Roadside perception distributed computing system and method based on vehicle-road-cloud integrated architecture

Abstract

The present invention relates to the technical field of intelligent connected vehicles, and specifically to a roadside perception distributed computing system and method based on a vehicle-road-cloud integrated architecture, comprising a convergence switch, an access switch, a computing chip module, a laser radar and a camera, wherein the access switch is connected to the convergence switch, the computing chip module is connected to the access switch, the laser radar is connected to the computing chip module, the camera is connected to the computing chip module, the laser radar is directly connected to the camera to the computing chip module, the original laser point cloud data and image data do not need to be transmitted through the switch, which greatly saves the transmission time of the original data, and at the same time, the computing chip module is installed in the all-in-one computer, each computing chip module only needs to process the data of a corresponding laser radar and a camera, thereby greatly reducing the computing power demand for the computing unit, and then the perception algorithm calculation task can be completed by a low-cost computing chip, thereby achieving the purpose of reducing costs.

Description

Technical Field

[0001] The present invention relates to the field of intelligent connected vehicle technology, and in particular to a roadside perception distributed computing system and method based on a vehicle-road-cloud integrated architecture. Background Art

[0002] Vehicle-road-cloud integration is a comprehensive intelligent transportation system concept that realizes intelligent management and services of the transportation system by closely integrating vehicles, road infrastructure and cloud computing platforms. It is based on system collaborative perception, decision-making and control to realize the information-physical system of safe, energy-saving, comfortable and efficient operation of intelligent networked vehicle transportation system. The construction of roadside infrastructure mainly includes roadside perception sensors (lidar, millimeter-wave radar, cameras, etc.), roadside edge computing units and communication units.

[0003] The current technical solutions for roadside perception systems all adopt a centralized computing architecture. At a standard intersection, at least four lidars and four cameras are required to transmit their respective raw point cloud data and image data to the edge computing unit through a switch. The edge computing unit then centrally processes and calculates all the raw data to obtain the final perception result.

[0004] However, in the above architecture, the edge computing unit needs to simultaneously receive and process a large amount of raw data from sensors, which places high demands on the computing power of edge computing. This requires a high-computing-power GPU chip, which is costly. Summary of the invention

[0005] The purpose of the present invention is to provide a roadside perception distributed computing system and method based on a vehicle-road-cloud integrated architecture, aiming to solve the problem that the edge computing unit needs to simultaneously receive and process a large number of raw data from sensors, which requires high computing power for edge computing and thus requires a high-computing-power GPU chip, which is costly.

[0006] To achieve the above-mentioned object, the present invention provides a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture, comprising a convergence switch, an access switch, a computing chip module, a laser radar and a camera, wherein the access switch is connected to the convergence switch, the computing chip module is connected to the access switch, the laser radar is connected to the computing chip module, and the camera is connected to the computing chip module;

[0007] The aggregation switch is used to concentrate the traffic data of multiple access ports to ports with higher bandwidth so as to better manage and transmit data;

[0008] The access switch is used to provide a network access point for the terminal device;

[0009] The computing chip module is used to perform real-time calculations on the received data to obtain the final perception result;

[0010] The laser radar is used to identify traffic targets;

[0011] The camera is used to collect traffic image information and identify traffic targets based on the image information.

[0012] The computing chip module includes a camera driving unit and a laser radar driving unit, the camera driving unit is connected to the camera, and the laser radar driving unit is connected to the laser radar;

[0013] The camera driving unit is used to drive the camera and obtain raw data from the camera;

[0014] The laser radar unit is used to drive the laser radar and obtain raw data from the laser radar.

[0015] The computing chip module further includes a front fusion perception algorithm unit, and the front fusion perception algorithm unit is connected to the camera driving unit and the laser radar driving unit respectively;

[0016] The pre-fusion perception algorithm unit is used to receive the collected original data and obtain the structured perception result within the coverage area of ​​the device after the perception operation.

[0017] Wherein, the computing power chip module further includes a distributed computing task scheduling unit, and the distributed computing task scheduling unit is connected to the front fusion perception algorithm unit;

[0018] The distributed computing task scheduling unit is used to receive the perception results so that each node can communicate with each other and transfer data.

[0019] Wherein, the computing power chip module further includes a post-fusion perception algorithm unit, and the post-fusion perception algorithm unit is connected to the distributed computing task scheduling unit;

[0020] The post-fusion perception algorithm unit is used to receive the perception results of all nodes, and fuse all the perception results to output the final perception result of the entire intersection.

[0021] A roadside perception distributed computing method based on a vehicle-road-cloud integrated architecture includes the following steps:

[0022] The camera collects traffic image information, and identifies traffic targets based on the image information, and the radar identifies traffic targets;

[0023] The computing chip module receives the traffic target data collected by the corresponding camera and the radar, and performs real-time calculations on the received data to obtain the final perception result;

[0024] The access switch provides a network access point for data transmitted by the computing chip module;

[0025] The aggregation switch concentrates the traffic data of multiple access ports to ports with higher bandwidth to better manage and transmit data.

[0026] The present invention provides a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture, wherein the laser radar and the camera are directly connected to the computing chip module, and the original laser point cloud data and image data do not need to be transmitted through a switch, thereby greatly saving the transmission time of the original data. At the same time, the computing chip module is installed and deployed in each of the integrated sensing and computing machines, and each of the computing chip modules only needs to process the data of the corresponding laser radar and the camera, thereby greatly reducing the computing power requirements for the computing unit, and then the perception algorithm calculation tasks can be completed by low-cost computing chips, thereby achieving the purpose of reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0028] Figure 1 It is a hardware architecture diagram of a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture according to the first embodiment of the present invention.

[0029] Figure 2 It is a schematic diagram of the structure of the computing power chip module of the first embodiment of the present invention.

[0030] Figure 3 It is a workflow diagram of a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture according to the first embodiment of the present invention.

[0031] Figure 4 It is a flowchart of the steps of a roadside perception distributed computing method based on a vehicle-road-cloud integrated architecture according to the second embodiment of the present invention.

[0032] In the figure: 101-aggregation switch, 102-access switch, 103-computing chip module, 104-lidar, 105-camera, 106-camera driver unit, 107-lidar driver unit, 108-front fusion perception algorithm unit, 109-distributed computing task scheduling unit, 110-back fusion perception algorithm unit. DETAILED DESCRIPTION

[0033] Embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0034] The first embodiment of the present application is:

[0035] See also Figures 1 to 3 ,in Figure 1 It is a hardware architecture diagram of a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture according to the first embodiment of the present invention. Figure 2 It is a schematic diagram of the structure of the computing power chip module of the first embodiment of the present invention. Figure 3 It is a workflow diagram of a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture according to the first embodiment of the present invention.

[0036] The present invention provides a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture, including a convergence switch 101, an access switch 102, a computing chip module 103, a laser radar 104 and a camera 105, wherein the computing chip module 103 includes a camera driving unit 106, a laser radar driving unit 107, a front fusion perception algorithm unit 108, a distributed computing task scheduling unit 109 and a back fusion perception algorithm unit 110. The aforementioned solution solves the problem that the edge computing unit needs to simultaneously receive and process a large amount of raw data from sensors, which requires high computing power for edge computing, thereby requiring a GPU chip with high computing power, which is costly.

[0037] In this embodiment, the laser radar 104 sensor, the camera 105 sensor and the computing chip module 103 are integrated together to form an all-in-one sensing and computing machine, which exchanges data with other roadside devices through the access switch 102 and the aggregation switch 101.

[0038] The access switch 102 is connected to the aggregation switch 101, the computing chip module 103 is connected to the access switch 102, the laser radar 104 is connected to the computing chip module 103, and the camera 105 is connected to the computing chip module 103;

[0039] The aggregation switch 101 is used to centralize the traffic data of multiple access ports to ports with higher bandwidth so as to better manage and transmit data;

[0040] The access switch 102 is used to provide a network access point for the terminal device;

[0041] The computing chip module 103 is used to perform real-time calculations on the received data to obtain the final perception result;

[0042] The laser radar 104 is used to identify traffic targets;

[0043] The camera 105 is used to collect traffic image information and identify traffic targets based on the image information.

[0044] The laser radar 104 and the camera 105 are directly connected to the computing chip module 103, and the original laser point cloud data and image data do not need to be transmitted through a switch, which greatly saves the transmission time of the original data. At the same time, the computing chip module 103 is installed and deployed in each of the all-in-one synergy computing machines. Each computing chip module 103 only needs to process the data of the corresponding laser radar 104 and the camera 105, thereby greatly reducing the computing power requirements for the computing unit, and then the perception algorithm calculation tasks can be completed through low-cost computing chips, thereby achieving the purpose of reducing costs.

[0045] For further information, see Figure 2 The computing chip module 103 includes a camera driving unit 106, a laser radar driving unit 107, a front fusion perception algorithm unit 108, a distributed computing task scheduling unit 109 and a back fusion perception algorithm unit 110. The camera driving unit 106 is connected to the camera 105, and the laser radar driving unit 107 is connected to the laser radar 104.

[0046] The camera driving unit 106 is used to drive the camera 105 and obtain raw data from the camera 105;

[0047] The laser radar 104 unit is used to drive the laser radar 104 and obtain raw data from the laser radar 104 .

[0048] The computing chip module 103 further includes a front fusion perception algorithm unit 108, and the front fusion perception algorithm unit 108 is connected to the camera driving unit 106 and the laser radar driving unit 107 respectively;

[0049] The pre-fusion perception algorithm unit 108 is used to receive the collected original data and obtain the structured perception result within the coverage area of ​​the device after the perception operation.

[0050] The computing chip module 103 further includes a distributed computing task scheduling unit 109, and the distributed computing task scheduling unit 109 is connected to the pre-fusion perception algorithm unit 108;

[0051] The distributed computing task scheduling unit 109 is used to receive the sensing results so that each node can communicate with each other and transfer data.

[0052] The computing chip module 103 further includes a post-fusion perception algorithm unit 110, and the post-fusion perception algorithm unit 110 is connected to the distributed computing task scheduling unit 109;

[0053] The post-fusion perception algorithm unit 110 is used to receive the perception results of all nodes, and fuse all the perception results to output the final perception result of the entire intersection.

[0054] The software architecture of this system consists of four nodes, namely the master node, slave node 1, slave node 2, and slave node 3. Each node is deployed in the computing unit of a synergistic computing machine.

[0055] Each node includes a camera driving unit 106, a lidar driving unit 107, a front fusion perception algorithm unit 108, a distributed computing task scheduling unit 109, and a post fusion perception algorithm unit 110.

[0056] After the camera driving unit 106 and the lidar driving unit 107 obtain the raw data from the sensor, they pass the data to the pre-fusion perception algorithm unit 108, obtain the structured perception results within the coverage of the device after the perception operation, and pass the perception results to the distributed computing task scheduling unit 109. Each node communicates with each other and transmits data through the distributed computing task scheduling unit 109.

[0057] The difference between the master node and the slave node is that the master node runs the post-fusion perception algorithm unit 110 in real time, receives all node perception results sent by the distributed computing task scheduling unit 109, and fuses all perception results to output the final perception result of the entire intersection. In the slave node, the post-fusion perception algorithm unit 110 is in a waiting state and is not actually running.

[0058] For further information, see Figure 3 , the working steps of this system include:

[0059] Each node of the system respectively obtains and analyzes the laser radar 104 point cloud data and the camera 105 image data in its own integrated computing machine;

[0060] Each node performs real-time calculations through its own computing chip to obtain structured perception results within the area covered by each all-in-one computing machine.

[0061] The distributed computing task scheduler running in each node publishes the operation status matrix S = [S1, S2, S3, S4] of each node, where S1 represents the working status of the master node, S2 represents the working status of slave node 1, S3 represents the working status of slave node 2, and S4 represents the working status of slave node 3. The value 1 represents that the node is working normally, and the value 0 represents that the node has crashed.

[0062] Calculate the product prod(S) of all elements of the matrix S;

[0063] If prod(S) is equal to 1, it proves that all nodes are in normal working condition. The master node obtains the perception results of all nodes through the distributed computing task scheduler, and runs the post-fusion perception algorithm program to obtain the fusion perception results of the entire intersection;

[0064] If prod(S) is equal to 0, it proves that there is a crash node, and the value of S(1) is determined;

[0065] If S(1)=1, it proves that the master node is running normally, and repeat step 5;

[0066] If S(1)=0, it proves that the master node has crashed, and the value of S(2) is determined;

[0067] If S(2)=1, it proves that slave node 1 is operating normally, and slave node 1 is converted to the master node, the post-fusion perception algorithm program is activated, and step 5 is repeated;

[0068] If S(2)=0, it proves that slave node 1 crashes, and determines the value of S(3);

[0069] If S(3)=1, it proves that slave node 2 is operating normally, and slave node 2 is converted to the master node, the post-fusion perception algorithm program is activated, and step 5 is repeated;

[0070] If S(3)=0, it proves that slave node 2 crashes, and determines the value of S(4);

[0071] If S(4)=1, it proves that slave node 3 is operating normally, and slave node 3 is converted to the master node, the post-fusion perception algorithm program is activated, and step 5 is repeated;

[0072] If S(4)=0, it proves that all nodes have crashed. Restart the system immediately and repeat step 1 after restart to obtain the original data.

[0073] When using a roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture of this embodiment, the laser radar 104 and the camera 105 are directly connected to the computing chip module 103, and the original laser point cloud data and image data do not need to be transmitted through a switch, which greatly saves the transmission time of the original data. At the same time, the computing chip module 103 is installed and deployed in each of the all-in-one sensing and computing machines. Each of the computing chip modules 103 only needs to process the data of the corresponding laser radar 104 and the camera 105, thereby greatly reducing the computing power requirements for the computing unit, and then the perception algorithm calculation tasks can be completed by low-cost computing chips, thereby achieving the purpose of reducing costs.

[0074] The second embodiment of the present application is:

[0075] Based on the first embodiment, please refer to Figure 4 ,in, Figure 4 This is a flowchart of the steps of a roadside perception distributed computing method based on a vehicle-road-cloud integrated architecture according to the second embodiment of the present invention. A roadside perception distributed computing method based on a vehicle-road-cloud integrated architecture according to this embodiment includes the following steps:

[0076] S201: the camera 105 collects traffic image information, and identifies traffic targets based on the image information, and the radar identifies traffic targets;

[0077] S202: receiving the traffic target data collected by the corresponding camera 105 and the radar through the computing chip module 103, and performing real-time calculation on the received data to obtain a final perception result;

[0078] S203: The access switch 102 provides a network access point for data transmitted by the computing chip module 103;

[0079] S204: The aggregation switch 101 aggregates the traffic data of multiple access ports to ports with higher bandwidth to better manage and transmit data.

[0080] Specifically, the camera 105 collects traffic image information and identifies traffic targets based on the image information. The radar identifies traffic targets. The laser radar 104 is directly connected to the computing chip module 103 with the camera 105, and transmits the collected raw data to the computing chip module 103 for processing. The computing chip module 103 receives the corresponding traffic target data collected by the camera 105 and the radar, and performs real-time calculations on the received data to obtain the final perception result. The access switch 102 provides a network access point for the data transmitted by the computing chip module 103, and cooperates with the The aggregation switch 101 concentrates the traffic data of multiple access ports to ports with higher bandwidth for better management and transmission of data, so that the original laser point cloud data and image data do not need to be transmitted through the switch, which greatly saves the transmission time of the original data. The computing chip module 103 is installed and deployed in each of the tele-sensing and computing all-in-one machines. Each of the computing chip modules 103 only needs to process the data of the corresponding laser radar 104 and the camera 105, thereby greatly reducing the computing power requirements for the computing unit, and then the perception algorithm calculation tasks can be completed through low-cost computing chips, thereby achieving the purpose of reducing costs.

[0081] What is disclosed above is only one or more preferred embodiments of the present application, and cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture, characterized in that: It includes a convergence switch, an access switch, a computing chip module, a laser radar and a camera, wherein the access switch is connected to the convergence switch, the computing chip module is connected to the access switch, the laser radar is connected to the computing chip module, and the camera is connected to the computing chip module; The aggregation switch is used to concentrate the traffic data of multiple access ports to ports with higher bandwidth so as to better manage and transmit data; The access switch is used to provide a network access point for the terminal device; The computing chip module is used to perform real-time calculations on the received data to obtain the final perception result; The laser radar is used to identify traffic targets; The camera is used to collect traffic image information and identify traffic targets based on the image information; The computing chip module includes a camera driving unit and a laser radar driving unit, wherein the camera driving unit is connected to the camera, and the laser radar driving unit is connected to the laser radar; The camera driving unit is used to drive the camera and obtain raw data from the camera; The laser radar unit is used to drive the laser radar and obtain raw data from the laser radar; The computing chip module also includes a front fusion perception algorithm unit, which is connected to the camera driving unit and the lidar driving unit respectively; the front fusion perception algorithm unit is used to receive the collected raw data and obtain the structured perception result within the coverage range of the device after the perception operation; The computing power chip module also includes a distributed computing task scheduling unit, and the distributed computing task scheduling unit is connected to the pre-fusion perception algorithm unit; The distributed computing task scheduling unit is used to receive the sensing results so that each node can communicate with each other and transfer data; Integrate the laser radar, the camera and the computing chip module together to form a sensory computing all-in-one machine; The software architecture of this system consists of at least two nodes, namely a master node and multiple slave nodes, each of which is deployed in the computing unit of a synergistic computing machine; The computing chip module also includes a post-fusion perception algorithm unit, which is connected to the distributed computing task scheduling unit; the post-fusion perception algorithm unit is used to receive the perception results of all nodes, and fuse all the perception results to output the final perception result of the entire intersection. The master node runs the post-fusion perception algorithm unit in real time, the slave node runs the pre-fusion perception algorithm unit in real time, and the post-fusion perception algorithm unit is in a waiting state.

2. A roadside perception distributed computing method based on a vehicle-road-cloud integrated architecture, applicable to the roadside perception distributed computing system based on a vehicle-road-cloud integrated architecture as claimed in claim 1, characterized in that: The following steps are involved: The camera collects traffic image information and identifies traffic targets based on the image information, and the radar identifies traffic targets; The computing chip module receives the traffic target data collected by the corresponding camera and the radar, and performs real-time calculations on the received data to obtain the final perception result; The access switch provides a network access point for data transmitted by the computing chip module; The aggregation switch concentrates the traffic data of multiple access ports to ports with higher bandwidth to better manage and transmit data.

Images