A method for establishing an unmanned driving model based on a big data platform and an unmanned driving system

Abstract

The present invention provides a method for establishing an unmanned driving model based on a big data platform and an unmanned driving system. Through vehicle-mounted camera and radar recognition technologies, the characteristics of moving targets, marking lines, working media, obstacles, etc. in the working scenario are recognized, and feature learning and calculation are performed through corresponding algorithms to establish discrete models. Finally, the discrete models are fused to establish a common autonomous driving model, which is uploaded to the big data platform. The big data performs machine learning and calculation on different autonomous driving models to establish a massive big data model library. When the vehicle works in a fixed occasion, the corresponding big data model is retrieved from the big data platform.

Description

Technical Field

[0001] The present invention relates to a method for establishing an unmanned driving model based on a big data platform and an unmanned driving system, belonging to the technical field of unmanned driving. Background Art

[0002] Currently, with the increasingly widespread application of unmanned driving technology in passenger cars, the demand for unmanned driving in construction machinery vehicles is also becoming more and more urgent. An economical and efficient unmanned driving technology has become increasingly important. The working environment of construction machinery vehicles is complex and changeable, with numerous driving models, making the implementation of unmanned driving technology more complex.

[0003] The existing driving models have the following defects:

[0004] 1. The scenario is single. The main working scenario of traditional passenger cars is the road, and the driving space scenario is single. During the process of autonomous driving, it is only necessary to judge the road vehicles and obstacles to achieve it. The working space scenario of construction machinery vehicles is relatively complex. In the working scenario, the driving path is not fixed, the working medium is different, and the surrounding obstacles also vary greatly.

[0005] 2. The hardware computing power requirement is relatively high. The autonomous driving of traditional passenger cars mainly relies on the computing power of the vehicle body CPU. The vehicle body performs algorithm learning and storage, and conducts environmental perception and vehicle control through multi-threading.

[0006] 3. The cloud platform function is single. The traditional passenger car market mainly uses the big data platform for vehicle condition data collection and analysis, with a single function. Summary of the Invention

[0007] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for establishing an unmanned driving model based on a big data platform and an unmanned driving system, which can establish a large number of working models and solve the pain points of complex and changeable environments.

[0008] To achieve the above object, the present invention is implemented by the following technical solutions:

[0009] In the first aspect, the present invention provides a method for establishing an unmanned driving model based on a big data platform, including the following steps:

[0010] Partition the environment around the vehicle into six partitions: left, right, front, rear, upper, and lower, and establish a working scenario model through radar sensors and camera sensors;

[0011] Identify and judge the medium of the image information collected by the camera through image recognition technology, judge the distance, volume, and height of the obstacles, and match the optimal algorithm;

[0012] Perform cross-variation processing on the six collected orientation models to obtain multiple different fusion scenarios, mainly including the main vehicle cab scenario, the working component scenario, and the vehicle peripheral scenario.

[0013] Perform computational analysis on multiple fusion scenarios through edge-side autonomous driving algorithms, traverse and search for the optimal model of scenario fitness on the big data cloud platform, and load it into the vehicle body storage medium.

[0014] According to the working models of each scenario, screen and fuse a panoramic scenario from each scenario and establish a model. After the model is completed, upload it to the cloud platform.

[0015] Furthermore, the method for establishing a working scenario model through radar sensors and camera sensors includes:

[0016] Based on the image segmentation technology of deep learning, achieve the purpose of image segmentation through the image segmentation method based on neural networks.

[0017] Based on the computational method of neural networks, achieve the accuracy of object perception through the combination of multi-point sampling and multi-scale feature fusion, as well as the perception of multiple neural networks.

[0018] Based on the target motion model and tracking algorithm of radar technology, perceive the moving objects in the scenario, and obtain the characteristics of the working medium, moving objects, obstacles, and standard markings in the working scenario.

[0019] Perform identification and judgment on the medium, markings, working environment, etc. of the images collected by the camera through image recognition technology, and the radar judges the distance, volume, and height of the obstacles and matches the optimal algorithm.

[0020] Furthermore, the method for traversing and searching for the optimal model of scenario fitness on the big data platform and obtaining the optimal automatic algorithm model includes:

[0021] Obtain an approximate model through the approximate matching algorithm, mainly by obtaining the similarity degree of the common subgraph and the common hypergraph through the approximate matching algorithm.

[0022] Screen the models from the algorithms obtained in the previous step through the exact matching algorithm, and screen out the algorithm models with consistent attributes.

[0023] Perform algorithm fusion among the multiple obtained algorithm models to generate the optimal algorithm.

[0024] Furthermore, the method also includes:

[0025] When the current working scenario does not match the scenario of the loaded autonomous driving working model, manual access operation is performed. The new influencing factors brought about by the manual operation are cached in the vehicle storage medium, and according to the importance of the new influencing factors, the current autonomous driving model is updated, a new autonomous driving model is established, and uploaded to the big data platform.

[0026] Furthermore, the construction method of the cloud platform includes:

[0027] The autonomous driving models transmitted by the vehicle to the big database through wireless communication technology are used to establish a big data model. New data models are mined from the old data, and various data models are fused to generate new data models.

[0028] Classify the big data models according to the influencing factors, where the influencing factors include vehicle type, working environment, vehicle components, road vehicles, and non-road vehicles, and classify all big data models in a pyramid manner to form a big data model database.

[0029] Furthermore, the big data model database is divided into road vehicles and non-road vehicles.

[0030] Each model database is divided into an automatic obstacle avoidance module, a medium recognition module, a path planning module, a vehicle control module, a working condition analysis module, and an artificial intelligence module;

[0031] In the application of the automatic obstacle avoidance module in road vehicles, it mainly avoids obstacles such as road vehicles, road pedestrians, and road obstacles. In non-road vehicles, in addition to avoiding vehicles, personnel, and obstacles in the construction scenario, it must also be able to identify obstacles above the vehicle, gullies, and construction site markers in the construction scenario for obstacle avoidance.

[0032] In the application of the medium recognition module in road vehicles, it mainly recognizes road signs, landmark lines, and traffic lights, while in non-road vehicles, it also needs to recognize working media, working signs, working vehicles, and working environments in the construction scenario.

[0033] In the application of the vehicle control module in road vehicles, it mainly performs functions such as emergency braking, automatic obstacle avoidance, automatic driving, and speed switching on the vehicle. In non-road vehicles, it also needs to control working structures such as crane booms and pile driver bits.

[0034] The path planning module can draw a path through prior environmental modeling or automatically establish a path model through cameras and sensors to work on the specified path.

[0035] The working condition analysis module mainly monitors parameters such as the oil temperature, oil pressure, hydraulic pressure, engine, and braking of the vehicle, and analyzes the working condition of the vehicle in real time through these data to control the healthy operation of the vehicle.

[0036] The artificial intelligence module mainly splits and analyzes the numerous collected models, extracts the influencing factors during the autonomous driving process, automatically constructs new models, and can automatically match the optimal autonomous driving model through the vehicle uploading the working environment model, and trains a better and more reasonable autonomous driving model.

[0037] In a second aspect, the present invention provides a vehicle driverless system, including a cloud platform, a vehicle body control system and a vehicle body sensing system which are connected to each other;

[0038] The vehicle body control system obtains the scene space information through the vehicle body sensing system and uploads the scene space information to the cloud platform;

[0039] The cloud platform processes and searches according to the scene space information, matches the most suitable driving model and sends it to the vehicle body control system; the vehicle body control system loads the driving model and controls the underlying driving system to perform autonomous driving according to the driving model.

[0040] If the current working scene does not match the scene of the loaded autonomous driving working model, manual access operation is performed. The new influencing factors brought by the manual operation are cached in the vehicle body storage medium, and the current autonomous driving model is updated according to the importance of the new influencing factors, a new autonomous driving model is established and uploaded to the big data platform.

[0041] Furthermore, the vehicle body sensing system includes a radar sensor and a camera sensor installed on the vehicle;

[0042] The vehicle body sensing system performs data interaction with vehicle components such as the vehicle control system and the engine controller system through the CAN bus, and obtains the scene space data of the space scene through the vehicle body camera and the radar system, and transmits the scene space data to the vehicle body control system;

[0043] The vehicle body control system extracts the autonomous driving model factors from the scene space data. The autonomous driving model factors include the construction site medium, vehicle speed, position, obstacles, and engine speed, and establishes an optimal autonomous driving algorithm according to the binding force and free combination technology of these factors.

[0044] Furthermore, the method for matching the most suitable driving model includes:

[0045] After the vehicle starts, the vehicle body control system traverses and searches in the big data platform according to the current working space scene and the current influencing factors collected by the vehicle body sensing system to obtain an initial autonomous driving model.

[0046] When the current working scenario does not match the scenario of the loaded autonomous driving working model, manual access operation is performed. The new influencing factors brought by the manual operation are cached in the vehicle storage medium, and according to the importance of the new influencing factors, the current autonomous driving model is updated, a new autonomous driving model is established, and uploaded to the big data platform.

[0047] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0048] 1. The present application establishes a large number of working models through technical means such as AI recognition technology, edge computing technology, big data cloud platform, and cloud computing to solve the pain points of complex and changeable environments;

[0049] 2. The present application reduces the pain point of excessive dependence on vehicle hardware by combining the big data platform and the vehicle body, greatly reducing the hardware cost and development difficulty;

[0050] 3. While collecting working condition data, the present application makes full use of technologies such as big data analysis, cloud computing, device profiling, and cloud computing. Through the splitting and combination of intelligent driving models generated in multiple scenarios, the vehicle traverses and searches the database for the working scenario of the current working condition to find a suitable autonomous driving model, so that the vehicle can work under the optimal working model. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flowchart of the end-cloud training method for the autonomous driving model;

[0052] Figure 2 is a structure diagram of the cloud platform model;

[0053] Figure 3 is a flowchart of the end-cloud training method for the autonomous driving model. DETAILED DESCRIPTION OF THE INVENTION

[0054] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the protection scope of the present invention.

[0055] Embodiment 1:

[0056] The present invention is mainly used for construction machinery vehicles. The construction machinery vehicles continuously learn and build models in complex working scenarios to establish a complete driverless database. It mainly includes:

[0057] Edge-side training method, such as Figure 1 First, the vehicle surrounding environment is partitioned, mainly divided into six partitions: left, right, front, rear, upper, and lower. And a working scenario model is established through radar sensors and camera sensors;

[0058] The image information collected by the camera is used for medium recognition and judgment through image recognition technology, and the radar is used to judge information such as the distance, volume, and height of obstacles, and the optimal algorithm is matched;

[0059] The six azimuth models collected are subjected to crossover and mutation processing to obtain multiple different fusion scenarios. It mainly includes the main vehicle cab scenario, the working component scenario, the vehicle peripheral scenario, etc.;

[0060] Multiple fusion scenarios are calculated and analyzed through the edge-side autonomous driving algorithm, and the model with the optimal scene fitness is traversed and searched on the platform and loaded into the vehicle body storage medium;

[0061] According to the working models of each scenario, the panoramic scenario is screened and fused from each scenario, and a model is established. After the model is completed, it is uploaded to the cloud platform;

[0062] Big data modeling technology Figure 2 , the autonomous driving model transmitted to the big database through the vehicle body via wireless communication technology is used to establish a big data model. New data models are mined from the old data, and various data models are fused to generate new data models. The big data models are classified, and they can be classified according to influencing factors such as vehicle type, working environment, vehicle components, road vehicles, and non-road vehicles. Try to classify all big data models in a pyramid style, which can be faster and more convenient when traversing and searching the data model.

[0063] For the big data model database, it is mainly divided into road vehicles and non-road vehicles. Each model is further divided into six modules: automatic obstacle avoidance module, medium recognition module, path planning module, vehicle control module, working condition analysis module, and artificial intelligence module.

[0064] In the application of the automatic obstacle avoidance module in road vehicles, it mainly avoids obstacles such as road vehicles, road pedestrians, and road obstacles. In non-road vehicles, in addition to avoiding vehicles, personnel, and obstacles in the construction scene, it must also be able to identify obstacles above the vehicle, gullies, and construction site markers in the construction scene for obstacle avoidance.

[0065] In the application of the medium recognition module in road vehicles, it mainly recognizes road signs, landmark lines, and traffic lights, while in non-road vehicles, it also needs to recognize working media, working signs, working vehicles, and working environments in the construction scene.

[0066] In the application of the vehicle control module in road vehicles, it mainly performs functions such as emergency braking, automatic obstacle avoidance, automatic driving, and speed switching on the vehicle. In non-road vehicles, it also needs to control working structures such as the boom of a crane and the drill bit of a pile driver.

[0067] The path planning module can draw paths through prior environmental modeling or automatically establish a path model through cameras and sensors, and work on the specified paths.

[0068] The working condition analysis mainly monitors parameters such as the oil temperature, oil pressure, hydraulics, engine, and braking of the vehicle, and analyzes the working condition of the vehicle in real time through these data, so as to control the healthy operation of the vehicle.

[0069] The artificial intelligence module mainly splits and analyzes the numerous collected models, extracts the influencing factors during the automatic driving process, automatically constructs new models, and can automatically match the optimal automatic driving model through the working environment model uploaded by the vehicle, and trains a better and more reasonable automatic driving model.

[0070] Data interaction is carried out with vehicle components such as the vehicle control system and the engine controller system through the CAN bus, and spatial scene data of the spatial scene is obtained through the vehicle body camera and the radar system, and automatic driving model factors are extracted from the spatial scene data. These factors include the medium at the construction site, vehicle speed, position, obstacles, engine speed, etc., and an optimal automatic driving algorithm is established according to the binding force and free combination technology of these factors.

[0071] After the vehicle starts, the vehicle body control system makes full use of the 5G low-latency and high-bandwidth technology, traverses and searches in the big data platform according to the current working space scene and the current influencing factors, and obtains the initial automatic driving model. When the current working scene does not match the scene of the loaded automatic driving working model, manual access operation is performed, and the new influencing factors brought by the manual operation are cached in the vehicle body storage medium, and the current automatic driving model is updated according to the importance of the new influencing factors, a new automatic driving model is established, and uploaded to the big data platform.

[0072] Key points of the present invention

[0073] 1) When establishing the automatic driving model, algorithms need to be established for numerous influencing factors and the importance needs to be analyzed, and the algorithms need to learn in real time.

[0074] The algorithm learning is mainly analyzed through the edge side, mainly analyzing the working characteristics of different engineering vehicles. For example, when a mining truck is loading and unloading goods, if the medium of ore is not recognized in the inherent algorithm, it is necessary to determine that this influencing factor is important and re-learn.

[0075] 2) When performing big data modeling, numerous automatic driving models need to be classified, which can be classified according to numerous influencing factors, and the closest automatic driving model can be obtained immediately according to the influencing factors obtained immediately when the vehicle starts.

[0076] 3) According to the new workspace scenario, the corresponding autonomous driving model learns and updates to make the autonomous driving model more perfect.

[0077] 4) When the vehicle does not recognize the new workspace scenario, manual intervention is required. When manual intervention is carried out, the local storage medium and core computing power should be fully utilized to calculate, learn, and store the current data.

[0078] This embodiment provides an autonomous driving system, including

[0079] Cloud platform - vehicle body control system - vehicle body sensing system

[0080] The vehicle body control system obtains scene control information through the vehicle body sensing system and uploads the scene space information to the cloud platform;

[0081] The cloud platform processes and searches according to the scene space information, finds the most suitable driving model and sends it to the vehicle body control system;

[0082] The vehicle body control system loads the driving model and controls the underlying driving system to perform autonomous driving according to the driving model. If the current working scene does not match the scene of the loaded autonomous driving working model, manual access operation is carried out. The new influencing factors brought by the manual operation are cached in the vehicle body storage medium, and the current autonomous driving model is updated according to the importance of the new influencing factors, a new autonomous driving model is established, and uploaded to the big data platform.

[0083] As Figure 3 shown, the end-cloud training method of the autonomous driving model is as shown in the figure:

[0084] Learn the current environment and train the model through sensor and camera parameters, and establish multiple parallel autonomous driving models;

[0085] Fuse the multiple trained parallel autonomous driving models into a current environment model and upload it to the cloud platform to establish an autonomous driving database; the database stores the vehicle working status and audio-video data models;

[0086] When the vehicle is working, according to the environmental perception and AI recognition technology, match the database autonomous driving model, match the nearest autonomous driving model, and load it to the edge side;

[0087] In the new working environment, calculate new driving parameters at the edge side, update the current autonomous driving model, and upload the new autonomous driving model to the database;

[0088] Learn and freely combine the massive big data in the big database to split out more autonomous driving models.

[0089] Embodiment 2:

[0090] This embodiment provides a vehicle driverless system, including a cloud platform, a vehicle body control system, and a vehicle body sensing system that are connected to each other;

[0091] The vehicle body control system obtains scene control information through the vehicle body sensing system and uploads the scene space information to the cloud platform;

[0092] The cloud platform processes and searches based on the scene space information, finds the most suitable driving model, and sends it to the vehicle body control system;

[0093] The vehicle body control system loads the driving model and controls the underlying driving system to perform autonomous driving according to the driving model. If the current working scene does not match the scene of the loaded autonomous driving working model, manual intervention is performed. The new influencing factors brought by the manual operation are cached in the end-side storage medium, and the current autonomous driving model is updated according to the importance of the new influencing factors, a new autonomous driving model is established, and uploaded to the big data platform.

[0094] The vehicle body sensing system includes a radar sensor and a camera sensor installed on the vehicle;

[0095] The vehicle body sensing system performs data interaction with vehicle components such as the vehicle control system and the engine controller system through the CAN bus, and obtains the scene space data of the space scene through the vehicle body camera and the radar system, and transmits the scene space data to the vehicle body control system;

[0096] The vehicle body control system extracts autonomous driving model factors from the scene space data. The autonomous driving model factors include construction site medium, vehicle speed, position, obstacles, and engine speed, and an optimal autonomous driving algorithm is established according to the binding force and free combination technology of these factors.

[0097] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0098] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0099] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means that implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0100] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0101] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art of the present technology, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for establishing an unmanned driving model based on a big data cloud platform, characterized in that, it includes the following steps: Partition the surrounding environment of the vehicle into six partitions: left, right, front, rear, upper, and lower, and establish a working scenario model through radar sensors and camera sensors; Based on the image segmentation technology of deep learning, perform image segmentation through the image segmentation method based on neural networks; Based on the calculation method of neural networks, perform object perception by adopting multi-point sampling combined with multi-scale feature fusion and the perception of multiple neuron networks; Based on the target motion model and tracking algorithm of radar technology, perceive the moving objects in the scenario, and obtain the characteristics of working media, moving objects, obstacles, and standard markings in the working scenario; Use image recognition technology to identify and judge media, signs and markings, and working environment for the image information collected by the camera, and use radar to judge the distance, volume, and height of obstacles; Perform cross-variation processing on the six azimuth models collected to obtain multiple different fusion scenarios; mainly including the main vehicle cab scenario, working component scenario, and vehicle surrounding scenario; Perform computational analysis on multiple fusion scenarios through edge-side autonomous driving algorithms, traverse and search for the optimal model of scenario fitness on the big data cloud platform, obtain the optimal autonomous algorithm model, and load it into the vehicle body storage medium; According to the working models of each scenario, screen and fuse panoramic scenarios from each scenario and establish a model; After the model is completed, upload it to the cloud platform; The method further includes: When the current working scenario does not match the loaded autonomous driving working model scenario, manual access operation is performed. The new influencing factors brought by manual operation are cached in the vehicle body storage medium, and according to the importance of the new influencing factors, the current autonomous driving model is updated, a new autonomous driving model is established, and uploaded to the big data cloud platform; The construction method of the cloud platform includes: The autonomous driving model of the vehicle body is transmitted to the big database through wireless communication technology to establish a big data model; mine new data models from old data, and fuse various data models to generate new data models; Classify the big data models according to the influencing factors, and the influencing factors include vehicle type, working environment, vehicle components, road vehicles, and non-road vehicles, and classify all big data models in a pyramid manner to form a big data model database.

2. The method for establishing an unmanned driving model based on a big data cloud platform according to claim 1, characterized in that, The method for traversing and searching for the optimal model of scenario fitness on the big data cloud platform and obtaining the optimal autonomous algorithm model includes: Obtain an approximate model through the approximate matching algorithm, mainly obtaining the similarity degree of the common subgraph and the common hypergraph through the approximate matching algorithm; Screen the algorithm models from the algorithms obtained in the previous step through the exact matching algorithm, and screen out the algorithm models with consistent attributes; Among the obtained multiple algorithm models, perform algorithm fusion to generate the optimal algorithm.

3. The method for establishing an unmanned driving model based on a big data cloud platform according to claim 1, characterized in that, The big data model database is divided into on-road vehicles and off-road vehicles; Each model database is divided into an automatic obstacle avoidance module, a medium recognition module, a path planning module, a vehicle control module, a working condition analysis module, and an artificial intelligence module; In the application of the automatic obstacle avoidance module in on-road vehicles, it mainly avoids obstacles for on-road vehicles, pedestrians, and road obstacles; in off-road vehicles, in addition to avoiding vehicles, personnel, and obstacles in the construction scene, it must also be able to identify obstacles above the vehicle, gullies, and construction site markers in the construction scene for obstacle avoidance; In the application of the medium recognition module in on-road vehicles, it mainly recognizes road signs, landmark lines, and traffic lights, while in off-road vehicles, it also needs to recognize working media, working signs, working vehicles, and working environments in the construction scene; In the application of the vehicle control module in on-road vehicles, it mainly performs emergency braking, automatic obstacle avoidance, automatic driving, and speed switching functions on the vehicle. In off-road vehicles, it also needs to control the working structures of the crane boom and pile driver bit; The path planning module can draw a path through prior environmental modeling or automatically establish a path model through cameras and sensors to work on the specified path; The working condition analysis module mainly monitors the oil temperature, oil pressure, hydraulic pressure, engine, and braking parameters of the vehicle, and analyzes the working condition of the vehicle in real time through these data to control the healthy operation of the vehicle; The artificial intelligence module mainly splits and analyzes the numerous collected models, extracts the influencing factors during the automatic driving process, automatically forms new models, and can automatically match the optimal automatic driving model through the vehicle uploading the working environment model, and trains a better and more reasonable automatic driving model.

4. A vehicle unmanned driving system for implementing the method for establishing an unmanned driving model based on a big data cloud platform as described in claim 1, characterized in that, it includes a connected cloud platform, a vehicle body control system, and a vehicle body sensing system; The vehicle body control system obtains the scene space information through the vehicle body sensing system and uploads the scene space information to the cloud platform; The cloud platform processes and searches according to the scene space information, matches the most suitable driving model, and sends it to the vehicle body control system; The vehicle body control system loads the driving model and controls the underlying driving system for automatic driving according to the driving model.

5. The vehicle unmanned driving system according to claim 4, characterized in that, The vehicle body sensing system includes a radar sensor and a camera sensor installed on the vehicle; The vehicle body sensing system conducts data interaction with vehicle control system and engine controller system vehicle components through the CAN bus, and obtains the scene space data of the spatial scene through the vehicle body camera and radar system, and transmits the scene space data to the vehicle body control system; The vehicle body control system extracts the automatic driving model factors from the scene space data. The automatic driving model factors include construction site media, vehicle speed, position, obstacles, and engine speed, and establishes the optimal automatic driving algorithm according to the constraints and free combinations of these factors.

6. The vehicle unmanned driving system according to claim 4, characterized in that, The method for the vehicle body control system to establish the optimal automatic driving algorithm includes: After the vehicle starts, the vehicle body control system traverses and searches in the big data cloud platform according to the current working space scenario and current influencing factors collected by the vehicle body sensing system, and obtains the initial automatic driving model; When the current working scenario does not match the scenario of the loaded automatic driving working model, manual access operation is performed. The new influencing factors brought by the manual operation are cached in the vehicle body storage medium, and according to the importance of the new influencing factors, the current automatic driving model is updated, a new automatic driving model is established, and uploaded to the big data cloud platform.

Images