A self-organizing path planning method and device for autonomous driving

Abstract

This application discloses a self-organizing path planning method and device for autonomous driving, which uses the limited traffic information obtained to provide a reasonable path planning method, effectively alleviates traffic congestion, improves traffic safety, and achieves an improvement in the overall performance of the road network based on independent vehicle decision-making.

Description

Technical Field

[0001] This application relates to, but is not limited to, intelligent transportation technology, and in particular to a self-organizing path planning method and device for autonomous driving. Background Art

[0002] The intelligent vehicle-road cooperative system adopts advanced wireless communication and new generation Internet technologies to realize all-round dynamic real-time information interaction between vehicles and roads, and carries out vehicle-road cooperative safety and active road control based on the collection and integration of dynamic traffic information in all time and space, fully realizing the effective coordination of people, vehicles and roads, ensuring traffic safety, improving traffic efficiency, and thus forming a safe, efficient and environmentally friendly road traffic system.

[0003] Autonomous driving based on a cooperative vehicle-infrastructure system is a brand-new technology approach. With the assistance of a cooperative vehicle-infrastructure system, the requirements for sensor accuracy and algorithm performance can be greatly reduced, providing a low-cost, high-precision, and real-time autonomous driving solution.

[0004] Planning a vehicle's path within a traffic network is fundamental to achieving autonomous driving in a collaborative vehicle-infrastructure environment. Proper path planning can effectively alleviate traffic congestion and improve traffic safety. However, achieving this goal is an urgent issue. Summary of the Invention

[0005] The present application provides a self-organizing path planning method and device for autonomous driving, which can provide a reasonable path planning method, effectively alleviate traffic congestion and improve traffic safety.

[0006] An embodiment of the present invention provides a self-organizing path planning method for autonomous driving, including:

[0007] Obtaining, at an intersection node, traffic status information for each sub-segment in a road network where the current intersection node is located; wherein the road network is a pre-set basic unit for planning the next local driving path starting from the current intersection node; each road segment in the road network is divided into a preset number of sub-segments; and the traffic status information reflects critical density and traffic density;

[0008] Encode the traffic status information of each sub-segment obtained respectively, and combine the encoding results of each sub-segment into path planning target information;

[0009] According to the path planning target information, the corresponding path is selected from the pre-trained candidate path set as the next local driving path of the vehicle starting from the current intersection node.

[0010] In an exemplary embodiment, the road network includes: the current intersection node, all next intersection nodes of the current intersection node, and the road sections connected to the current intersection node and all next intersection nodes of the current intersection node, except the road sections entering the current intersection node.

[0011] In an exemplary embodiment, the road network includes: the current intersection node, several preset intersection nodes near the current intersection node, and road sections connected to these intersection nodes, except the road section where the vehicle enters the current intersection node.

[0012] In an exemplary embodiment, the obtained traffic status information of each sub-road section is encoded respectively according to a preset encoding method.

[0013] In an exemplary embodiment, the preset encoding method includes a one-hot encoding method.

[0014] In an exemplary embodiment, encoding the obtained traffic status information of each sub-road section respectively according to a one-hot encoding method includes:

[0015] According to the critical density n c Divide the traffic density space into 10 intervals; c The range of kn is divided into 9 density intervals of the same size, corresponding to the 1st to 9th dimensions of the state vector. c ~∞ is divided into a density interval, corresponding to the 10th dimension of the state vector; where k is a preset constant;

[0016] According to the one-hot encoding method, a 10×1 vector is used to represent the traffic status information in each sub-segment, and each dimension corresponds to each interval of the traffic density space in turn.

[0017] An embodiment of the present application also provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute any of the above-mentioned self-organizing path planning methods for autonomous driving.

[0018] An embodiment of the present application further provides a device for self-organizing path planning for autonomous driving, including a memory and a processor, wherein the memory stores the following instructions that can be executed by the processor: for executing the steps of the method for self-organizing path planning for autonomous driving described in any of the above items.

[0019] The embodiment of the present application further provides a self-organizing path planning device for autonomous driving, comprising: an acquisition module, a processing module, and a path selection module; wherein,

[0020] The acquisition module is configured to acquire, at an intersection node, traffic status information of each sub-section in a road network where the current intersection node is located; wherein the road network is a pre-set basic unit for planning the next local driving path starting from the current intersection node; each road section in the road network is divided into a preset number of sub-sections; and the traffic status information reflects critical density and traffic density;

[0021] The processing module is used to encode the obtained traffic status information of each sub-segment respectively, and combine the encoding results of each sub-segment into path planning target information;

[0022] The path selection module is used to select a corresponding path from a pre-trained set of candidate paths according to the path planning target information as the next local driving path starting from the current intersection node.

[0023] In an exemplary embodiment, it also includes: a training module, which is used to encode the sample traffic status information of each sub-section in the road network where the current intersection node is located according to the sample traffic status information of each sub-section obtained at the intersection node according to a preset encoding method, and combine the sample encoding results of each sub-section into path planning target sample information; use the path planning target sample information as the input of a preset training model, use the candidate path corresponding to the path planning target sample information in the candidate path set as the output of the preset training model, and train the parameters of the preset training model to obtain a path selection model.

[0024] In an exemplary embodiment, the path selection module is configured to input the path planning target information into the path selection model, and the path selection model selects and outputs a path corresponding to the path planning target information from a set of candidate paths.

[0025] The self-organizing path planning method for autonomous driving provided in the embodiment of the present application utilizes the limited traffic information obtained to provide a reasonable path planning method, effectively alleviates traffic congestion, improves traffic safety, and achieves an improvement in the overall performance of the road network based on independent vehicle decision-making.

[0026] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings are used to provide a further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0028] Figure 1 This is a flow chart of a self-organizing path planning method for autonomous driving in an embodiment of the present application;

[0029] FIG2( a ) is a schematic diagram illustrating a first embodiment of a road network according to an embodiment of the present application;

[0030] FIG2( b ) is a schematic diagram illustrating a second embodiment of a road network according to an embodiment of the present application;

[0031] FIG2( c ) is a schematic diagram illustrating a third embodiment of a road network according to an embodiment of the present application;

[0032] FIG3( a ) is a schematic diagram showing how each road section in a road network is divided into sub-road sections in an embodiment of the present application;

[0033] FIG3( b ) is a schematic diagram of traffic status information in a road network according to an embodiment of the present application;

[0034] FIG3( c ) is a schematic diagram of using One-Hot coding to describe the traffic status information of each cell in a road network according to an embodiment of the present application;

[0035] Figure 4 This is a schematic diagram of the composition structure of a self-organizing path planning device for autonomous driving in an embodiment of the present application. DETAILED DESCRIPTION

[0036] To make the purpose, technical solutions and advantages of this application more clear, the embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other in any way.

[0037] In a typical configuration of the present application, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0038] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0039] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include non-transitory media such as modulated data signals and carrier waves.

[0040] The steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions. Also, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be performed in an order different from that shown here.

[0041] In the future, vehicles in the transportation system will have only limited information exchange capabilities, that is, they can only obtain limited traffic information. In this way, on the one hand, it is difficult for vehicles to obtain vehicle information within the global transportation network. On the other hand, direct interaction between vehicles is difficult to obtain absolute network communication guarantees.

[0042] Most related technologies plan vehicle routes within the framework of traffic flow distribution, hoping to achieve traffic balance. However, these technologies focus on static, global traffic flow routing, making it difficult to flexibly respond to changes in local traffic demand. Furthermore, the performance of the methods provided by these technologies relies on simplified traffic flow dynamics models, making it difficult to accurately characterize the traffic flow characteristics of real traffic systems. Furthermore, these technologies generally require global traffic information, which is difficult to meet given the limited information exchange capabilities of vehicles over a long period of time in the future.

[0043] The inventors of this application have conducted research on the current autonomous driving based on the vehicle-road cooperative system. On the one hand, given that vehicles can only obtain limited traffic information, they have proposed a self-organizing path planning mechanism for autonomous driving, which improves the overall performance of the road network by optimizing the local traffic status. On the other hand, given that direct interaction between vehicles is difficult to ensure, they have proposed a path planning method based on reinforcement learning under the guidance of the self-organizing path planning mechanism, and by designing a reward mechanism, they have achieved indirect interaction between vehicles under the premise of independent vehicle decision-making, ultimately reducing the average travel time of vehicles. In view of this, the embodiments of this application provide a self-organizing path planning method for autonomous driving, which uses the limited traffic information obtained to provide a reasonable path planning method, effectively alleviate traffic congestion, improve traffic safety, and improve the overall performance of the road network on the basis of independent vehicle decision-making.

[0044] Figure 1 This is a flow chart of a self-organizing path planning method for autonomous driving in an embodiment of the present application. Figure 1 As shown, the following steps are included:

[0045] Step 100: Obtain, at the intersection node, traffic status information of each sub-section in the road network where the current intersection node is located.

[0046] The road network is a pre-defined basic unit used to plan the next local driving route starting from the current intersection node. Each road segment in the road network is divided into a preset number of sub-segments. Traffic status information reflects critical density and traffic density.

[0047] In an exemplary embodiment, taking a vehicle for autonomous driving as an example, the road network may include at least: the current intersection node of the vehicle, all next intersection nodes of the current intersection node of the vehicle, and the road segments connected to the current intersection node of the vehicle, and the road segments connected to all next intersection nodes of the current intersection node, excluding the road segment where the vehicle enters the current intersection node. The demonstration diagram of the road network is as follows Figure 2(a) to Figure 2(c) shown.

[0048] As shown in Figure 2(a), taking intersection node 0 as an intersection node (i.e., the current intersection node) that the vehicle passes through as an example, the road network where intersection node 0 is located includes intersection node 0, the next intersection node 1 of intersection node 0, another next intersection node 2 of intersection node 0, and the next intersection node on the left side of intersection node 0 (not shown in Figure 2(a)), as well as the following road sections: section 1 (Route1) from intersection node 0 to intersection node 1, section 2 (Route2) from intersection node 0 to intersection node 2, sections from intersection node 0 to the intersection node on its left (not shown in Figure 2(a)), and sections from intersection node 1 to its upper, lower, left, and right directions. 2(a)). In FIG2(a), the vehicle enters intersection node 0 from above intersection node 0. Therefore, the road network does not include the road section from intersection node 0 to the intersection node above it.

[0049] As shown in Figure 2(b), assuming that the vehicle chooses to reach intersection node 2 (i.e., the current intersection node) from intersection node 0 via path 2, then the road network where intersection node 2 is located includes intersection node 2, the next intersection node 3 of intersection node 2, another next intersection node 4 of intersection node 2, and the next intersection node on the left side of intersection node 2 (not shown in Figure 2(b)), as well as the following road sections: section 1 (Route1) from intersection node 2 to intersection node 3, section 2 (Route2) from intersection node 2 to intersection node 4, section 3 from intersection node 2 to its left intersection node (not shown in Figure 2(b)), and section 4 from intersection node 3 to its upper and lower intersection nodes. Four road sections in the left and right directions (it should be noted that only one of the repeated road sections 1 needs to be taken), four road sections from the intersection node 4 to its upper, lower, left and right directions (it should be noted that only one of the repeated road sections 2 needs to be taken), and four road sections from the next intersection node on the left side of the intersection node 2 to its upper, lower, left and right directions (not shown in Figure 2(b)) (it should be noted that the road section on its right is repeated with the road section from the intersection node 2 to the intersection node on its left (not shown in Figure 2(b)), so only one of them needs to be taken). In Figure 2(b), the vehicle enters the intersection node 2 from above the intersection node 2. Therefore, the road network does not include the road section from the intersection node 2 to the intersection node above it.

[0050] As shown in Figure 2(c), assuming that the vehicle chooses to reach intersection node 3 (i.e., the current intersection node) from intersection node 2 via path 1, then the road network where intersection node 3 is located includes intersection node 3, the next intersection node 5 of intersection node 3, the other next intersection node 6 of intersection node 3, and the next next intersection node 7 above intersection node 3, as well as the following road sections: section 1 (Route1) from intersection node 3 to intersection node 5, section 2 (Route2) from intersection node 3 to intersection node 6, the section from intersection node 3 to intersection node 7, the four road sections from intersection node 5 to its upper, lower, left, and right directions (it should be noted that only one of the repeated road sections 1 needs to be taken), and the four road sections from intersection node 6 to its upper, lower, left, and right directions (it should be noted that only one of the repeated road sections 2 needs to be taken). In Figure 2(c), the vehicle enters intersection node 3 from the left side of intersection node 3. Therefore, the road network does not include the section from intersection node 3 to the intersection node on its left.

[0051] It should be noted that Figure 2(a) to Figure 2(c) The embodiment shown is merely an embodiment of a road network setting and is not intended to limit the scope of protection of this application.

[0052] In another exemplary embodiment, taking a vehicle for autonomous driving as an example, the road network may include: the intersection node where the vehicle is currently located, several preset intersection nodes near the intersection node where the vehicle is currently located, and the road sections connected to these intersection nodes, excluding the road section where the vehicle enters the intersection node where it is currently located. Figure 2(a) to Figure 2(c) For example, the road network can be set to be composed of several nodes to the right and lower right of the intersection node where the vehicle is currently located, as roughly shown in the shaded part of Figure 2(a). In this embodiment, the road network where intersection node 0 is located includes intersection node 0, the next intersection node 1 of intersection node 0, the other next intersection node 2 of intersection node 0 and intersection node 01, as well as the following road sections: section 1 (Route1) from intersection node 0 to intersection node 1, section 2 (Route2) from intersection node 0 to intersection node 2, and section 3 (Route4) from intersection node 0 to the intersection node on its left (not shown in Figure 2(a)). The road network includes the following sections: the road section from intersection node 1 to its upper, lower, left, and right directions (shown in Figure 2(a)), the four road sections from intersection node 1 to its upper, lower, left, and right directions (it should be noted that only one of the road sections 1 is repeated), the four road sections from intersection node 2 to its upper, lower, left, and right directions (it should be noted that only one of the road sections 2 is repeated), and the four road sections from intersection node 01 to its upper, lower, left, and right directions (it should be noted that only one of the road sections above and to the left is repeated). In Figure 2(a), the vehicle enters intersection node 0 from above, so the road network does not include the section from intersection node 0 to the intersection node above it. The road networks of Figures 2(b) and 2(c) are roughly shown as the shaded parts in the figures and will not be described in detail here.

[0053] From the above settings about the road network, it can be seen that the road network can be pre-set according to different needs based on the actual scenario. The specific setting scale (that is, the number of intersection nodes and road sections included in the road network) can be pre-set according to the actual scenario. In other words, the setting of the road network size can be large or small, as long as the vehicles in the road network can obtain the traffic information in the current road network in real time and the road network can be regarded as the basic unit of the global road network.

[0054] In one exemplary embodiment, each road segment in a road network is divided into a preset number of sub-segments, also known as cells. Figure 3(a) shows a portion of a road segment in a road network, which is divided into multiple cells of a predetermined length. The number of cells on each road segment is a fixed number. As shown in Figure 3(a), a road segment or a portion of a road segment is divided into sub-segment 1 (Cell 1) and sub-segment 2.

[0055] Because each road segment contains the same number of cells, it is understandable that if the lengths of different road segments are different, the lengths of the cells divided from different road segments may also be different. For example, if road segment 1 is 2 kilometers (km) and road segment 2 is 6 kilometers, and the number of pre-set sub-segments is 5, then road segment 1 is divided into 5 cells, and the length of each cell can be 2 kilometers divided by 5, which is equal to 0.4 kilometers. Road segment 2 is divided into 5 cells, and the length of each cell can be 6 kilometers divided by 5, which is equal to 1.2 kilometers.

[0056] By analyzing traffic flow dynamics, the inventors of this application have determined that critical density is a key parameter in traffic dynamics. Critical density represents the traffic density at which the vehicle flow on a road segment reaches its maximum. In one exemplary embodiment, the traffic state information in the present embodiment reflects both critical density and traffic density. As shown in Figure 3(b), the abscissa represents traffic density, and the ordinate represents vehicle flow. The point on the abscissa corresponding to the maximum vehicle flow is the critical density. The relationship between traffic density and critical density has a significant impact on traffic efficiency. Therefore, in the present embodiment, critical density information is used as part of the traffic state information to accurately characterize traffic characteristics at the current moment. In the present embodiment, the current traffic state within each cell in the road network (i.e., the local road network) at the intersection where the vehicle is currently located is used as the basis for planning the next local route. Compared to methods that directly consider the position of each vehicle, the cell-based approach in the present embodiment reduces computational complexity and improves the fault tolerance of route estimation. Furthermore, the cell-based approach in the present embodiment more accurately describes the current traffic state in the road network, ensuring high-quality route planning.

[0057] Step 101: Encode the acquired traffic status information of each sub-road segment respectively, and combine the encoding results of each sub-road segment into path planning target information.

[0058] In one exemplary embodiment, the traffic status information obtained for each sub-segment can be encoded using a pre-set encoding method. In one embodiment, the pre-set encoding method includes, but is not limited to, one-hot encoding. One-hot encoding is a process that converts categorical variables into a form that is easily usable by machine learning algorithms and can be used to reduce the size of the state space.

[0059] In one embodiment, in step 101, one-hot encoding technology may be used to encode the traffic status information of each sub-road segment, including:

[0060] According to the critical density n c Divide the traffic density space into 10 intervals. c The range of kn is divided into 9 density intervals of the same size, corresponding to the 1st to 9th dimensions of the state vector. c~∞ is divided into a density interval separately, corresponding to the 10th dimension of the state vector; where k is a preset constant, such as k=1.5, k=2, k=3, etc.

[0061] According to One-Hot encoding, a 10×1 vector is used to represent the traffic status information in each sub-segment, and each dimension corresponds to each interval of the traffic density space.

[0062] Figure 3(c) shows an example of using one-hot encoding to describe the traffic status information of each cell. As shown in Figure 3(c), one-hot encoding technology is used to represent the traffic status information in each cell, that is, the traffic density information in this embodiment. That is, if a dimension of the 10×1 vector is set to 1, it means that the traffic density in the cell at the current moment is in the interval of the traffic density space corresponding to the dimension. In this embodiment, 2n c is the upper bound, where n c Indicates critical density; 0 to 2n c Divide into 9 density intervals of equal size, and divide 2n c ∞ is divided into a density interval, corresponding to the 10th dimension of the state vector. As shown in Figure 3(c), the one-hot encoding information of sub-segment 1 (Cell 1) is: 0000100000, that is, the traffic density in Cell 1 at the current moment is in the upper limit of the traffic density space n c interval.

[0063] The traffic density space interval division mechanism in the embodiment of the present application fully considers the critical density information. In one embodiment, the traffic density space interval division can be dynamically adjusted according to factors such as traffic demand, cell length, speed limit, etc. For example, the length of the density interval can be determined according to the actual traffic demand, and it is not necessarily based on the 2n in the above embodiment. c This interval division mechanism can flexibly control the scale of the state space and fully consider the characteristics of traffic dynamics.

[0064] In an exemplary embodiment, combining the encoding results of each sub-segment into path planning target information in step 101 includes:

[0065] According to the pre-set splicing order of the sub-road sections, the encoding results of the sub-road sections obtained after encoding are concatenated to form the path planning target information.

[0066] After the road network is set up, the intersection nodes and corresponding road sections included in the road network are clear, and the sequence of splicing the road sections can be set in advance. For different road networks (i.e., local road networks), the sequence of splicing the road sections is set in the same order. For example, for intersection node A, assuming that the road network only includes the section A1 below intersection node A and the section A2 to the right of intersection node A, section A1 includes Cell A11, Cell A12, Cell A13, and section A2 includes Cell A21, Cell A22, and Cell A23. Assuming that the pre-set sequence of sections is the lower section first and then the right section, then the encoding results of each sub-section obtained after encoding can be concatenated as follows: Cell A11 encoding result - Cell A12 encoding result - Cell A13 encoding result - Cell A21 encoding result - Cell A22 encoding result - Cell A23 encoding result. The connector "-" here does not exist in the actual concatenation, and it only indicates a sequence. Afterwards, assuming that intersection node B is reached, for intersection node B, assuming that the road network only includes the section B1 below intersection node B, and section B1 includes Cell Cell B11, Cell B12, Cell B13, the encoding results of each sub-segment obtained after encoding are concatenated together to be: the encoding result of Cell B11 - the encoding result of Cell B12 - the encoding result of Cell AB3 - 0-0-0. Here, because for the intersection node B, there is no road section to the right of the intersection node B in the road network, the encoding results of each sub-segment in the section are all replaced by 0.

[0067] Step 102: Select a corresponding path from a pre-trained set of candidate paths according to the path planning target information as the next local driving path starting from the current intersection node.

[0068] In an exemplary embodiment, taking a vehicle for autonomous driving as an example, this step may also include:

[0069] According to the sample traffic status information of each sub-section in the road network where the vehicle is currently located at the intersection node, the sample traffic status information of each sub-section is encoded according to a preset encoding method, and the sample encoding results of each sub-section are combined into path planning target sample information; the path planning target sample information is used as the input of a preset training model, and the candidate path corresponding to the path planning target sample information in the candidate path set is used as the output of the preset training model, and the parameters of the preset training model are trained to obtain a path selection model.

[0070] The training model can adopt deep Q learning, etc. The choice of training model is not used to limit the scope of protection of this application. The embodiment of the present application emphasizes that in the process of training the training model, the design of the road network, each sub-road section, and traffic status information, as well as the encoding of the traffic status information of each sub-road section, etc. all adopt the methods of steps 100 and 101 in the self-organizing path planning method provided by the embodiment of the present application. The traffic status information in the embodiment of the present application reflects the critical density and traffic density, and directly reflects the traffic flow dynamics characteristics compared to directly using the number of vehicles in each cell as traffic status information, thereby improving the learning effect of training the set of candidate paths, and the state space will not be too large, thereby improving the learning efficiency of training the set of candidate paths.

[0071] In an exemplary embodiment, selecting a corresponding path from a pre-trained set of candidate paths according to the path planning target information in step 102 may include:

[0072] The path planning target information is input into a pre-trained path selection model, and the path selection model selects the path corresponding to the path planning target information from the set of candidate paths and outputs it.

[0073] In one embodiment, the correspondence between the route planning target information and each path in the set of candidate paths may be pre-set based on historical data and traffic conditions as shown by the historical data. In this way, step 102 can directly select the path corresponding to the route planning target information based on the correspondence.

[0074] The traffic system has strong randomness, time-varying and scalability, which makes it difficult for vehicles to accurately predict the traffic status of the entire traffic system in the future. In this case, if the traffic status is predicted according to relevant technical basis, the vehicle directly determines its path to the destination at the starting point at one time, which often results in poor path planning results. Moreover, directly solving the path selection problem of vehicles in a large-scale road network will result in extremely high computational complexity. In addition, the current information exchange capabilities of vehicles are limited, and vehicles can only obtain local traffic information, and direct interaction between vehicles cannot be absolutely guaranteed by the communication network. Therefore, the self-organizing path planning method for autonomous driving provided in the embodiment of the present application is a self-organizing vehicle path dynamic planning mechanism. After the vehicle determines the starting point (origin)-destination (destination), it performs a local path planning according to the self-organizing path planning method provided in the embodiment of the present application at each intersection node during driving until the vehicle reaches the destination. The embodiment of the present application achieves high efficiency in reaching the destination by solving the path planning problems of several road networks (i.e., basic units) in steps. The advantages of the self-organizing path planning mechanism provided by the application embodiment include but are not limited to: each time a vehicle arrives at an intersection, it will use the latest traffic status information to dynamically solve the path planning problem within a limited time and space area, effectively reducing the negative impact of the time-varying and random nature of the traffic system on the path planning performance; compared with directly solving the path planning problem of the vehicle in the entire road network, the self-organizing mechanism effectively reduces the scale of the problem and reduces the requirements for information completeness; the self-organizing mechanism is not sensitive to the size of the road network and can adapt to traffic networks of any size and any number of traffic trip (OD, Origin-Destination) pairs.

[0075] It should be noted that for ease of description, this article only uses autonomous driving vehicles as an example, but this does not limit the scope of protection of this application. The self-organizing path planning method for autonomous driving provided in the embodiments of this application is applicable to mobile objects for autonomous driving, such as the above-mentioned vehicles, and may also include autonomous driving trains, aircraft, etc.

[0076] The present application also provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute any one of the self-organizing path planning methods for autonomous driving.

[0077] The present application further provides a device for self-organizing path planning for autonomous driving, comprising a memory and a processor, wherein the memory stores the following instructions that can be executed by the processor: for executing the steps of the method for self-organizing path planning for autonomous driving as described in any of the above items.

[0078] Figure 4This is a schematic diagram of the structure of a self-organizing path planning device for autonomous driving in an embodiment of the present application. Figure 4 As shown, it at least includes: an acquisition module, a processing module, and a path selection module; wherein,

[0079] An acquisition module is used to obtain, at an intersection node, traffic status information for each sub-segment in the road network where the current intersection node is located. The road network is a pre-set basic unit used to plan the next local driving path starting from the current intersection node. Each road segment in the road network is divided into a preset number of sub-segments. The traffic status information reflects critical density and traffic density.

[0080] a processing module, configured to encode the obtained traffic status information of each sub-road section respectively, and combine the encoding results of each sub-road section into path planning target information;

[0081] The path selection module is used to select a corresponding path from a pre-trained set of candidate paths according to the path planning target information as the next local driving path starting from the current intersection node.

[0082] In an exemplary instance, the self-organizing path planning device for autonomous driving provided by an embodiment of the present application may also include: a training module, which is used to encode the sample traffic status information of each sub-section in the road network where the vehicle is currently located at the intersection node according to a preset encoding method, and combine the sample encoding results of each sub-section into path planning target sample information; use the path planning target sample information as the input of a preset training model, use the candidate path corresponding to the path planning target sample information in the candidate path set as the output of the preset training model, and train the parameters of the preset training model to obtain a path selection model.

[0083] In an exemplary embodiment, the path selection module may be configured to: input the path planning target information into a pre-trained path selection model; the path selection model selects and outputs a path corresponding to the path planning target information from a set of candidate paths.

[0084] In an exemplary embodiment, the processing module may use one-hot encoding technology to encode the traffic status information of each sub-segment. One-hot encoding is a process of converting categorical variables into a form that is easy to use by machine learning algorithms, which can be used to reduce the size of the state space.

[0085] In an exemplary embodiment, taking a vehicle for autonomous driving as an example, the road network in the acquisition module may include at least: the intersection node where the vehicle is currently located, all next intersection nodes of the vehicle's current intersection node, and the road section connected to the vehicle's current intersection node, and the road section connected to all next intersection nodes of the current intersection node, excluding the road section where the vehicle enters the current intersection node.

[0086] In an exemplary embodiment, taking a vehicle for autonomous driving as an example, the road network in the acquisition module may include at least: the intersection node where the vehicle is currently located, several preset intersection nodes near the intersection node where the vehicle is currently located, and the road sections connected to these intersection nodes, except for the road section where the vehicle enters the intersection node where it is currently located.

[0087] In the embodiment of the present application, critical density information is used as part of the traffic status information to accurately characterize the traffic characteristics at the current moment. In the embodiment of the present application, the current traffic status of each cell in the road network (i.e., the local road network) at the intersection where the vehicle is currently located is used as the basis for planning the next local path. On the one hand, compared with the method of directly considering the position of each vehicle, the cell-based method in the embodiment of the present application reduces the amount of calculation and improves the fault tolerance of the path estimation. On the other hand, the cell-based method in the embodiment of the present application more accurately describes the current traffic status in the road network, providing a guarantee for the high quality of path planning.

[0088] The present application also provides an autonomous driving vehicle, including any of the aforementioned self-organizing path planning devices for autonomous driving. The autonomous driving vehicle provided in the present application provides a reasonable path planning method, effectively alleviating traffic congestion and improving traffic safety.

Claims

1. A self-organizing path planning method for autonomous driving, characterized in that: include: Obtaining, at an intersection node, traffic status information for each sub-segment in a road network where the current intersection node is located; wherein the road network is a pre-set basic unit for planning the next local driving path starting from the current intersection node; each road segment in the road network is divided into a preset number of sub-segments; and the traffic status information reflects critical density and traffic density; The traffic status information of each sub-segment obtained is encoded respectively according to the one-hot encoding method, and the encoding results of each sub-segment are combined into the path planning target information; wherein, the traffic status information of each sub-segment obtained is encoded respectively according to the one-hot encoding method, including: according to the critical density Divide the traffic density space into 10 intervals; The range of is divided into 9 density intervals of the same size, corresponding to the 1st to 9th dimensions of the state vector. The traffic state information of each sub-segment is represented by a 10×1 vector according to the one-hot encoding method, with each dimension corresponding to each interval of the traffic density space. According to the path planning target information, the corresponding path is selected from the pre-trained candidate path set as the next local driving path of the vehicle starting from the current intersection node.

2. The self-organizing path planning method according to claim 1, wherein: The road network includes: the current intersection node, all next intersection nodes of the current intersection node, and the road sections connected to the current intersection node and all next intersection nodes of the current intersection node, excluding the road sections leading to the current intersection node.

3. The self-organizing path planning method according to claim 1, wherein: The road network includes: the current intersection node, several preset intersection nodes near the current intersection node, and road sections connected to these intersection nodes, excluding the road section leading to the current intersection node.

4. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the self-organizing path planning method for autonomous driving according to any one of claims 1 to 3.

5. A device for self-organizing path planning for autonomous driving, comprising a memory and a processor, wherein: The memory stores the following instructions that can be executed by the processor: used to execute the steps of the method for self-organizing path planning for autonomous driving as described in any one of claims 1 to 3.

6. A self-organizing path planning device for autonomous driving, characterized in that: include: Acquisition module, processing module, path selection module; among them, The acquisition module is configured to acquire, at an intersection node, traffic status information of each sub-section in a road network where the current intersection node is located; wherein the road network is a pre-set basic unit for planning the next local driving path starting from the current intersection node; each road section in the road network is divided into a preset number of sub-sections; and the traffic status information reflects critical density and traffic density; The processing module is used to encode the traffic status information of each sub-segment obtained in a one-hot encoding manner, and combine the encoding results of each sub-segment into path planning target information; wherein, the encoding of the traffic status information of each sub-segment obtained in a one-hot encoding manner includes: Divide the traffic density space into 10 intervals; The range of is divided into 9 density intervals of the same size, corresponding to the 1st to 9th dimensions of the state vector. It is divided into a density interval, corresponding to the 10th dimension of the state vector; wherein k is a preset constant; according to the one-hot encoding method, a 10×1 vector is used to represent the traffic state information in each sub-segment, and each dimension corresponds to each interval of the traffic density space in turn. The path selection module is used to select a corresponding path from a pre-trained set of candidate paths according to the path planning target information as the next local driving path starting from the current intersection node.

7. The self-organizing path planning device according to claim 6, further comprising: The training module is used to encode the sample traffic status information of each sub-section in the road network where the current intersection node is located according to the sample traffic status information of each sub-section obtained at the intersection node according to a preset encoding method, and combine the sample encoding results of each sub-section into path planning target sample information; use the path planning target sample information as the input of a preset training model, use the candidate path corresponding to the path planning target sample information in the candidate path set as the output of the preset training model, and train the parameters of the preset training model to obtain a path selection model.

8. The self-organizing path planning device according to claim 7, wherein: The path selection module is used to input the path planning target information into the path selection model, and the path selection model selects a path corresponding to the path planning target information from a set of candidate paths and outputs the selected path.

Images