Distributed traffic control system, method, and program

Abstract

A traffic distribution control system according to an aspect of the present disclosure includes a plurality of mobile bodies, and comprises: a state prediction unit that is configured to predict, using a state h'(tk) at a time tk of the plurality of mobile bodies, an optimal state h(tk+1) at a next time tk+1 that is expressed as a state simulation of the plurality of mobile bodies in a virtual space; and a control unit that is configured to control the plurality of mobile bodies such that a state h'(tk+1) at the time tk+1 of the plurality of mobile bodies becomes closer to the optimal state h(tk+1).

Description

[Technical Field]

[0001] The present invention relates to a technology for autonomous traffic control by a plurality of vehicles. [Background technology]

[0002] Conventionally, traffic control (which may also be referred to as traffic regulation) is performed using traffic lights installed at road intersections and other locations to allow vehicles to travel safely without collisions. However, such traffic control has the problem of causing chronic congestion, particularly in urban areas, due to frequent stops while waiting for traffic lights, temporary stops when merging or turning right or left, etc. In response to this problem, in recent years, studies have been conducted to provide feedback on traffic control optimized in a virtual space to real-world objects via digital twins, which simulate real-world objects in a virtual space (see, for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Digital twin computing white paper (ver. 2.0.0), 2019, Internet<URL:https: / / www.rd.ntt / dtc / DTC_Whitepaper_jp_2_0_0.pdf> Summary of the Invention [Problem to be solved by the invention]

[0004] As autonomous vehicles become more widespread in the future, it is expected that the traditional centralized control system using traffic lights will be replaced with a decentralized traffic control system that does not use traffic lights (signal-free traffic control). For this reason, it is anticipated that the above issues will be resolved by using a decentralized traffic control system via a digital twin.

[0005] The present disclosure has been made in consideration of the above points, and aims to provide a technology for realizing distributed traffic control without using traffic lights via a digital twin. [Means for solving the problem]

[0006] A traffic distribution control system according to one aspect of the present disclosure is a traffic distribution control system including a plurality of moving objects, the ... plurality of moving objects, and the traffic distribution control system including a plurality of moving objects. k In the state h'(t k ) to calculate the next time t k+1 The optimal state h(t k+1 ) and a state prediction unit configured to predict the state of the plurality of moving bodies at time t k+1 State h'(t k+1 ) is the optimal state h(t k+1 and a control unit configured to control the plurality of moving bodies to approach a target point. [Effects of the Invention]

[0007] The disclosed technology provides a technology for realizing distributed traffic control without using traffic lights by using a distributed traffic control system via a digital twin. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating a graph structure. [Figure 2] FIG. 1 is a diagram for explaining an overview of an embodiment. [Figure 3] FIG. 1 is a diagram for explaining an outline of inference. [Figure 4] FIG. 1 is a diagram for explaining an overview of learning. [Figure 5] FIG. 1 illustrates an example of a system configuration. [Figure 6] FIG. 10 is a diagram illustrating an example of a processing flow. [Figure 7] FIG. 10 is a diagram illustrating a first modification of the system configuration. [Figure 8]FIG. 10 is a diagram illustrating a second modification of the system configuration. [Figure 9] FIG. 2 illustrates an example of the hardware configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, an embodiment of the present invention (the present embodiment) will be described with reference to the drawings. The embodiment described below is merely an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

[0010] (Outline of the embodiment) An example of a graph structure for each vehicle included in the traffic distributed control system of this embodiment is shown in Figure 1. As shown in Figure 1, each vehicle included in this traffic distributed control system is represented by a graph structure in which vehicles are nodes and connections for wireless communication between neighboring vehicles are edges. In this traffic distributed control system, vehicles can move without colliding with each other.

[0011] In the example shown in Fig. 1, the nodes are vehicles, but the nodes are not limited to a specific object. The nodes may be moving bodies other than vehicles, such as aircraft, ships, robots, drones, etc. In the following description, the nodes are described as "vehicles."

[0012] In this embodiment, a digital twin of the real world, such as vehicles and roads, is constructed in a virtual space, and traffic is controlled by a feedback loop between the digital twin and real-world vehicles. More specifically, actual vehicle states (e.g., vehicle speed, position, etc.) are collected from the real world, and these states are used to determine the optimal state of each vehicle on the digital twin, and this optimal state (i.e., optimal speed, position, etc.) is fed back to the actual vehicles. By repeating this process, optimal traffic distribution control of vehicles in the real world is achieved.

[0013] An image of the above feedback loop is shown in Figure 2. As shown in Figure 2, the states of vehicles in the real world (speed, position, etc.) are collected, and in the digital twin, these states and wireless communications to prevent collisions with nearby vehicles are taken into consideration to predict the state of each vehicle (this may also be called a state update) and determine the optimal state for the next time. This optimal state is fed back to the vehicles in the real world, and each vehicle in the real world performs control (autonomous control) to approach this optimal state. This prevents collisions between each vehicle and other vehicles (for example, traffic accidents), and realizes traffic distribution control that meets desired objectives (for example, increasing speed, shortening travel time and transportation time to a destination, reducing acceleration and deceleration, etc.).

[0014] In this embodiment, speed and position are used as the state, but this is just an example. It is also possible to include some state other than speed and position, such as route, lane, steering direction, etc. Also, only one piece of information (for example, only speed or only position) may be used as the state, or multiple pieces of information (for example, both speed and position) may be used.

[0015] Here, the state prediction of each vehicle on the digital twin is realized by an ordinary differential equation (ODE) that expresses the state update dynamics. This ordinary differential equation can be expressed, for example, as in the following equation (1).

[0016] dh / dt=M(h,x,t,θ,G) =M1(h,x,t,θ,G)+M2(h,x,A,b,G) (1) However, the above formula (1) has a constraint that the vehicles must be kept at a certain distance or more.

[0017] Here, the meaning of each symbol is as follows:

[0018] M: Dynamics M1: (Sub)dynamics that updates the state of each vehicle M2: (Sub)dynamics representing the relationships (constraint relationships) between vehicles t: Time (discrete time) h=h(t):state x=x(t): External control input θ: parameter to be learned G=G(t):Graph A, b: constraint parameters The external control input x is information exchanged via wireless communication with nearby vehicles to prevent collisions with other vehicles, and is, for example, information indicating the vehicle's traveling position, destination, etc. Specific examples of the external control input include images from an onboard camera mounted on the vehicle, position information measured by a GPS (Global Positioning System) receiver, and sensor information measured by various other sensors. For example, if the number of vehicles is N and the number of external input devices mounted on the vehicles (e.g., onboard cameras, GPS receivers, various sensors, etc.) is L, the external control input x is expressed as an N × L matrix, with the (i, j) element representing information on the j-th external input device mounted on the i-th vehicle.

[0019] Time t represents a discrete time. In the following, the index of time t is represented by k, and the discrete time is represented by t k (k=0, ,K). In other words, the ordinary differential equation shown in the above equation (1) is a discretized ordinary differential equation. If a predetermined time width is Δt, then for each k=1, ,K, t k -t k-1 =Δt. Δt can be set to, for example, about 1 millisecond to 1 second.

[0020] Graph G is graph information with vehicles as nodes and connections for wireless communication between neighboring vehicles as edges. Because the connection relationships with neighboring vehicles can change from moment to moment as a vehicle moves, it is expressed as G = G(t).

[0021] As mentioned above, the state h is, for example, the speed or position. Because the state of each vehicle can change from moment to moment, it is expressed as h = h(t). For example, if the number of vehicles is N, the state h is expressed as an N-dimensional vector, with the i-th element representing the state of the i-th vehicle.

[0022] The constraint parameters A and b are parameters for expressing constraint conditions that require a certain distance between vehicles to be maintained. Examples of constraint conditions include inequality constraint conditions such as Ah+b≦0, equality constraint conditions such as Ah+b=0, or conditions that include both inequality constraint conditions and equality constraint conditions. Note that an inequality constraint condition such as Ah+b≦0 expresses a repulsive force that keeps the distance between vehicles at a certain distance or an attractive force that keeps the distance between vehicles within a certain distance, while an equality constraint condition such as Ah+b=0 expresses a consensus force that keeps the distance between vehicles at a certain distance. These repulsive forces and consensus forces prevent collisions between vehicles.

[0023] The learning parameter θ is a parameter representing a traffic control model. By appropriately learning this θ, an optimal state h is obtained according to the above formula (1), and as a result, traffic control is realized in which each vehicle does not collide with other vehicles and achieves predetermined objectives (such as increasing speed, shortening travel time or transportation time to a destination, and reducing acceleration and deceleration). This learning parameter θ is updated and learned using a commonly used method such as backpropagation (BP), assuming a machine learning model such as a neural network. In the following description, the machine learning model will be described as a neural network as an example, but the machine learning model usable in this embodiment is not limited to a neural network.

[0024] (Summary of reasoning) Figure 3 shows an overview of the inference process (i.e., when predicting the optimal state of each vehicle in real time). As shown in Figure 3, at each time t k At time t (t=1, ,K), the real-world vehicle states are collected and stored in the digital twin. k-1The predicted optimal state h=h(t k ) is fed back to the real-world vehicle. After wireless communication between the vehicles on the digital twin, the next optimal state h = h(t k+1 ) is predicted. Meanwhile, in the real world, the optimal state h = h(t k ) is controlled so that each vehicle approaches the optimal state. In this way, the optimal state is predicted on the digital twin and fed back to the real-world vehicle. k This is repeated (i.e., a feedback loop) at (t=1, ,K). This allows optimal traffic distribution control of each vehicle's state via the digital twin.

[0025] (Study Overview) Figure 4 shows an overview of the learning process (i.e., when learning the parameter θ to be learned). As shown in Figure 4, the parameter θ is learned by optimizing a pre-designed evaluation function using states collected from real-world vehicles. More specifically, the parameter θ to be learned is learned by using a recurrent neural network with a graph structure as a solver (ODE solver) for solving equation (1). For example, if the recurrent neural network with a graph structure is f, then h(t k+1 )=f(h(t k ),x(t k ), G(t k ),A,b;θ). In this case, for example, the learning target parameter θ can be updated and learned by using the backpropagation method or the like for a graph-structured recurrent neural network f. Furthermore, as an evaluation function, for example, if the state h is speed, then J(·) can be a function that returns the inverse of the average speed, and H can be a function that represents the observation model, so that J(H(h)) can be used. In this case, the learning target parameter θ is updated so as to minimize the evaluation function J(H(h)). Note that the observation model is a system that returns the state that is actually observed when an actual vehicle is controlled to approach a certain state.

[0026] (System configuration example) An example configuration of a traffic distribution control system in this embodiment is shown in Figure 5. As shown in Figure 5, the traffic distribution control system in this embodiment includes a digital twin system 10 and a plurality of vehicles 20. Furthermore, the digital twin system 10 and each vehicle 20 are communicatively connected via a communication network 30 including the Internet or the like.

[0027] The digital twin system 10 is a computer or computer system that constructs a digital twin that simulates the real world in a virtual space and executes the above-mentioned inference and learning on this digital twin. Here, the digital twin system 10 has an inference unit 110 that performs the above-mentioned inference (i.e., predicting the optimal state of each vehicle 20 on the digital twin), and a learning unit 120 that learns the learning target parameter θ. Note that the inference unit 110 and the learning unit 120 are realized, for example, by processing in which one or more programs contained in the digital twin system 10 are executed by a processor such as a CPU (Central Processing Unit).

[0028] Vehicles 20 are moving objects that move in the real world. Each vehicle 20 is equipped with a computer or computer system such as an on-board device. Here, each vehicle 20 has a control unit 200 that controls the state of the vehicle 20 (for example, by controlling the engine, brakes, etc.) so as to approach the optimal state fed back from the digital twin system 10. Note that the control unit 200 is realized, for example, by a process in which one or more programs possessed by the on-board device of the vehicle 20 are executed by a processor such as a CPU. The on-board device having one or more programs that realize the control unit 200 may be referred to, for example, as a control device.

[0029] Hereinafter, assuming that the number of vehicles is N, each vehicle 20 will be distinguished from the others by "vehicle 201," "vehicle 202," ..., "vehicle 20 N " etc. Also, vehicle 20 i (i=1, . . . , N) is referred to as "control unit 200 i " is written as ".

[0030] (Processing flow) An example of a processing flow of the traffic distribution control system in this embodiment will be described with reference to Fig. 6. Here, steps S101 to S106 in Fig. 6 are repeatedly executed until a predetermined termination condition is satisfied. An example of such a termination condition is when an instruction to terminate traffic distribution control is given.

[0031] First, at time t=t0, ,t K Steps S101 to S105 are repeatedly executed for a certain time t=t k Steps S101 to S105 relating to this will be described below.

[0032] The inference unit 110 of the digital twin system 10 calculates the vehicle speed and the vehicle speed. i (i=1, ,N) from time t k The state h(t k )=(h1(t k ),···,h N (t k )) are collected (S101). k The state of each vehicle 20 in (i.e., the state of each vehicle 20 in the real world) h(t k ) to h'=h'(t k )=(h1'(t k ),···,h N '(t k )) where h i '(t k ) is time t k Vehicle 20 i These states h'=h'(t k ) is stored, for example, in a storage area of ​​a storage device provided in the digital twin system 10.

[0033] Next, the inference unit 110 of the digital twin system 10 calculates the time t k-1 The predicted time t k The optimal state of ^h=^h(t k )=(^h1(t k ),···,^hN (t k )) for each vehicle 20 i (i=1, ,N) (S102). i (t k ) is time t k Vehicle 20 i It should be noted that the inference unit 110 may use, for example, ^h i (t k ) Only 20 vehicles i may be sent to

[0034] The control unit 200 of each vehicle 20 k The vehicle 20 is controlled so as to approach the optimum state (S103). i Control unit 200 i is the optimal state ^h i (t k ) so that the vehicle 20 i Controls the state of

[0035] However, when t=t0, the above steps S102 to S103 do not have to be executed, because there is no optimal state for t0 predicted at the immediately previous time.

[0036] The inference unit 110 of the digital twin system 10 performs wireless communication between each vehicle and nearby vehicles on the digital twin (S104). k External control input x=x(t k ) can be obtained. Note that these external control inputs x=x(t k ) is stored, for example, in a storage area of ​​a storage device provided in the digital twin system 10.

[0037] The inference unit 110 of the digital twin system 10 then calculates the real-world vehicle 20 i Using the state h' of (i=1,...,N) and the external control input x, the next time t k+1 The optimal state h=h(t k+1 )=(h1(t k+1 ),···,h N (t k+1The optimal state h = h(t k+1 ) to ^h=^h(t k+1 )=(^h1(t k+1 ),···,^h N (t k+1 )).

[0038] Time t=t0,...,t K After S101 to S105 have been repeatedly executed for f, when learning the learning parameter θ, the learning unit 120 of the digital twin system 10 learns the learning parameter θ (S106). That is, the learning unit 120 updates and learns the learning parameter θ so as to minimize the evaluation function J(H(h)), for example, by using an error backpropagation method or the like for the recurrent neural network f having a graph structure.

[0039] The learning parameter θ may be updated, for example, when a preset period (e.g., one day, one week, etc.) has elapsed. However, the learning parameter θ may also be updated when some other preset condition is satisfied.

[0040] (Variation) A modified example of the traffic distribution control system according to this embodiment will now be described.

[0041] Variation 1 In this embodiment, the digital twin system 10 has an inference unit 110 and a learning unit 120. However, for example, as shown in FIG. 7, each vehicle 20 may have an inference unit 110, and a learning device 40 having a learning unit 120 may exist. In this case, the vehicle 20 i Inference section 110 i is its own state h i ' and 20 other vehicles j (j≠i, j∈{1, ,N}) j ' and the optimal state h (or optimal state h) at the next time on the digital twin. iOn the other hand, the learning unit 120 of the learning device 40 learns the learning target parameter θ in the same manner as in S106 above. However, the information required for learning is the i is transmitted to the learning device 40.

[0042] In addition, each vehicle has 20 i 20 other vehicles j (j≠i, j∈{1, ,N}) to state h j When collecting the vehicle speed information ', direct wireless communication may be performed between the vehicles 20, or wireless communication may be performed via some kind of relay device.

[0043] Variation 2 In this embodiment, the digital twin system 10 has the inference unit 110 and the learning unit 120. However, for example, as shown in FIG. 8, each vehicle 20 may have the inference unit 110 and the learning unit 120, and the digital twin system 10 may not be necessary. In this case, as in the first modification, the vehicle 20 i Inference section 110 i is its own state h i ' and 20 other vehicles j (j≠i, j∈{1, ,N}) j ' and the optimal state h (or optimal state h) at the next time on the digital twin. i ) is inferred. On the other hand, vehicle 20 i Learning section 120 i learns the learning parameter θ in the same manner as in S106 above.

[0044] In addition, each vehicle has 20 i 20 other vehicles j (j≠i, j∈{1, ,N}) to state h j When collecting the vehicle speed information ', direct wireless communication may be performed between the vehicles 20, or wireless communication may be performed via some kind of relay device.

[0045] (Example of hardware configuration) 9 shows an example of the hardware configuration of a computer 500 that can realize the above-described digital twin system 10, the on-board device installed in the vehicle 20, and the learning device 40. As shown in Fig. 9, the computer 500 has an input device 501, a display device 502, an external I / F 503, a communication I / F 504, a RAM (Random Access Memory) 505, a ROM (Read Only Memory) 506, an auxiliary storage device 507, and a processor 508. Each of these pieces of hardware is connected to each other via a bus 509 so as to be able to communicate with each other.

[0046] The input device 501 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 502 is, for example, a display, a display panel, etc. Note that the computer 500 does not necessarily have to have at least one of the input device 501 and the display device 502, for example.

[0047] The external I / F 503 is an interface with an external device such as a recording medium 503a. The computer 500 can read from and write to the recording medium 503a via the external I / F 503. Examples of the recording medium 503a include a flexible disk, a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0048] The communication I / F 504 is an interface that enables the computer 500 to communicate with other devices and apparatuses. The RAM 505 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 506 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 507 is a storage device (storage device) such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory. The processor 508 is an arithmetic device such as a CPU or a GPU (Graphics Processing Unit).

[0049] 9 is an example and is not intended to be limiting. For example, the computer 500 may have multiple auxiliary storage devices 507 or multiple processors 508, may not have some of the hardware shown in the figure, or may have various hardware other than the hardware shown in the figure.

[0050] (summary) As described above, the traffic distribution control system in this embodiment collects the state of each moving body in the real world at each time, and uses these states to predict the optimal state of each moving body at the next time on the digital twin, and feeds this back to each moving body in the real world. This allows optimal traffic distribution control of moving bodies (signal-free traffic control), and as a result, autonomous traffic control can be realized that achieves predetermined goals, such as increasing speed and reducing travel and transportation times, without moving bodies colliding with each other.

[0051] The present invention is not limited to the above-described specifically disclosed embodiments, and various modifications, changes, and combinations with known technologies are possible without departing from the scope of the claims. [Explanation of symbols]

[0052] 10 Digital Twin System 20 vehicles 30 Communication Network 40 Learning Device 110 Reasoning Department 120 Learning Department 200 control section

Claims

1. A distributed traffic control system including a plurality of moving objects, The time t ｋ In the state h'(t ｋ ) to calculate the next time t ｋ＋１ The optimal state h(t ｋ＋１ a state predictor configured to predict the state of the The time t ｋ＋１ State h'(t ｋ＋１ ) is the optimal state h(t ｋ＋１ a control unit configured to control the plurality of moving objects to approach a target object; and A distributed traffic control system, wherein the optimal state h(t k+1 ) includes optimal speeds, optimal positions, optimal routes, optimal lanes, and optimal steering directions of the plurality of mobile bodies at time t k+1 .

2. The state prediction unit Among the plurality of moving bodies expressed as state simulations of the plurality of moving bodies, ｋ The optimal state h(t ｋ＋１ 2. The distributed traffic control system of claim 1, configured to predict a traffic flow.

3. The traffic distribution control system of claim 2, wherein the information transmitted and received between the neighboring mobile bodies is information for preventing collisions between the neighboring mobile bodies, and includes at least one of camera images, location information, and sensor information.

4. The state prediction unit The parameters of the traffic control model to achieve a predetermined objective are also used to calculate the optimal state h(t ｋ＋１ ) and The traffic distribution control system includes: The traffic distribution control system according to claim 1 , further comprising a learning unit configured to learn the parameters using an evaluation function according to the purpose.

5. The learning unit State h(t ｋ ) is input at least to the state h(t ｋ＋１ 5. The traffic distribution control system according to claim 4, wherein the traffic distribution control system is configured to output a parameter θ, and to learn the parameter that optimizes the evaluation function based on backpropagation of errors through a neural network having the parameter θ.

6. A distributed traffic control system involving multiple moving objects The time t ｋ In the state h'(t ｋ ) to calculate the next time t ｋ＋１ The optimal state h(t ｋ＋１ ) and a state prediction procedure for predicting the The time t ｋ＋１ State h'(t ｋ＋１ ) is the optimal state h(t ｋ＋１ a control procedure for controlling the plurality of moving bodies so that they approach each other; Run The method, wherein the optimal state h(t k+1 ) includes an optimal speed, an optimal position, an optimal route, an optimal lane, and an optimal steering direction of the plurality of moving objects at time t k+1 .

7. In a distributed traffic control system that includes multiple moving objects, The time t ｋ In the state h'(t ｋ ) to calculate the next time t ｋ＋１ The optimal state h(t ｋ＋１ ) and a state prediction procedure for predicting the The time t ｋ＋１ State h'(t ｋ＋１ ) is the optimal state h(t ｋ＋１ a control procedure for controlling the plurality of moving bodies so that they approach each other; Execute The optimal state h(t k+1 ) includes optimal speeds, optimal positions, optimal routes, optimal lanes, and optimal steering directions of the plurality of moving bodies at time t k+1 .

Images