the existing allocation method.

[0004] With the continuous development of Internet of Vehicles technology and its organic integration with MEC, Vehicular Edge Computing (VEC), a new paradigm for improving vehicle computing performance, has developed rapidly. As VEC nodes, constantly moving vehicles lead to rapid changes in the topology of the Internet of Vehicles. Compared with MEC, VEC puts forward new requirements for the selection of target computing nodes, multi-hop transmission task allocation, and opportunistic computing network construction. For example, for the node selection problem, Hamdi et al. constructed a service level agreement to ensure the selection of the best computing nodes, thereby achieving efficient utilization of idle computing resources of vehicles; for the multi-hop transmission problem, Liu et al. made full use of the computing resources of multi-hop vehicles on the basis of single-hop transmission task allocation to achieve a significant reduction in system latency; for the opportunistic computing problem, Rahman et al. considered the real-time status information such as vehicle speed, direction and position, and designed a context-aware opportunistic network construction scheme, which effectively reduced the task waiting time and improved the success rate of collaborative computing.

[0005] In summary, most current research aims to reduce system computing latency by optimizing the allocation of computing tasks and improving computing power utilization efficiency. Few studies focus on vehicle-road collaborative computing and design simulation systems to verify the variation patterns of vehicle-road collaborative computing latency in different scenarios. The present invention focuses on smart road facilities, vehicle operation control, vehicle-road communication computing and other elements involved in the vehicle-road collaborative system, and designs a multi-layer intelligent agent simulation model that includes an infrastructure layer, a traffic operation layer, and a collaborative computing layer. Summary of the invention

[0006] The purpose of this invention is to propose a multi-layer intelligent agent simulation method and system for vehicle-road collaborative computing, and to verify the variation law of vehicle-road collaborative computing delay under different scenarios.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A multi-layer intelligent agent simulation method for vehicle-road collaborative computing, comprising:

[0009] The S1 basic data input module reads satellite maps, road facilities, road-area intelligent devices, and vehicle information data;

[0010] S2 road scene construction module constructs the road scene;

[0011] The S3 traffic operation control module builds a traffic control model and continuously updates the control plan;

[0012] The S4 vehicle-road collaborative computing module distributes, processes and provides feedback on computing tasks;

[0013] The S5 result visualization module outputs traffic and communication evaluation indicators under various traffic scenarios.

[0014] A further improvement of the present invention is that in step S1, the steps performed by the basic data input module include: reading satellite maps, road facilities, road area intelligent devices, and vehicle information data; wherein:

[0015] The basic elements of road infrastructure include materials, alignment, signs and markings;

[0016] The road domain intelligent equipment includes perception, computing, communication and control units; the perception unit collects vehicle location and operation status information and transmits it to the traffic operation control module through the communication module;

[0017] The vehicle’s information includes computing power parameters, communication range, speed, driving direction, current location, and vehicle type.

[0018] A further improvement of the present invention is that the process of constructing the road scene by the road scene construction module in step S2 includes:

[0019] (1) Extracting road plane linear elements based on satellite maps and performing coordinate conversion to generate a linear road network topology, i.e., the basic layer of road infrastructure;

[0020] (2) Construct an optimized deployment model for road-domain communication computing units, update model parameters to meet different smart road service requirements, and generate a device layer based on point elements;

[0021] (3) Importing smart road facility information and equipment information into the database;

[0022] (4) A multi-layer road network and database together form a geographic information map to construct a road scene.

[0023] A further improvement of the present invention is that in step S3, the traffic control model constructed by the traffic operation control module includes: signal control, variable speed limit, and tidal lane; the process of generating a control plan includes:

[0024] (1) Setting the driving behavior parameters of the vehicle agent, including the parameters of the following behavior and the lane changing behavior, and setting the initial speed, acceleration and other parameters of different vehicle agents to follow a given distribution;

[0025] (2) Construct a traffic control model, including signal control, variable speed limits, and tidal lanes;

[0026] (3) Based on indicators such as vehicle travel delay and intersection queue length, feedback is provided to update the control plan and its parameters.

[0027] A further improvement of the present invention is that in step S4, the process of the vehicle-road collaborative computing module distributing, processing and feeding back computing tasks includes:

[0028] (1) The vehicle agent generates computing tasks of a certain size at a certain frequency and includes all vehicles and road communication computing units within its communication range into the candidate computing node set;

[0029] (2) Based on the given task allocation rules, select computing nodes from the candidate set and assign computing tasks;

[0030] (3) After the selected node completes the task processing, it sends the result back to the original vehicle.

[0031] A further improvement of the present invention is that in step S4, the vehicle-road collaborative computing module adopts a vehicle-road collaborative computing task allocation method based on a water injection method to select computing nodes from a candidate set and allocate computing tasks.

[0032] A further improvement of the present invention is that, in step S5, the content output by the result visualization output module includes:

[0033] (1) Output traffic evaluation indicators for various traffic scenarios, including vehicle travel delay, intersection queue length, and road service level;

[0034] (2) Output communication evaluation indicators under various traffic scenarios; in the S4 vehicle-road collaborative computing module, after the computing task is successfully assigned, it will go through four stages: sending the task to the target node, queuing the task before the target node processes it, computing the task, and sending the computing result back to the original node. The system calculates the delay T latency The calculation formula is as follows:

[0035] T latency＝ T request +T queuing +T process +T response

[0036] Among them, T request is the sending delay, T queuing is the waiting delay, T process To handle the delay, T response Return delay.

[0037] The present invention also provides a multi-layer intelligent agent simulation system for vehicle-road collaborative computing, which includes: a basic data input module, a road scene construction module, a traffic operation control module, a vehicle-road collaborative computing module and a result visualization module; and is used to execute the above-mentioned multi-layer intelligent agent simulation method for vehicle-road collaborative computing.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention provides a multi-layer intelligent agent simulation method and system for vehicle-road collaborative computing, constructs smart road simulation scenarios under different road basic environments, traffic flows, road-vehicle and other external information interaction strategies, optimizes road-domain intelligent equipment deployment, traffic control and vehicle-road collaborative computing strategies, realizes the full-process simulation of the allocation, processing and feedback of computing tasks between road-domain communication computing units and intelligent vehicles, outputs traffic and communication evaluation indicators under various scenarios, and can verify the variation law of vehicle-road collaborative computing delay under different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0041] Figure 2 The multi-layer intelligent agent simulation model architecture of the present invention;

[0042] Figure 3 The multi-layer intelligent agent simulation model process of the present invention;

[0043] Figure 4A schematic diagram of a road scene in a specific embodiment of the present invention;

[0044] Figure 5 This is a diagram of the vehicle-road collaborative calculation delay results in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0046] like Figure 1 As shown, a multi-layer intelligent agent simulation method for vehicle-road collaborative computing includes:

[0047] S1 basic data input module: read satellite maps, road facilities, road domain intelligent equipment, vehicles and other data. Road facilities include basic elements such as materials, line shapes, signs and markings. Road domain intelligent equipment includes perception, calculation, communication and control units. Vehicle information includes computing power parameters, communication range, speed, etc.

[0048] S2 Road Scene Construction Module: Extract road plane linear elements based on satellite maps and perform coordinate conversion to generate a linear road network topology, that is, the basic layer of road facilities. Aiming at different smart road service needs, optimize the layout of road domain communication computing units and generate a device layer based on point elements. Then import the smart road facility information and equipment information into the database to form a geographic information base map to realize the construction of road scenes;

[0049] S3 Traffic Operation Control Module: Constructs various traffic control models such as signal control, variable speed limit, tidal lanes, etc. Based on real-time perception of traffic operation status, with the goal of reducing vehicle travel delays, intersection queue lengths and other indicators, continuously updates control plans and improves road service levels;

[0050] S4 vehicle-road collaborative computing module: includes the entire collaborative computing process of computing task distribution, processing and feedback. After the vehicle generates a computing task, it establishes communication with neighboring vehicles and road communication computing units within a certain range, selects the best node to form a transmission link and allocates computing tasks according to given rules. After the selected node completes the task processing, it feeds back the computing results to the original vehicle;

[0051] S5 result visualization module: This module outputs traffic and communication evaluation indicators under various traffic scenarios;

[0052] This specific embodiment uses AnyLogic as a simulation platform to specifically illustrate the invention. Figure 2 The multi-layer intelligent agent simulation model architecture of the present invention is: Figure 3This is the multi-layer intelligent agent simulation model process of the present invention.

[0053] The specific process of step S1 is as follows:

[0054] Step 1.1: Read the satellite map and take the urban road section with a flat intersection as a typical specific example scenario. The length of each entrance direction section is 200m, including a left-turn lane and a straight-through lane. The right-turn lane adopts channelization diversion measures;

[0055] Step 1.2: Read the road infrastructure, including basic elements such as materials, alignment, signs and markings;

[0056] Step 1.3: Read the road domain intelligent devices, including perception, calculation, communication and control units. The perception unit collects vehicle location and operation status information and transmits it to the traffic operation control module through the communication module;

[0057] Step 1.4: Read vehicle information, including computing power parameters, communication range, speed, driving direction, current location, vehicle type, etc.

[0058] The specific process of step S2 is as follows:

[0059] Step 2.1: Based on the satellite base map, extract the key nodes and distance reference information of the road alignment, convert the Cartesian plane coordinates of each point in the simulation system according to the map scale, and draw the road alignment in the simulation system;

[0060] Step 2.2: In this specific embodiment, in order to weaken the impact of the layout of the road domain communication calculation unit on the vehicle-road collaborative calculation effect, only one set of road domain communication calculation units is introduced. At the same time, in order to ensure that the road domain unit has the same distance attenuation for vehicles in all directions, it is deployed in the center of the intersection. In addition, its coverage range is set to ensure that communication links can be established with all vehicles in the simulation instance scenario. The road domain communication calculation unit can fully cover the spatial range of this instance;

[0061] Step 2.3: According to the actual road plane information and pavement design, set the parameters such as lane direction, number, connectivity, and pavement material, complete the construction of the basic base map of road infrastructure, and import the smart road facility information (such as pavement performance parameters such as anti-skid) and equipment information (such as equipment performance and layout location parameters) into the database;

[0062] Step 2.4: The multi-layer road network and database together form a geographic information map to construct a road scene, such as Figure 4 As shown;

[0063] The specific process of step S3 is as follows:

[0064] Step 3.1: Set the driving behavior parameters of the vehicle agent. The target speed of the vehicle agent is 60km / h and the maximum acceleration is 1.8m / s 2 , the maximum deceleration is 4.2m / s 2 ;

[0065] Step 3.2: Set the traffic flow input of the minimum and maximum traffic volume scenarios to 80pcu / h and 800pcu / h (i.e., saturated traffic volume), respectively, and increase the input by 80pcu / h in each scenario. Other traffic flow parameters (such as the proportion of each turn on the entrance road, etc.) are all actual data obtained from the survey at the intersection.

[0066] Step 3.3: Design an intersection signal control scheme. This specific embodiment sets four phases: north-south straight, north-south left turn, east-west straight, and east-west left turn, and adopts a timing control strategy to calculate the signal control cycle, the green light duration of each phase, and other parameters according to the Webster timing method;

[0067] The specific process of step S4 is as follows:

[0068] Step 4.1: This specific embodiment sets the mean computing power of the vehicle agent to 400TOPS (trillion operations per second). Considering the differences in computing power among different vehicles in actual road traffic operation scenarios, different degrees of vehicle computing power heterogeneity are introduced in the simulation scenario design. Specifically, the absolute value of the difference between the extreme value and the mean of the vehicle computing power in the system is set to 0%, 25%, 50%, 75% and 100% of the mean computing power (i.e. 400), and obeys a triangular distribution. In addition, in the extreme scenario with a ratio of 100%, when a vehicle agent with a computing power of 0 appears, the computing power of the vehicle is set to 0.1TOPS to avoid the situation where the vehicle cannot handle computing tasks, thereby ensuring that all vehicles can complete the processing of computing tasks within a limited time.

[0069] Step 4.2: In this specific embodiment, the calculation task generation interval and the calculation task amount of the vehicle intelligent body are set to follow a symmetrical triangular distribution. The calculation task generation interval is μ~Δ(50,150), in milliseconds; the task calculation amount is τ~Δ(20,40), in trillion operations.

[0070] Step 4.3: In this specific embodiment, the computing power of the road communication computing unit is set to 0, 10, and 10 of the average computing power of the intelligent vehicle. 0 , 10 1 , 10 2 with 10 3 That is, the computing power of the road domain computing communication unit is 0, 400×10 0 , 400×10 1 , 400×10 2With 400×10 3 Among them, the scenario when the computing power of the road domain unit is 0 represents a pure Internet of Vehicles environment without the support of smart roads.

[0071] Step 4.4: This specific embodiment designs a vehicle-road collaborative computing task allocation method based on the water injection method, according to the maximum computing capacity c of the vehicle or road domain communication computing unit node v v , based on the total computing tasks that have been processed, find the threshold of computing tasks to be allocated, that is, the water filling line WL, and maximize the use of the total computing capacity of the system on the basis of balancing system performance and resource occupancy. The pseudo code is shown in Table 1.

[0072] As shown in Table 1: The input of the computational task allocation algorithm based on the water injection method includes: vehicle or road domain communication computing unit node v s , calculate the task size τ; vehicle communication coverage radius r v , the coverage radius of the road domain communication computing unit is r r ; The total amount of calculation tasks of the vehicle is f(t), the total amount of calculation tasks of other vehicles is g(t), and the total amount of calculation tasks of the road is h(t); its output is; the best calculation node

[0073] The computational task allocation algorithm based on the water injection method includes (1) searching for the best water injection line, (2) calculating the amount of tasks to be assigned, (3) selecting collaborative computing nodes, and (4) judging whether the task can be successfully sent and updating the amount of computational tasks.

[0074] (1) In the process of searching for the best water injection line, the water injection line WL is first initialized:

[0075] WL=min(f(t),g(t),h(t))+c v / 3

[0076] Define the reference value ρ for updating the water injection line t :

[0077] ρ t =(WL-f(t)) + +(WL-g(t)) + +(WL-h(t)) +

[0078] Continuously update the water injection line WL and the reference value ρ t , until |c v -ρ t |>ε; where: the expression for updating the water injection line WL is:

[0079] WL=WL+(c v -ρ t ) / 3

[0080] Update reference value ρ t The expression is:

[0081] ρ t =(WL-f(t)) + +(WL-g(t)) + +(WL-h(t)) +

[0082] (2) The stage of calculating the amount of tasks to be assigned specifically includes:

[0083] Calculate the amount of computing resources that can be allocated under the waterfill line c t , whose expression is:

[0084] c t =WL-(f(t)+g(t)+h(t))

[0085] Calculate the distribution ratio λ of computing tasks of other nodes, and its expression is:

[0086]

[0087] Calculate the computing task volume w of other nodes k , whose expression is:

[0088] w k =λ×τ

[0089] (3) The stage of selecting collaborative computing nodes specifically includes:

[0090] Define the candidate computing node set S, whose expression is:

[0091]

[0092] According to the node computing capacity c s , calculate the maximum task completion time t max , whose expression is:

[0093] t max =max{∑τ / c s :v s =S}

[0094] Initialize the compute nodes For the car v v , initialize the score:

[0095]

[0096] In v s ∈S∧v s ≠v v In this case, the following steps are executed in a loop:

[0097] Determine whether the calculation node is consistent with the direction of movement of the vehicle

[0098] If they match, the score is calculated based on the relative speed. Its expression is:

[0099]

[0100] Otherwise, the score is calculated based on the relative speed. Its expression is:

[0101]

[0102] Calculate scores based on task handling ability Its expression is:

[0103]

[0104] Compare current node scores and candidate node score

[0105] If it is less than, update the calculation node and its score, the expression is:

[0106]

[0107] (4) The stage of judging whether the task can be successfully sent and updating the calculation task amount specifically includes:

[0108] Determine whether the sending task fails, the expression is random(0,1)<p;

[0109] If the sending fails, the calculation task f(t) of the ego vehicle is updated, and the calculation nodes and scores are updated back to the ego vehicle; where: the expression for updating the calculation task f(t) of the ego vehicle is:

[0110] f(t)=f(t)+τ

[0111] The expression for calculating nodes and updating scores back to the vehicle is:

[0112]

[0113] If the task is sent successfully, there are two situations:

[0114] (4.1) If the computing node is other intelligent vehicles Then update the calculation task volume f(t) of the self-car and the calculation task volume g(t) of the other car, the expressions are:

[0115] f(t)=f(t)+τ-wk

[0116] g(t)=g(t)+w k

[0117] (4.2) If the computing node is a road domain communication computing unit node, then update the vehicle computing task f(t) and update the road domain communication computing unit computing task h(t), the expressions are:

[0118] f(t)=f(t)+τ-w k

[0119] h(t)=h(t)+w k

[0120] Table 1 Pseudo code of the computational task allocation algorithm based on water injection method

[0121]

[0122]

[0123] Step 4.5: After the selected node completes the task processing, it transmits the result back to the original vehicle; the specific process of step S5 is as follows:

[0124] Step 5.1: Output traffic evaluation indicators under various traffic scenarios. In this specific embodiment, the intersection service level basically reaches Class C;

[0125] Step 5.2: Output the communication evaluation indicators under various traffic scenarios. The results are as follows: Figure 5 As shown, Figure 5 FIG. 1 is a diagram showing the delay result of vehicle-road collaborative calculation in a specific embodiment of the present invention. Figure 5 (a) to 5(e) represent five different levels of vehicle computing power heterogeneity;

[0126] This specific embodiment illustrates that the present invention can realize the full-process simulation of the allocation, processing and feedback of computing tasks between the road communication computing unit and the intelligent vehicle, output traffic and communication evaluation indicators under various scenarios, and verify the variation law of vehicle-road collaborative computing delay under different scenarios.

[0127] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A multi-layer intelligent agent simulation method for vehicle-road collaborative computing, characterized in that include: The S1 basic data input module reads satellite maps, road facilities, road-area intelligent devices, and vehicle information data; S2 road scene construction module constructs the road scene; The S3 traffic operation control module builds a traffic control model and continuously updates the control plan; The S4 vehicle-road collaborative computing module distributes, processes and provides feedback on computing tasks; The S5 result visualization module outputs traffic and communication evaluation indicators under various traffic scenarios.

2. A multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: In step S1, the steps performed by the basic data input module include: reading satellite maps, road facilities, road area intelligent devices, and vehicle information data; wherein: The basic elements of road infrastructure include materials, alignment, signs and markings; The road domain intelligent equipment includes perception, computing, communication and control units; the perception unit collects vehicle location and operation status information and transmits it to the traffic operation control module through the communication module; The vehicle’s information includes computing power parameters, communication range, speed, driving direction, current location, and vehicle type.

3. The multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: The process of constructing the road scene by the road scene construction module in step S2 includes: (1) Extracting road plane linear elements based on satellite maps and performing coordinate conversion to generate a linear road network topology, i.e., the basic layer of road infrastructure; (2) Construct an optimized deployment model for road-domain communication computing units, update model parameters to meet different smart road service requirements, and generate a device layer based on point elements; (3) Importing smart road facility information and equipment information into the database; (4) A multi-layer road network and database together form a geographic information map to construct a road scene.

4. The multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: In step S3, the traffic control model constructed by the traffic operation control module includes: signal control, variable speed limit, and tidal lane; the process of generating a control plan includes: (1) Setting the driving behavior parameters of the vehicle agent, including the parameters of the following behavior and the lane-changing behavior; and setting the initial speed, acceleration and other parameters of different vehicle agents to follow a given distribution; (2) Construct a traffic control model, including signal control, variable speed limits, and tidal lanes; (3) Based on indicators such as vehicle travel delay and intersection queue length, feedback is provided to update the control plan and its parameters.

5. The multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: In step S4, the process of the vehicle-road collaborative computing module distributing, processing and providing feedback on computing tasks includes: (1) The vehicle agent generates computing tasks of a certain size at a certain frequency and includes all vehicles and road communication computing units within its communication range into the candidate computing node set; (2) Based on the given task allocation rules, select computing nodes from the candidate set and assign computing tasks; (3) After the selected node completes the task processing, it sends the result back to the original vehicle.

6. The multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: In step S4, the vehicle-road collaborative computing module adopts a vehicle-road collaborative computing task allocation method based on the water injection method to select computing nodes from the candidate set and allocate computing tasks.

7. The multi-layer intelligent agent simulation method for vehicle-road collaborative computing according to claim 1, characterized in that: In step S5, the content output by the result visualization output module includes: (1) Output traffic evaluation indicators for various traffic scenarios, including vehicle travel delay, intersection queue length, and road service level; (2) Output communication evaluation indicators under various traffic scenarios; in the S4 vehicle-road collaborative computing module, after the computing task is successfully assigned, it will go through four stages: sending the task to the target node, queuing the task before the target node processes it, computing the task, and sending the computing result back to the original node. The system calculates the delay T latency The calculation formula is as follows: T latency＝ T request +T queuing +T process +T response Among them, T request is the sending delay, T queuing is the waiting delay, T process To process the delay, T response Return delay.

8. A multi-layer intelligent agent simulation system for vehicle-road collaborative computing, characterized in that include: Basic data input module, road scene construction module, traffic operation control module, vehicle-road collaborative calculation module and result visualization module; A multi-layer intelligent agent simulation method for vehicle-road collaborative computing as described in any one of claims 1 to 7.

Images