WO2022257056A1 - 车辆自动驾驶系统、方法、装置、计算机设备和存储介质 - Google Patents

车辆自动驾驶系统、方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2022257056A1
WO2022257056A1 PCT/CN2021/099280 CN2021099280W WO2022257056A1 WO 2022257056 A1 WO2022257056 A1 WO 2022257056A1 CN 2021099280 W CN2021099280 W CN 2021099280W WO 2022257056 A1 WO2022257056 A1 WO 2022257056A1
Authority
WO
WIPO (PCT)
Prior art keywords
inference
reasoning
request
vehicle
cloud
Prior art date
Application number
PCT/CN2021/099280
Other languages
English (en)
French (fr)
Inventor
庄奇
Original Assignee
深圳元戎启行科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳元戎启行科技有限公司 filed Critical 深圳元戎启行科技有限公司
Priority to PCT/CN2021/099280 priority Critical patent/WO2022257056A1/zh
Priority to CN202180050153.5A priority patent/CN115956045A/zh
Publication of WO2022257056A1 publication Critical patent/WO2022257056A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions

Definitions

  • the present application relates to the technical field of automatic driving, in particular to a vehicle automatic driving system, method, device, computer equipment and storage medium.
  • Deep learning is one of the most important solutions of artificial intelligence in the field of autonomous driving, especially in perception.
  • the design purpose of the deep learning inference engine is to speed up the inference of the deep learning model, and the inference of the deep learning model is completed by GPU or AI chip.
  • the perception module of automatic driving is mainly realized by deep learning method. Deep learning needs to use sensor information such as vehicle laser radar and camera as input, and obtain output through neuron network calculation, which is the perception result. The process of neuron network calculation is also called reasoning. process.
  • the computing platforms of self-driving car companies are all deployed on self-driving cars.
  • the power of the computing platform is related to the level of autonomous driving.
  • L1-L3 autonomous driving a small computing chip or a camera integrated with a small chip can be realized.
  • L4-level self-driving vehicles since the required deep learning models are more complex, more types, and more computationally intensive, it is impossible to perform real-time calculations with only a small chip at this time.
  • the solution of the L4 self-driving car company is to deploy the high-performance computing platform directly on the car, such as in the trunk or on the roof.
  • the existing technology because a large number of core algorithms are placed on the computing platform in the self-driving vehicle, there is a risk that the core algorithm of the computing platform will be leaked.
  • a vehicle automatic driving system comprising: a cloud reasoning server and an automatic driving vehicle; the automatic driving vehicle and the cloud reasoning server are connected through wireless communication;
  • the self-driving vehicle is used to generate and send an inference request to the cloud inference server, receive an inference result sent by the cloud inference server, and execute a corresponding driving instruction according to the inference result;
  • the cloud reasoning server is configured to receive the reasoning request sent by the self-driving vehicle, generate a corresponding reasoning result according to the reasoning request, and send the reasoning result to the self-driving vehicle;
  • the inference result is an instruction generated by the cloud inference server according to the inference request;
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform inference services;
  • the above data package is the inference data required for inference based on the request type.
  • a method for automatic driving of a vehicle comprising: receiving an inference request sent by an automatic driving vehicle; the inference request includes a request type and/or a data packet; the request type is a type that requires a cloud inference server to perform an inference service; The above data packet is the reasoning data required for reasoning according to the request type;
  • it also includes: parsing the reasoning request, and selecting a corresponding deep learning model according to the reasoning request;
  • it also includes: parsing the reasoning request, and judging whether the required deep learning model has been selected in the reasoning request;
  • the required deep learning model has been selected in the reasoning request, select the selected deep learning model in the reasoning request as the deep learning model used in reasoning processing.
  • it also includes: if the required deep learning model is not selected in the reasoning request, obtaining the reasoning purpose according to the reasoning request;
  • it also includes: receiving the registered inference engine service request sent by the self-driving vehicle; the registered inference engine service request is sent by the self-driving vehicle to apply for the cloud reasoning server to perform the inference request sent by it. offer information for inferential processing;
  • a vehicle automatic driving method comprising: generating and sending an inference request to a cloud inference server; the inference request includes a request type and/or a data packet; the request type is a type that requires a cloud inference server to perform an inference service; The data packet is the reasoning data required for reasoning according to the request type;
  • it also includes: when the instruction of the reasoning request is received and/or the operating state of the autonomous vehicle meets the generation conditions of the reasoning request, generating a corresponding reasoning request;
  • it also includes: receiving the reasoning result sent by the cloud reasoning server;
  • the driving of the automatic driving vehicle is controlled according to the driving instruction.
  • it also includes: sending a registered reasoning engine service request to the cloud reasoning server; the registered reasoning engine service request is sent by the self-driving vehicle for applying to the cloud reasoning server to perform reasoning on the reasoning request sent by it processed offer information;
  • the return information includes the return information of agreeing to register and the return information of rejecting registration.
  • the method further includes: obtaining a data packet through the vehicle-mounted road surface information acquisition device, the data packet including road condition image information and road condition point cloud information.
  • a vehicle automatic driving device comprising:
  • the inference request receiving module is used to receive the inference request sent by the self-driving vehicle; the inference request includes a request type and/or a data packet; the request type is a type that requires a cloud inference server to perform an inference service; the data packet is based on The inference data required for inference of the request type;
  • An inference result generation module configured to generate a corresponding inference result according to the inference request, and send the inference result to the self-driving vehicle; the inference result is generated by a cloud inference server through a deep learning model according to the inference request instructions for controlling the driving of the self-driving vehicle.
  • a vehicle automatic driving device comprising:
  • the reasoning request generation module is used to generate and send the reasoning request to the cloud reasoning server;
  • the reasoning request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud reasoning server to perform reasoning services;
  • the data package is Inference data required for inference based on request type;
  • the reasoning result receiving module is used to receive the reasoning result sent by the cloud reasoning server, and execute the corresponding driving instruction according to the reasoning result; the reasoning result is generated by the cloud reasoning server through the deep learning model according to the reasoning request Instructions for controlling the driving of the autonomous vehicle.
  • the vehicle automatic driving device further includes: a vehicle-mounted road surface information acquisition device, a vehicle-mounted driving control device, and a vehicle-mounted intelligent processing terminal;
  • the vehicle-mounted road surface information acquisition device is used to acquire data packets, the data packets include road condition image information and road condition point cloud information;
  • the vehicle-mounted intelligent processing terminal is set in the inference request generation module, and is used to generate an inference request, send the inference request to a cloud inference server, and receive an inference result generated by the cloud inference server;
  • the on-vehicle driving control device is connected with the inference result receiving module, and is used to control the driving of the self-driving vehicle according to the inference result generated by the cloud inference server.
  • the vehicle-mounted road surface information acquisition device includes a vehicle-mounted road surface image acquisition device and a vehicle-mounted radar system;
  • the vehicle-mounted road surface image collection device is used to collect road condition image information, and send the road condition image information to the vehicle-mounted intelligent processing terminal;
  • the vehicle-mounted radar system is used to collect road condition point cloud information, and send the road condition point cloud information to the vehicle-mounted intelligent processing terminal.
  • the vehicle-mounted intelligent processing terminal includes:
  • a processor configured to receive the road condition image information and road condition point cloud information, and generate an inference request
  • the communication device is used to realize the wireless communication between the vehicle-mounted intelligent processing and the cloud reasoning server.
  • the vehicle radar system includes at least one of a laser radar system, an ultrasonic radar system, a millimeter wave radar system and a microwave radar system.
  • the communication device includes at least one of WIFI device, Bluetooth device, GPRS device, 3G, 4G and 5G modules connected to the processor.
  • a computer device comprising a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires a cloud inference server to perform an inference service;
  • the data packet is required for inference according to the request type inference data;
  • a computer device comprising a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform an inference service;
  • the data packet is generated when reasoning according to the request type Inference data needed;
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires a cloud inference server to perform an inference service;
  • the data packet is required for inference according to the request type inference data;
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform an inference service;
  • the data packet is generated when reasoning according to the request type Inference data needed;
  • the above vehicle automatic driving system, method, device, computer equipment and storage medium generate and send reasoning requests to the cloud reasoning server wirelessly connected to the self-driving vehicle through the self-driving vehicle, and receive the reasoning request sent by the self-driving vehicle through the cloud reasoning server , generate the corresponding reasoning result according to the reasoning request, and send the reasoning result to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • Fig. 1 is a structural diagram of a vehicle automatic driving system in an embodiment
  • Fig. 2 is an information interaction diagram between the cloud reasoning server and the self-driving vehicle in the vehicle automatic driving system in one embodiment
  • Fig. 3 is the flow chart of information processing of the vehicle-mounted intelligent processing terminal in the vehicle automatic driving system in one embodiment
  • Fig. 4 is a structure diagram of reasoning request in the vehicle automatic driving system in one embodiment
  • Fig. 5 is a schematic flow chart of a vehicle automatic driving method in an embodiment
  • Fig. 6 is a schematic flow chart of a vehicle automatic driving method in another embodiment
  • Fig. 7 is a structural block diagram of a vehicle automatic driving device in an embodiment
  • Fig. 8 is a structural block diagram of a vehicle automatic driving device in another embodiment
  • Figure 9 is a block diagram of a computer device in one embodiment.
  • a vehicle automatic driving system including: a cloud reasoning server and an automatic driving vehicle; the automatic driving vehicle and the cloud reasoning server are connected by wireless communication;
  • the type of self-driving vehicle is not unique, specifically, it may be a car, a van, or a truck.
  • the cloud inference server is a server device used to perform neuron network calculations to obtain output, that is, perception results.
  • the process of neuron network calculations is also called the inference process.
  • the vehicle automatic driving system includes a cloud inference server and an automatic driving vehicle; wherein, each self-driving vehicle is wirelessly connected to the cloud inference server, and the wireless communication connection method is not specifically limited, for example, it can be connected through 5G communication.
  • the self-driving vehicle is used to generate and send an inference request to the cloud inference server, receive an inference result sent by the cloud inference server, and execute a corresponding driving instruction according to the inference result;
  • an inference request is an instruction sent by an autonomous vehicle to request a cloud inference server to perform an inference process, including a request type and/or a data packet.
  • the self-driving vehicle applies for registration of a new inference engine service
  • the self-driving vehicle selects the deep learning model that it wants to infer
  • the self-driving vehicle gives the input data that needs to be inferred
  • the self-driving vehicle requests
  • the output of the cloud inference server after inference and the self-driving vehicle need to obtain the last inference time, etc.
  • the inference result sent by the cloud inference server is an indispensable command for the self-driving vehicle to drive automatically.
  • the control system of the self-driving vehicle performs corresponding actions according to the received inference result to realize the safe driving of the self-driving vehicle.
  • the cloud reasoning server is configured to receive the reasoning request sent by the self-driving vehicle, generate a corresponding reasoning result according to the reasoning request, and send the reasoning result to the self-driving vehicle;
  • the cloud reasoning server is a server remotely connected to the self-driving vehicle for performing the reasoning process, and the cloud reasoning server may be a single server or a server group composed of multiple servers.
  • Figure 2 is a structural diagram of the cloud inference server in the vehicle automatic driving system. As shown in Figure 2, the cloud inference server is connected to each autonomous vehicle through wireless communication, and can receive the inference request sent by each autonomous vehicle in real time, including the request type and/or or data packets, and according to the content of the inference request, perform corresponding processing according to the preset program, generate the inference result, and send the inference result to the corresponding self-driving vehicle in real time, so that each self-driving vehicle can execute the corresponding operation according to the reasoning result in real time action.
  • the cloud reasoning server processes the data sent by the self-driving vehicle for predicting the trajectory of the self-driving vehicle according to the type of deep learning model selected by the self-driving vehicle, obtains the corresponding trajectory information of the self-driving vehicle, and sends it to The corresponding self-driving vehicle enables the self-driving vehicle to drive safely and smoothly.
  • the inference result is an instruction generated by the cloud inference server according to the inference request;
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform inference services;
  • the above data package is the inference data required for inference based on the request type.
  • the reasoning result is the instruction and information required by the self-driving vehicle generated by the cloud reasoning server according to the reasoning request.
  • Different reasoning requests have different corresponding reasoning results. For example, if the inference request is that the self-driving vehicle applies to register a new inference engine service, the inference result is the success or failure of the registration of the corresponding self-driving vehicle; if the reasoning request is that the self-driving vehicle selects the deep learning is the deep learning model selected by the self-driving vehicle; if the inference request is the output result after the self-driving vehicle requests the cloud inference server for inference, the corresponding inference result is the returned inference result; if the reasoning request is the self-driving vehicle needs to obtain the last If the inference time is , the inference result is the time spent in the last inference process.
  • the above vehicle automatic driving system generates and sends an inference request to a cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, receives the inference request sent by the self-driving vehicle through the cloud inference server, and generates a corresponding inference result according to the inference request, And send the reasoning result to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a method for automatic driving of a vehicle is provided, which is described by taking the method applied to a cloud reasoning server as an example, including the following steps:
  • Step 502 receiving an inference request sent by an autonomous vehicle; the inference request includes a request type and/or a data packet; the request type is a type that requires a cloud inference server to perform inference services; the data packet is inferred according to the request type the inference data needed.
  • the cloud reasoning server is a cloud reasoning server remotely connected to the self-driving vehicle for performing the reasoning process, and the cloud reasoning server may be a single server or a server group composed of multiple servers.
  • the cloud inference server is connected to each autonomous vehicle through wireless communication, and can receive inference requests sent by each autonomous vehicle in real time through wireless communication, including request types and/or data packets.
  • An inference request is an instruction issued by an autonomous vehicle to request the cloud inference server to perform an inference process, including the request type and/or data packet.
  • the self-driving vehicle applies for registration of a new inference engine service
  • the self-driving vehicle selects the deep learning model that it wants to infer
  • the self-driving vehicle gives the input data that needs to be inferred
  • the self-driving vehicle requests The output of the cloud inference server after inference and the self-driving vehicle need to obtain the last inference time, etc.
  • the data packet is the inference data required for inference according to the request type, and the type is not specifically limited, for example, it includes road condition image information and road condition point cloud information.
  • Step 504 generate a corresponding reasoning result according to the reasoning request, and send the reasoning result to the self-driving vehicle; the reasoning result is generated by the cloud reasoning server through a deep learning model according to the reasoning request Instructions for controlling the driving of the autonomous vehicle.
  • the cloud reasoning server After the cloud reasoning server receives the reasoning request sent by the self-driving vehicle, it performs corresponding processing according to the content of the reasoning request according to the preset program, generates the reasoning result, and sends the reasoning result to the corresponding self-driving vehicle in real time, so that Each self-driving vehicle can perform corresponding actions in real time according to the inference results.
  • the cloud reasoning server processes the data sent by the self-driving vehicle for predicting the trajectory of the self-driving vehicle according to the type of deep learning model selected by the self-driving vehicle, obtains the corresponding trajectory information of the self-driving vehicle, and sends it to The corresponding self-driving vehicle enables the self-driving vehicle to drive safely and smoothly.
  • the inference result sent by the cloud inference server is an indispensable command for the self-driving vehicle to drive automatically.
  • the control system of the self-driving vehicle performs corresponding actions according to the received inference result to realize the safe driving of the self-driving vehicle.
  • the inference request sent by the self-driving vehicle is received through the cloud reasoning server, the corresponding inference result is generated according to the reasoning request, and the reasoning result is sent to the self-driving vehicle so that the self-driving vehicle can execute the corresponding inference result according to the reasoning result. driving action. Since the core algorithm of reasoning processing is arranged on the cloud reasoning server wirelessly connected with the self-driving vehicle, the core algorithm is prevented from being leaked.
  • the generating a corresponding reasoning result according to the reasoning request, and sending the reasoning result to the self-driving vehicle includes:
  • the cloud reasoning server when the cloud reasoning server receives the reasoning request sent by the self-driving vehicle, it needs to analyze the reasoning request, and select the corresponding deep learning model according to the specific content of the reasoning request. After the deep learning model is selected, the data packet is processed according to the deep learning model. Usually, the data packet is input into the corresponding deep learning model to obtain the generated inference result. Finally, the inference result is sent to the self-driving vehicle, so that the self-driving vehicle executes the corresponding driving action instruction according to the content of the inference result.
  • the cloud reasoning server analyzes the reasoning request, selects the corresponding deep learning model according to the reasoning request, and performs reasoning processing on the data packet according to the deep learning model, obtains the reasoning result, and finally sends the reasoning result to the self-driving vehicle.
  • the processing of inference requests by the cloud inference server avoids the risk of the core algorithm being leaked due to the core algorithm being set on the vehicle computing platform.
  • the parsing the reasoning request, and selecting the corresponding deep learning model according to the reasoning request includes:
  • the required deep learning model has been selected in the inference request, select the selected deep learning model in the inference request as the deep learning model used during inference processing.
  • the cloud reasoning platform selects a deep learning model, it first analyzes the reasoning request, and judges whether the reasoning request contains the relevant content of the deep learning model selected by the autonomous driving platform; if the required deep learning model has been selected in the reasoning request, Or the deep learning model selected by the self-driving vehicle can be determined according to the content of the reasoning request, then select the selected or determinable deep learning model in the reasoning request as the deep learning model used in the reasoning process as the depth used in the reasoning process learning model.
  • the reasoning request by analyzing the reasoning request, it is judged whether the reasoning request contains the relevant content of the deep learning model selected by the automatic driving platform, so that the reasoning request used for the reasoning process can be quickly selected, and the reasoning process of the reasoning request is improved. s efficiency.
  • the parsing the reasoning request, and selecting the corresponding deep learning model according to the reasoning request further includes:
  • the reasoning purpose is obtained according to the reasoning request.
  • the reasoning purpose is obtained according to the reasoning request, and the reasoning purpose refers to the self-driving vehicle The purpose to be achieved by sending an inference request. After obtaining the reasoning purpose, select the deep learning model with the highest prediction accuracy that can achieve the reasoning purpose in the cloud reasoning server.
  • the required deep learning model is not selected in the reasoning request, or the deep learning model selected by the self-driving vehicle cannot be determined according to the content of the reasoning request, the one with the highest prediction accuracy that can achieve the purpose of reasoning is selected.
  • the deep learning model improves the accuracy of inference processing for inference requests.
  • the autonomous vehicle before receiving the reasoning request sent by the autonomous vehicle, it also includes:
  • the registered reasoning engine service request is the offer information sent by the self-driving vehicle for applying to the cloud reasoning server to perform reasoning processing on the reasoning request sent by it;
  • the cloud inference server before the cloud inference server receives and processes the inference request sent by the self-driving vehicle, the cloud inference server also receives the registered inference engine service request sent by the self-driving vehicle; the registered inference engine service request is sent by the self-driving vehicle for application The offer information for the inference processing of the inference request sent by the cloud inference server. And determine whether the self-driving vehicle meets the preset registration conditions, and when the self-driving vehicle meets the preset registration conditions, agree to the registration reasoning engine service request of the self-driving vehicle, and send a return message agreeing to the registration to the self-driving vehicle. Otherwise, send a return message denying registration to the self-driving vehicle.
  • the cloud inference server before the cloud inference server receives the inference request sent by the self-driving vehicle and processes it, it also receives the registration inference engine service request sent by the self-driving vehicle, and judges whether the self-driving vehicle meets the preset registration conditions.
  • the self-driving vehicle satisfies the preset registration conditions, agrees to the registration reasoning engine service request of the self-driving vehicle, and sends a return message agreeing to the registration to the self-driving vehicle.
  • a method for automatic driving of a vehicle is provided, and the application of the method to an automatic driving vehicle is used as an example for illustration, including the following steps:
  • Step 602 Generate and send an inference request to the cloud inference server; the inference request includes a request type and/or a data packet; the request type is a type that requires the cloud inference server to perform inference services; the data packet is processed according to the request type Inference data required for inference.
  • the self-driving vehicle generates an inference request when needed, and sends the inference request to the cloud inference server through the wireless communication device set by the self-driving vehicle.
  • An inference request is an instruction issued by an autonomous vehicle to request the cloud inference server to perform an inference process, including the request type and/or data packet.
  • the self-driving vehicle applies for registration of a new inference engine service
  • the self-driving vehicle selects the deep learning model that it wants to infer
  • the self-driving vehicle gives the input data that needs to be inferred
  • the self-driving vehicle requests The output of the cloud inference server after inference and the self-driving vehicle need to obtain the last inference time, etc.
  • the data packet is the inference data required for inference according to the request type, and the type is not specifically limited, for example, it includes road condition image information and road condition point cloud information.
  • Step 604 receiving the inference result sent by the cloud inference server, and executing the corresponding driving instruction according to the inference result; the inference result is generated by the cloud inference server through a deep learning model according to the inference request and used to control the automatic Instructions to drive the vehicle.
  • the cloud inference server processes the inference request, it will generate a corresponding inference result and send it to the autonomous vehicle.
  • the reasoning result is an instruction for the self-driving vehicle generated by the cloud reasoning server through the deep learning model according to the reasoning request.
  • the reasoning request is different, and the corresponding reasoning result is also different.
  • the self-driving vehicle receives the inference result sent by the cloud inference server, it executes the corresponding driving instruction according to the inference result.
  • the cloud reasoning server processes the data sent by the self-driving vehicle for predicting the trajectory of the self-driving vehicle according to the type of deep learning model selected by the self-driving vehicle, obtains the corresponding trajectory information of the self-driving vehicle, and sends it to The corresponding self-driving vehicle enables the self-driving vehicle to drive safely and smoothly.
  • the self-driving vehicle sends an inference request to the cloud reasoning server, and the cloud reasoning server receives the reasoning request sent by the self-driving vehicle, generates a corresponding reasoning result according to the reasoning request, and sends the reasoning result to the self-driving vehicle for further processing. It enables the self-driving vehicle to perform corresponding driving actions according to the inference results. Since the core algorithm of reasoning processing is arranged on the cloud reasoning server wirelessly connected with the self-driving vehicle, the core algorithm is prevented from being leaked.
  • the generating and sending the reasoning request to the cloud reasoning server includes:
  • the self-driving vehicle when it generates an inference request, it includes receiving an artificially generated inference request instruction, and generating a corresponding inference request according to the artificially generated inference request instruction.
  • a vehicle-mounted intelligent processing terminal is installed in a self-driving vehicle, and when the vehicle-mounted intelligent processing terminal of the self-driving vehicle receives an instruction to manually generate an inference request and/or the operating state of the self-driving vehicle meets the conditions for generating a reasoning request, a corresponding reasoning request.
  • the self-driving vehicle generates a corresponding reasoning request, for example, when the vehicle-mounted intelligent processing terminal meets the reasoning request generation conditions when the running state of the self-driving vehicle, if the vehicle is in an abnormal state
  • the self-driving vehicle sends the corresponding inference request to the cloud inference server for the cloud inference server to process the inference request and generate an inference result.
  • the vehicle-mounted intelligent processing terminal of the self-driving vehicle when the vehicle-mounted intelligent processing terminal of the self-driving vehicle receives an instruction to manually generate an inference request and/or the running state of the self-driving vehicle meets the conditions for generating an inference request, it generates a corresponding inference request and sends the inference request to
  • the cloud inference server performs processing, which realizes the process of the cloud inference server processing inference requests, and avoids the risk of the core algorithm of the computing platform being leaked.
  • the receiving the inference result sent by the cloud inference server, and executing the corresponding driving instruction according to the inference result includes:
  • the driving of the automatic driving vehicle is controlled according to the driving instruction.
  • the self-driving vehicle receives the inference result sent by the cloud inference server, it analyzes the inference result through the vehicle-mounted intelligent processing terminal, generates a driving instruction corresponding to the inference result; and sends the driving instruction to the vehicle driving control device, and finally The driving of the self-driving vehicle is controlled by the on-board driving control device according to the driving instructions.
  • the on-vehicle driving control device is used to control the driving of the self-driving vehicle according to the inference result generated by the cloud inference server, the on-board driving control device is connected to the on-board intelligent processing terminal, and can receive the inference sent by the cloud vehicle server through the on-vehicle intelligent processing terminal As a result, the driving state of the self-driving vehicle is controlled according to the inference result, and the corresponding control action is executed.
  • the self-driving vehicle receives the inference result sent by the cloud reasoning server, analyzes and processes the reasoning result, generates a driving instruction, and finally controls the driving of the self-driving vehicle according to the driving instruction, realizing the reasoning through the cloud reasoning server.
  • Inference processing of requests improves the processing speed of inference requests.
  • the generating and sending the reasoning request to the cloud reasoning server also includes:
  • the registered inference engine service request is offer information sent by the self-driving vehicle for applying to the cloud inference server to perform inference processing on the inference request sent by it;
  • the return information includes the return information of agreeing to register and the return information of rejecting registration.
  • the self-driving vehicle in order for the self-driving vehicle to perform inference processing on its inference requests through the cloud inference server, before the cloud inference server receives and processes the inference request sent by the self-driving vehicle, the self-driving vehicle needs to send a registered inference engine service request to the cloud inference Server: Registering the inference engine service request is the offer information sent by the self-driving vehicle to apply for the cloud inference server to perform inference processing on the inference request sent by it.
  • the cloud inference server needs to judge whether the self-driving vehicle meets the preset registration conditions, and when the self-driving vehicle meets the preset registration conditions, agree to the registration inference engine service request of the self-driving vehicle, and send a return message agreeing to the registration to the self-driving vehicle. Otherwise, send a return message denying registration to the self-driving vehicle.
  • the self-driving vehicle before the self-driving vehicle sends an inference request to the cloud reasoning server and requests the cloud reasoning server to process it, the self-driving vehicle also needs to send a registration reasoning engine service request to the cloud reasoning server, and determine whether the self-driving vehicle meets the requirements.
  • the registration conditions are set.
  • the cloud reasoning server agrees to the registration reasoning engine service request of the self-driving vehicle, and sends the return information agreeing to register to the self-driving vehicle, which improves the security of the cloud reasoning server.
  • the cloud inference server before sending the inference request to the cloud inference server, it also includes:
  • the data package is obtained by the vehicle-mounted road surface information acquisition device, and the data package includes road condition image information and road condition point cloud information.
  • the self-driving vehicle is provided with a vehicle-mounted road surface information acquisition device for acquiring data packets, including road condition image information and road condition point cloud information.
  • the vehicle-mounted road surface information acquisition device specifically includes: a vehicle-mounted road surface image acquisition device and a vehicle-mounted radar system.
  • the vehicle-mounted road surface image acquisition device is used to collect road condition image information, and send the road condition image information to the vehicle-mounted intelligent processing terminal;
  • the road condition image information is the road condition information around the self-driving vehicle, mainly including surrounding road surface information, surrounding vehicles Information and surrounding obstacles, pedestrian information, etc.
  • the vehicle-mounted radar system is used to collect road condition point cloud information and send the road condition point cloud information to the vehicle-mounted intelligent processing terminal; the vehicle-mounted radar system can not only detect road condition information in the surrounding range, but also detect Road condition information is conducive to the safe driving of self-driving vehicles.
  • the vehicle-mounted road surface information acquisition device acquires a data packet, the data packet includes road condition image information and road condition point cloud information, so that the cloud inference server can perform inference based on the road condition image information and road condition point cloud information.
  • FIGS. 5-6 are displayed sequentially as indicated by the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in FIGS. 5-6 may include multiple steps or stages. These steps or stages are not necessarily executed at the same moment, but may be executed at different moments. The steps or stages The order of execution is not necessarily performed sequentially, but may be performed alternately or alternately with other steps or at least a part of steps or stages in other steps.
  • a vehicle automatic driving device including: an inference request receiving module 701 and an inference result generating module 702, wherein:
  • the reasoning request receiving module 701 is used to receive the reasoning request sent by the self-driving vehicle; the reasoning request includes a request type and/or a data packet; the request type is a type that requires a cloud reasoning server to perform reasoning services; the data package is Inference data required for inference based on request type;
  • the reasoning result generating module 702 is configured to generate a corresponding reasoning result according to the reasoning request, and send the reasoning result to the self-driving vehicle; the reasoning result is obtained by the cloud reasoning server through the deep learning model according to the reasoning request The generated instructions are used to control the driving of the self-driving vehicle.
  • the inference result generation module 702 is further configured to: analyze the inference request, select a corresponding deep learning model according to the inference request; perform inference processing on the data packet according to the deep learning model , obtaining an inference result; sending the inference result to the self-driving vehicle.
  • the inference result generating module 702 is further configured to: analyze the inference request, and judge whether the required deep learning model has been selected in the inference request; if the inference request has selected For the required deep learning model, select the deep learning model selected in the inference request as the deep learning model used during inference processing.
  • the reasoning result generating module 702 is further configured to: if the required deep learning model is not selected in the reasoning request, obtain the reasoning purpose according to the reasoning request; according to the content of the reasoning purpose, Select the deep learning model with the highest prediction accuracy that can achieve the reasoning purpose.
  • the inference request receiving module 701 is further configured to: receive the registered inference engine service request sent by the self-driving vehicle; the registered inference engine service request is sent by the self-driving vehicle to apply for cloud reasoning Offer information for reasoning processing of the inference request sent by the server; judging whether the self-driving vehicle meets the preset registration conditions, and agreeing to the registration reasoning of the self-driving vehicle if the self-driving vehicle meets the preset registration conditions Engine service request, sending return information agreeing to registration to the self-driving vehicle.
  • the above vehicle automatic driving system device generates and sends an inference request to the cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, receives the inference request sent by the self-driving vehicle through the cloud inference server, and generates a corresponding inference result according to the inference request , and send the reasoning result to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a vehicle automatic driving device including: an inference request generating module 801 and an inference result receiving module 802, wherein:
  • Inference request generating module 801 configured to generate and send an inference request to a cloud inference server; the inference request includes a request type and/or a data packet; the request type is a type that requires a cloud inference server to perform an inference service; the data packet It is the inference data required for inference based on the request type;
  • the reasoning result receiving module 802 is configured to receive the reasoning result sent by the cloud reasoning server, and execute the corresponding driving instruction according to the reasoning result; the reasoning result is generated by the cloud reasoning server through the deep learning model according to the reasoning request. Instructions for controlling the driving of the self-driving vehicle.
  • the inference request generation module 801 is further configured to: generate a corresponding inference request when receiving an inference request instruction and/or the operating state of the autonomous vehicle meets the inference request generation condition; The reasoning request is sent to the cloud reasoning server.
  • the reasoning result receiving module 802 is further configured to: receive the reasoning result sent by the cloud reasoning server; analyze and process the reasoning result to generate a driving instruction; control the driving instruction according to the driving instruction.
  • the inference request generating module 801 is further configured to: send a registered inference engine service request to the cloud inference server; the registered inference engine service request is sent by an autonomous vehicle to apply for a cloud inference server Offer information for reasoning processing of the reasoning request sent by it; receiving the return information sent by the cloud reasoning server; the return information includes the return information of agreeing to register and the return information of rejecting registration.
  • the reasoning request generating module 801 is further configured to: acquire a data packet through a vehicle-mounted road surface information acquisition device, where the data packet includes road condition image information and road condition point cloud information.
  • the vehicle automatic driving system device further includes: a vehicle-mounted road surface information acquisition device, a vehicle-mounted driving control device, and a vehicle-mounted intelligent processing terminal;
  • the vehicle-mounted road surface information acquisition device is used to acquire data packets, the data packets include road condition image information and road condition point cloud information;
  • the vehicle-mounted intelligent processing terminal is set in the inference request generation module, and is used to generate an inference request, send the inference request to a cloud inference server, and receive an inference result generated by the cloud inference server;
  • the on-vehicle driving control device is connected to the inference result receiving module, and is used to control the driving of the self-driving vehicle according to the inference result generated by the cloud inference server
  • the vehicle-mounted road surface information acquisition device is used to acquire data packets, the data packets include road condition image information and road condition point cloud information;
  • the vehicle-mounted intelligent processing terminal is used to generate an inference request, send the inference request to a cloud inference server, and receive an inference result generated by the cloud inference server;
  • the on-vehicle driving control device is used to control the driving of the self-driving vehicle according to the inference result generated by the cloud inference server.
  • the self-driving vehicle includes a vehicle-mounted road surface information acquisition device, a vehicle-mounted driving control device, and a vehicle-mounted intelligent processing terminal.
  • the vehicle-mounted road surface information acquisition device is used to acquire data packets, and the data packets include road condition image information and road condition point cloud information.
  • the data packets not only include road condition image information and road condition point cloud information, but also different inference request data packets, for example, also include vehicle driving status information, etc.
  • the vehicle-mounted intelligent processing terminal is used to generate inference requests, send the inference requests to the cloud inference server, and receive inference results generated by the cloud inference server.
  • Fig. 3 is an information processing flow chart of the vehicle-mounted intelligent processing terminal in the vehicle automatic driving system in an embodiment.
  • the information sent by the inference server includes human-operated instructions or preset instructions caused by interrupted operations; it can also send information to the cloud inference server according to the received instructions.
  • the on-vehicle driving control device is used to control the driving of the self-driving vehicle according to the inference result generated by the cloud inference server, the on-board driving control device is connected to the on-board intelligent processing terminal, and can receive the inference sent by the cloud vehicle server through the on-vehicle intelligent processing terminal As a result, the driving state of the self-driving vehicle is controlled according to the inference result, and the corresponding control action is executed.
  • the data packet is obtained by the vehicle-mounted road surface information acquisition device, and the inference request is generated by the vehicle-mounted intelligent processing terminal, and the inference request is sent to the cloud inference server, and the inference result generated by the cloud inference server is received.
  • the driving of the self-driving vehicle is also controlled by the on-vehicle driving control device according to the inference result generated by the cloud reasoning server, thereby realizing the generation of reasoning results by the cloud reasoning server and the control of the self-driving vehicle.
  • the vehicle-mounted road surface information acquisition device includes a vehicle-mounted road surface image acquisition device and a vehicle-mounted radar system;
  • the vehicle-mounted road surface image collection device is used to collect road condition image information, and send the road condition image information to the vehicle-mounted intelligent processing terminal;
  • the vehicle-mounted radar system is used to collect road condition point cloud information, and send the road condition point cloud information to the vehicle-mounted intelligent processing terminal.
  • the vehicle-mounted road surface information acquisition device includes a vehicle-mounted road surface image acquisition device and a vehicle-mounted radar system.
  • the vehicle-mounted road surface image acquisition device is used to collect road condition image information, and send the road condition image information to the vehicle-mounted intelligent processing terminal;
  • the road condition image information is the road condition information around the self-driving vehicle, mainly including surrounding road surface information, surrounding vehicles Information and surrounding obstacles, pedestrian information, etc.
  • the vehicle-mounted radar system is used to collect road condition point cloud information, and send the road condition point cloud information to the vehicle-mounted intelligent processing terminal;
  • the vehicle-mounted intelligent processing terminal can not only detect the road condition information in the surrounding range, but also detect the road condition information in the far area through the radar, which is conducive to the safe driving of autonomous vehicles.
  • the road condition image information is collected by the vehicle-mounted road surface image acquisition device, and the road condition image information is sent to the vehicle-mounted intelligent processing terminal; the vehicle-mounted radar system is used to collect road condition point cloud information, and the road condition point cloud information It is sent to the vehicle-mounted intelligent processing terminal, so that the cloud inference information can be inferred according to the road condition image information and the road condition point cloud information.
  • the vehicle-mounted intelligent processing terminal includes: a processor, configured to receive the road condition image information and road condition point cloud information, and generate an inference request; wireless communication between.
  • FIG. 4 is a structure diagram of an inference request in an automatic vehicle driving system in an embodiment.
  • an inference request includes a request type and/or a data packet; the request type requires a cloud inference server to perform an inference service type; the data packet is inference data required for inference according to the request type.
  • the vehicle-mounted intelligent processing terminal includes a processor and a communication device, the processor is connected to the vehicle-mounted road surface information acquisition device and the vehicle-mounted driving control device, and the processor is used to process information, such as processing information obtained by the road surface information acquisition device; the communication
  • the device is connected to the processor for communicating with the cloud reasoning server according to the instructions of the processor, for example, sending the reasoning request generated by the processor or the road surface information obtained by the vehicle-mounted road surface information acquisition device to the cloud reasoning server.
  • the processor is also connected to the vehicle-mounted travel control device, so that the vehicle-mounted travel control device can be controlled to execute corresponding travel instructions according to the inference result.
  • the processor in the vehicle-mounted intelligent processing terminal is connected to the vehicle-mounted road surface information acquisition device, the vehicle-mounted driving control device and the communication device, so as to realize the acquisition, processing and communication with the cloud reasoning server of road condition information.
  • the vehicle radar system includes at least one of a laser radar system, an ultrasonic radar system, a millimeter wave radar system and a microwave radar system.
  • the on-board radar system When the self-driving vehicle is driving at a high speed, it may not be possible to see the breaks or steps ahead in time due to weather or field of view.
  • the abnormal road conditions ahead can be scanned, and the abnormal road conditions can be sent to the on-board intelligent processing terminal of the self-driving vehicle in time, and the abnormal road information can be sent to the cloud reasoning server through the communication device, and then generated by reasoning of the cloud reasoning server Inference results are sent to autonomous vehicles.
  • autonomous vehicles obtain road anomaly information mainly by means of vehicle-mounted radar systems capable of detecting relatively long distances.
  • the types of vehicle-mounted radar systems include at least one of lidar systems, ultrasonic radar systems, and microwave radar systems.
  • the vehicle radar system is a laser radar system
  • the vehicle radar system generates road condition laser radar point cloud information.
  • the vehicle-mounted radar system is used to detect the abnormal road section in front of the self-driving vehicle, and the abnormal road surface information is sent to the cloud reasoning server, and the reasoning result is generated and sent to the self-driving vehicle for automatic driving through the processing of the cloud reasoning server
  • the corresponding actions are carried out on the vehicle, which improves the driving safety of the self-driving vehicle.
  • the communication device includes at least one of WIFI device, Bluetooth device, GPRS device, 3G, 4G and 5G modules connected to the processor.
  • an appropriate wireless communication device can be selected according to actual needs, and has wide applicability.
  • the wireless communication device of the vehicle-mounted intelligent processing terminal uses at least one of WIFI devices, Bluetooth devices, GPRS devices, 3G, 4G and 5G modules connected to the processor; Scenarios, such as self-driving vehicles driving automatically in relatively closed and narrow areas such as parks; GPRS devices, 3G, 4G and 5G modules are suitable for longer-distance communications, such as driving on intercity roads.
  • the self-driving vehicle and the cloud reasoning server are connected wirelessly through at least one of WIFI device, Bluetooth device, GPRS device, 3G, 4G and 5G modules, and the self-driving vehicle and the cloud reasoning server are realized.
  • the stable wireless communication between them avoids the risk of the core algorithm being leaked due to the storage of the core algorithm of the computing platform in the self-driving vehicle.
  • each module in the above-mentioned vehicle automatic driving device can be fully or partially realized by software, hardware and a combination thereof.
  • the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, and can also be stored in the memory of the computer device in the form of software, so that the processor can invoke and execute the corresponding operations of the above-mentioned modules.
  • the above vehicle automatic driving device generates and sends an inference request to a cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, receives the inference request sent by the self-driving vehicle through the cloud inference server, and generates a corresponding inference result according to the inference request, And send the reasoning result to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a computer device is provided.
  • the computer device may be a server, and its internal structure may be as shown in FIG. 9 .
  • the computer device includes a processor, memory and a network interface connected by a system bus. Wherein, the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, computer programs and databases.
  • the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
  • the network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for distributing video customer service is realized.
  • FIG. 9 is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation on the computer equipment on which the solution of this application is applied.
  • the specific computer equipment can be More or fewer components than shown in the figures may be included, or some components may be combined, or have a different arrangement of components.
  • a computer device including a memory and a processor, a computer program is stored in the memory, and the processor implements the following steps when executing the computer program:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires a cloud inference server to perform an inference service;
  • the data packet is required for inference according to the request type inference data;
  • the processor executes the computer program, the following steps are also implemented: parsing the reasoning request, selecting a corresponding deep learning model according to the reasoning request; performing reasoning processing on the data packet according to the deep learning model, Acquiring an inference result; sending the inference result to the self-driving vehicle.
  • the processor executes the computer program, the following steps are further implemented: parsing the reasoning request, and judging whether the required deep learning model has been selected in the reasoning request; if the required deep learning model has been selected in the reasoning request The desired deep learning model is selected, and the selected deep learning model in the inference request is selected as the deep learning model used in the inference process.
  • the processor executes the computer program, the following steps are also implemented: if the required deep learning model is not selected in the reasoning request, obtain the reasoning purpose according to the reasoning request; according to the content of the reasoning purpose, select The deep learning model with the highest prediction accuracy that can achieve the reasoning purpose.
  • the processor when the processor executes the computer program, it also implements the following steps: receiving the registered reasoning engine service request sent by the self-driving vehicle; the registered reasoning engine service request is sent by the self-driving vehicle to apply for a cloud reasoning server Offer information for performing inference processing on the inference request sent by it; judging whether the self-driving vehicle meets the preset registration conditions, and if the self-driving vehicle meets the preset registration conditions, agree to the registration inference engine of the self-driving vehicle Service request, sending return information agreeing to registration to the self-driving vehicle.
  • the above-mentioned computer device generates and sends an inference request to a cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, and receives the inference request sent by the self-driving vehicle through the cloud inference server, generates a corresponding inference result according to the inference request, and sends The reasoning result is sent to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a computer device including a memory and a processor, a computer program is stored in the memory, and the processor implements the following steps when executing the computer program:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform an inference service;
  • the data packet is generated when reasoning according to the request type Inference data needed;
  • the processor executes the computer program, the following steps are also implemented: when the instruction of the reasoning request is received and/or the operating state of the autonomous vehicle meets the generation conditions of the reasoning request, a corresponding reasoning request is generated; The request is sent to the cloud reasoning server.
  • the processor executes the computer program, the following steps are further implemented: receiving the inference result sent by the cloud inference server; analyzing and processing the inference result to generate a driving instruction; controlling the automatic driving instruction according to the driving instruction. driving the vehicle.
  • the processor executes the computer program, the following steps are also implemented: sending a registration reasoning engine service request to the cloud reasoning server; the registration reasoning engine service request is sent by the self-driving vehicle to apply for the cloud reasoning server
  • the reasoning request sent by it sends offer information for reasoning processing; receiving the return information sent by the cloud reasoning server; the return information includes the return information of agreeing to register and the return information of rejecting registration.
  • the vehicle-mounted road surface information acquisition device acquires a data package, the data package includes road condition image information and road condition point cloud information.
  • the above-mentioned computer device generates and sends an inference request to a cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, and receives the inference request sent by the self-driving vehicle through the cloud inference server, generates a corresponding inference result according to the inference request, and sends The reasoning result is sent to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires a cloud inference server to perform an inference service;
  • the data packet is required for inference according to the request type inference data;
  • the following steps are also implemented: parsing the reasoning request, selecting a corresponding deep learning model according to the reasoning request; performing reasoning processing on the data packet according to the deep learning model , obtaining an inference result; sending the inference result to the self-driving vehicle.
  • the following steps are further implemented: parsing the inference request, and judging whether the required deep learning model has been selected in the inference request; if the inference request has selected For the required deep learning model, select the deep learning model selected in the inference request as the deep learning model used during inference processing.
  • the following steps are also implemented: if the required deep learning model is not selected in the reasoning request, obtain the reasoning purpose according to the reasoning request; according to the content of the reasoning purpose, Select the deep learning model with the highest prediction accuracy that can achieve the reasoning purpose.
  • the following steps are also implemented: receiving the registration inference engine service request sent by the self-driving vehicle; the registration reasoning engine service request is sent by the self-driving vehicle to apply for cloud reasoning Offer information for reasoning processing of the inference request sent by the server; judging whether the self-driving vehicle meets the preset registration conditions, and agreeing to the registration reasoning of the self-driving vehicle if the self-driving vehicle meets the preset registration conditions Engine service request, sending return information agreeing to registration to the self-driving vehicle.
  • the above storage medium generates and sends an inference request to the cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, receives the inference request sent by the self-driving vehicle through the cloud inference server, generates a corresponding inference result according to the inference request, and sends The reasoning result is sent to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the inference request includes a request type and/or a data packet;
  • the request type is a type that requires the cloud inference server to perform an inference service;
  • the data packet is generated when reasoning according to the request type Inference data needed;
  • the following steps are further implemented: when the instruction of the inference request is received and/or the operating state of the autonomous vehicle meets the conditions for generating the inference request, generate a corresponding inference request;
  • the reasoning request is sent to the cloud reasoning server.
  • the following steps are further implemented: receiving the inference result sent by the cloud inference server; analyzing and processing the inference result to generate a driving instruction; controlling the driving instruction according to the driving instruction.
  • the movement of autonomous vehicles is further implemented: receiving the inference result sent by the cloud inference server; analyzing and processing the inference result to generate a driving instruction; controlling the driving instruction according to the driving instruction.
  • the following steps are also implemented: sending a registration reasoning engine service request to the cloud reasoning server; the registration reasoning engine service request is sent by the self-driving vehicle to apply for the cloud reasoning server Offer information for reasoning processing of the reasoning request sent by it; receiving the return information sent by the cloud reasoning server; the return information includes the return information of agreeing to register and the return information of rejecting registration.
  • the vehicle-mounted road surface information acquisition device acquires a data package, the data package includes road condition image information and road condition point cloud information.
  • the above storage medium generates and sends an inference request to the cloud inference server wirelessly connected to the self-driving vehicle through the self-driving vehicle, receives the inference request sent by the self-driving vehicle through the cloud inference server, generates a corresponding inference result according to the inference request, and sends The reasoning result is sent to the self-driving vehicle; finally, the self-driving vehicle receives the reasoning result sent by the cloud reasoning server, and executes the corresponding driving instruction according to the reasoning result. Since the core algorithm is placed on the cloud reasoning server wirelessly connected to the self-driving vehicle, the core algorithm is prevented from being leaked.
  • Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory or optical memory, etc.
  • Volatile memory can include Random Access Memory (RAM) or external cache memory.
  • RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Human Computer Interaction (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Traffic Control Systems (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)

Abstract

一种车辆自动驾驶系统、方法、装置、计算机设备和存储介质。该系统包括：云端推理服务器和自动驾驶车辆；自动驾驶车辆和云端推理服务器之间通过无线通信连接；自动驾驶车辆用于生成和发送推理请求至云端推理服务器，接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令；云端推理服务器用于接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；其中，推理结果是云端推理服务器根据推理请求通过深度学习模型生成的用于控制自动驾驶车辆行驶的指令。本系统能够提高运算平台核心算法的保护，避免核心算法被泄露的风险。

Description

车辆自动驾驶系统、方法、装置、计算机设备和存储介质 技术领域
本申请涉及自动驾驶技术领域，特别是涉及一种车辆自动驾驶系统、方法、装置、计算机设备和存储介质。
背景技术
在自动驾驶中，人工智能起到重要作用。深度学习是人工智能在自动驾驶领域最重要的解决方案之一，尤其在感知方面。深度学习推理引擎的设计目的就是要让深度学习模型推理的速度加快，深度学习模型推理由GPU或AI芯片完成。当前，自动驾驶的感知模块实现以深度学习方法为主，深度学习需要通过车载激光雷达和摄像头等传感器信息作为输入，通过神经元网络计算获得输出即感知结果，神经元网络计算的过程也叫做推理过程。
现有技术中，自动驾驶汽车公司的计算平台均部署在自动驾驶汽车上。计算平台的功率和自动驾驶的级别相关，对于低级别L1-L3级的自动驾驶，一块小小的计算芯片或者一个集成了小小芯片的摄像头既可以实现。但是对于L4级的自动驾驶车辆，由于所需要的深度学习模型更复杂，种类更多，运算量更大，此时仅仅靠一块小芯片是无法进行实时运算的。在此情形下，L4级自动驾驶汽车公司的解决方案是把高性能计算平台直接部署在车上，例如后备箱之中，或车顶之上。但是，现有技术因为大量核心算法被放置在自动驾驶车辆中的计算平台上，导致了运算平台的核心算法被泄露的风险。
发明内容
基于此，有必要针对上述技术问题，提供一种能够避免自动驾驶核心算法被泄露的车辆自动驾驶系统、方法、装置、计算机设备和存储介质。
一种车辆自动驾驶系统，所述系统包括：云端推理服务器和自动驾驶车辆；所述自动驾驶车辆和所述云端推理服务器之间通过无线通信连接；
所述自动驾驶车辆用于生成和发送推理请求至所述云端推理服务器，接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；
所述云端推理服务器用于接收所述自动驾驶车辆发送的推理请求，根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；
其中，所述推理结果是所述云端推理服务器根据所述推理请求生成的指令；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。
一种车辆自动驾驶方法，所述方法包括：接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
在其中一个实施例中，还包括：对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；
根据所述深度学习模型对数据包进行推理处理，获取推理结果；
将所述推理结果发送至所述自动驾驶车辆。
在其中一个实施例中，还包括：对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；
若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理处理时使用的深度学习模型。
在其中一个实施例中，还包括：若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；
根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
在其中一个实施例中，还包括：接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
一种车辆自动驾驶方法，所述方法包括：生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在其中一个实施例中，还包括：当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；
将所述推理请求发送至所述云端推理服务器。
在其中一个实施例中，还包括：接收所述云端推理服务器发送的推理结果；
对所述推理结果进行解析处理，生成驾驶指令；
根据所述驾驶指令控制所述自动驾驶车辆的行驶。
在其中一个实施例中，还包括：向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
在其中一个实施例中，还包括：通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
一种车辆自动驾驶装置，所述装置包括：
推理请求接收模块，用于接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
推理结果生成模块，用于根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
一种车辆自动驾驶装置，所述装置包括：
推理请求生成模块，用于生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
推理结果接收模块，用于接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在其中一个实施例中，所述车辆自动驾驶装置还包括：车载路面信息获取装置、车载行驶控制装置和车载智能处理终端；
所述车载路面信息获取装置用于获取数据包，所述数据包包括路况图像信息和路况点云信息；
所述车载智能处理终端设置于所述推理请求生成模块内，用于生成推理请求，并将推理请求发送至云端推理服务器，以及接收所述云端推理服务器生成的推理结果；
所述车载行驶控制装置与所述推理结果接收模块连接，用于根据所述云端推理服务器生成的推理结果 控制所述自动驾驶车辆的行驶。
在其中一个实施例中，所述车载路面信息获取装置包括车载路面图像采集装置和车载雷达系统；
所述车载路面图像采集装置用于采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；
所述车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端。
在其中一个实施例中，所述车载智能处理终端包括：
处理器，用于接收所述路况图像信息和路况点云信息，以及生成推理请求；
通信装置，用于实现车载智能处理和云端推理服务器之间的无线通信。
在其中一个实施例中，所述车载雷达系统包括激光雷达系统、超声波雷达系统、毫米波雷达系统和微波雷达系统中的至少一种。
在其中一个实施例中，所述通信装置包括连接所述处理器的WIFI装置、蓝牙装置、GPRS装置、3G、4G和5G模块中的的至少一种。
一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：
接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：
生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
上述车辆自动驾驶系统、方法、装置、计算机设备和存储介质，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
附图说明
图1为一个实施例中车辆自动驾驶系统结构图；
图2为一个实施例中车辆自动驾驶系统中云端推理服务器与自动驾驶车辆的信息交互图；
图3为一个实施例中车辆自动驾驶系统中车载智能处理终端的信息处理流程图；
图4为一个实施例中车辆自动驾驶系统中推理请求结构图；
图5为一个实施例中车辆自动驾驶方法流程示意图；
图6为另一个实施例中车辆自动驾驶方法流程示意图；
图7为一个实施例中车辆自动驾驶装置的结构框图；
图8为另一个实施例中车辆自动驾驶装置的结构框图；
图9为一个实施例中计算机设备的结构框图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
在一个实施例中，如图1所示，提供了一种车辆自动驾驶系统，包括：云端推理服务器和自动驾驶车辆；所述自动驾驶车辆和所述云端推理服务器之间通过无线通信连接；
具体地，自动驾驶车辆的类型并不唯一，具体可以为轿车、面包车、卡车等。云端推理服务器是用于执行神经元网络计算获得输出即感知结果的服务器设备，神经元网络计算的过程也叫做推理过程。车辆自动驾驶系统包括云端推理服务器和自动驾驶车辆；其中，每辆自动驾驶车辆均与云端推理服务器无线通信连接，无线通信连接的方法不作具体限定，例如可通过5G通信连接。
所述自动驾驶车辆用于生成和发送推理请求至所述云端推理服务器，接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；
具体地，推理请求是自动驾驶车辆发出的请求云端推理服务器执行推理过程的指令，包括请求类型和/或数据包。其中，请求类型不少于一种，例如：自动驾驶车辆申请注册新的推理引擎服务、自动驾驶车辆选取想要推理的深度学习模型、自动驾驶车辆给出需要推理的输入数据、自动驾驶车辆请求云端推理服务器推理之后的输出结果和自动驾驶车辆需要获得上一次的推理用时等。云端推理服务器发送的推理结果是自动驾驶车辆进行自动驾驶时必不可少的指令，自动驾驶车辆的控制系统根据接收的推理结果执行对应的动作，实现自动驾驶车辆的安全行驶。
所述云端推理服务器用于接收所述自动驾驶车辆发送的推理请求，根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；
具体地，云端推理服务器是与自动驾驶车辆远程连接的用于执行推理过程的服务器，云端推理服务器可以是一台服务器，也可以是多台服务器组成的服务器群组。图2为车辆自动驾驶系统中云端推理服务器结构图，如图2所示，云端推理服务器与各自动驾驶车辆通过无线通信连接，能够实时接收各自动驾驶车辆发送的推理请求，包括请求类型和/或数据包，并根据推理请求的内容按照预设程序进行对应的处理，生成推理结果，并实时将推理结果发送至对应的自动驾驶车辆，以使得各自动驾驶车辆能够实时根据推理结果执行对应的动作。例如，云端推理服务器根据自动驾驶车辆选择的深度学习模型类型对自动驾驶车辆发送的用于预判自动驾驶车辆行驶轨迹的数据进行处理，获取对应的自动驾驶车辆行驶的轨迹路线信息，并发送至对应的自动驾驶车辆以使得自动驾驶车辆能够安全平稳的行驶。
其中，所述推理结果是所述云端推理服务器根据所述推理请求生成的指令；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。
具体地，推理结果是云端推理服务器根据推理请求生成的自动驾驶车辆所需要的指令和信息，推理请求不同，对应的推理结果也不同。例如，若推理请求是自动驾驶车辆申请注册新的推理引擎服务，则推理结果即是对应的自动驾驶车辆注册成功或失败；若推理请求是自动驾驶车辆选取想要推理的深度学习模型，推理结果是自动驾驶车辆所选择的深度学习模型；若推理请求是自动驾驶车辆请求云端推理服务器推理之后的输出结果，则对应的推理结果是返回的推理结果；若推理请求是自动驾驶车辆需要获得上一次的推理用时，则推理结果是上一次推理过程所耗费的时间。
上述车辆自动驾驶系统，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推 理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，如图5所示，提供了一种车辆自动驾驶方法，以该方法应用于云端推理服务器为例进行说明，包括以下步骤：
步骤502，接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。
具体地，云端推理服务器是与自动驾驶车辆远程连接的用于执行推理过程的云端推理服务器，云端推理服务器可以是一台服务器，也可以是多台服务器组成的服务器群组。云端推理服务器与各自动驾驶车辆通过无线通信连接，能够通过无线通信方式实时接收各自动驾驶车辆发送的推理请求，包括请求类型和/或数据包。
推理请求是自动驾驶车辆发出的请求云端推理服务器执行推理过程的指令，包括请求类型和/或数据包。其中，请求类型不少于一种，例如：自动驾驶车辆申请注册新的推理引擎服务、自动驾驶车辆选取想要推理的深度学习模型、自动驾驶车辆给出需要推理的输入数据、自动驾驶车辆请求云端推理服务器推理之后的输出结果和自动驾驶车辆需要获得上一次的推理用时等。数据包是根据请求类型进行推理时所需要的推理数据，类型不作具体限定，例如包括路况图像信息和路况点云信息。
步骤504，根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
具体地，云端推理服务器接收到自动驾驶车辆发送的推理请求以后，根据推理请求的内容按照预设程序进行对应的处理，生成推理结果，并实时将推理结果发送至对应的自动驾驶车辆，以使得各自动驾驶车辆能够实时根据推理结果执行对应的动作。例如，云端推理服务器根据自动驾驶车辆选择的深度学习模型类型对自动驾驶车辆发送的用于预判自动驾驶车辆行驶轨迹的数据进行处理，获取对应的自动驾驶车辆行驶的轨迹路线信息，并发送至对应的自动驾驶车辆以使得自动驾驶车辆能够安全平稳的行驶。
云端推理服务器发送的推理结果是自动驾驶车辆进行自动驾驶时必不可少的指令，自动驾驶车辆的控制系统根据接收的推理结果执行对应的动作，实现自动驾驶车辆的安全行驶。
上述车辆自动驾驶方法中，通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆以使得自动驾驶车辆能够根据推理结果执行对应的驾驶动作。由于推理处理的核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，所述根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆包括：
对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；
根据所述深度学习模型对数据包进行推理处理，获取推理结果；
将所述推理结果发送至所述自动驾驶车辆。
具体地，当云端推理服务器接收到自动驾驶车辆发送的推理请求以后，需要对推理请求进行解析，根据推理请求的具体内容选择对应的深度学习模型。选择深度学习模型以后，根据深度学习模型对数据包进行处理，通常是将数据包输入至对应的深度学习模型内，获取生成的推理结果。最后将推理结果发送至自动驾驶车辆，以使得自动驾驶车辆根据推理结果的内容执行对应的驾驶动作指令。
本实施例中，云端推理服务器对推理请求进行解析，根据推理请求选择对应的深度学习模型，并根据深度学习模型对数据包进行推理处理，获取推理结果，最后将推理结果发送至自动驾驶车辆。通过云端推理服务器对推理请求的处理，避免了由于核心算法被设置于车载运算平台导致的核心算法被泄露的风险。
在一个实施例中，所述对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型包括：
对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；
若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理 处理时使用的深度学习模型。
具体地，云端推理平台在选择深度学习模型时，首先对推理请求进行解析，判断推理请求中是否包含自动驾驶平台选择深度学习模型的相关内容；如果推理请求中已选择所需的深度学习模型，或者根据推理请求的内容能够确定自动驾驶车辆所选择的深度学习模型，则选择推理请求中已选择的或能够确定的深度学习模型作为推理处理时使用的深度学习模型作为进行推理处理时使用的深度学习模型。
本实施例中，通过对推理请求进行解析，判断推理请求中是否包含自动驾驶平台选择深度学习模型的相关内容，从而能够快速选择出推理处理所使用的推理请求，提高了对推理请求进行推理处理的效率。
在一个实施例中，所述对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型还包括：
若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；
根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
具体地，如果推理请求中没有选择所需的深度学习模型，或者根据推理请求的内容不能够确定自动驾驶车辆所选择的深度学习模型，则根据推理请求获取推理目的，推理目的是指自动驾驶车辆发送推理请求所欲实现的目的。获取推理目的以后，在云端推理服务器中选择能够实现该推理目的的预测准确率最高的深度学习模型。
本实施例中，通过在推理请求中没有选择所需的深度学习模型，或者根据推理请求的内容不能够确定自动驾驶车辆所选择的深度学习模型时，选择能够实现推理目的的预测准确率最高的深度学习模型作为推理处理的深度学习模型，提高了对推理请求进行推理处理的准确率。
在一个实施例中，所述接收自动驾驶车辆发送的推理请求，之前还包括：
接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
具体地，在云端推理服务器接收自动驾驶车辆发送的推理请求并进行处理之前，云端推理服务器还接收自动驾驶车辆发送的注册推理引擎服务请求；注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息。并判断自动驾驶车辆是否满足预设的注册条件，在自动驾驶车辆满足预设的注册条件，同意自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至自动驾驶车辆。否则，发送拒绝注册的返回信息至自动驾驶车辆。
本实施例中，通过在云端推理服务器接收自动驾驶车辆发送的推理请求并进行处理之前，还接收自动驾驶车辆发送的注册推理引擎服务请求，并判断自动驾驶车辆是否满足预设的注册条件，在自动驾驶车辆满足预设的注册条件，同意自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至自动驾驶车辆。提高了云端推理服务器的安全性。
在一个实施例中，如图6所示，提供了一种车辆自动驾驶方法，以该方法应用于自动驾驶车辆为例进行说明，包括以下步骤：
步骤602，生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。
具体地，自动驾驶车辆在需要时生成推理请求，通过自动驾驶车辆设置的无线通信装置将推理请求发送至云端推理服务器。推理请求是自动驾驶车辆发出的请求云端推理服务器执行推理过程的指令，包括请求类型和/或数据包。其中，请求类型不少于一种，例如：自动驾驶车辆申请注册新的推理引擎服务、自动驾驶车辆选取想要推理的深度学习模型、自动驾驶车辆给出需要推理的输入数据、自动驾驶车辆请求云端推理服务器推理之后的输出结果和自动驾驶车辆需要获得上一次的推理用时等。数据包是根据请求类型进行推理时所需要的推理数据，类型不作具体限定，例如包括路况图像信息和路况点云信息。
步骤604，接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
具体地，云端推理服务器在对推理请求进行处理以后，会生成对应的推理结果并发送至自动驾驶车辆。 推理结果是云端推理服务器根据推理请求通过深度学习模型生成的供自动驾驶车辆行驶的指令，推理请求不同，对应的推理结果也不同。自动驾驶车辆接收到云端推理服务器发送的推理结果以后，根据推理结果执行对应驾驶指令。例如，云端推理服务器根据自动驾驶车辆选择的深度学习模型类型对自动驾驶车辆发送的用于预判自动驾驶车辆行驶轨迹的数据进行处理，获取对应的自动驾驶车辆行驶的轨迹路线信息，并发送至对应的自动驾驶车辆以使得自动驾驶车辆能够安全平稳的行驶。
上述车辆自动驾驶方法中，自动驾驶车辆将推理请求发送至云端推理服务器，云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆以使得自动驾驶车辆能够根据推理结果执行对应的驾驶动作。由于推理处理的核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，所述生成并发送推理请求至云端推理服务器包括：
当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；
将所述推理请求发送至所述云端推理服务器。
具体地，自动驾驶车辆生成推理请求时，包括接收人工生成的推理请求指令，根据人工生成的推理请求指令生成对应的推理请求。例如，自动驾驶车辆中安装有车载智能处理终端，当自动驾驶车辆的车载智能处理终端接收到人工生成推理请求的指令和/或所述自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求。
和/或自动驾驶车辆的运行状态符合推理请求生成条件时，自动驾驶车辆生成对应的推理请求，例如车载智能处理终端在自动驾驶车辆的运行状态符合推理请求生成条件时，如车辆出现运行异常状态时，自动驾驶车辆将对应的推理请求发送至云端推理服务器，以供云端推理服务器对推理请求进行处理，生成推理结果。
本实施例中，当自动驾驶车辆的车载智能处理终端接收到人工生成推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求，并将推理请求发送至云端推理服务器进行处理，实现了云端推理服务器对推理请求进行处理的过程，避免了运算平台核心算法被泄露的风险。
在一个实施例中，所述接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令包括：
接收所述云端推理服务器发送的推理结果；
对所述推理结果进行解析处理，生成驾驶指令；
根据所述驾驶指令控制所述自动驾驶车辆的行驶。
具体地，自动驾驶车辆接收所述云端推理服务器发送的推理结果以后，通过车载智能处理终端对推理结果进行解析，生成与推理结果对应的驾驶指令；并将驾驶指令发送至车载行驶控制装置，最后通过车载行驶控制装置根据驾驶指令控制自动驾驶车辆的行驶。车载行驶控制装置用于根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶，车载行驶控制装置与车载智能处理终端相连接，能够通过车载智能处理终端接收云端车辆服务器发送的推理结果，并根据推理结果对自动驾驶车辆的行驶状态进行控制，执行对应的控制动作。
本实施例中，自动驾驶车辆接收云端推理服务器发送的推理结果，并对推理结果进行解析处理，生成驾驶指令，最后根据驾驶指令控制所述自动驾驶车辆的行驶，实现了通过云端推理服务器对推理请求的推理处理，提高了推理请求的处理速度。
在一个实施例中，所述生成并发送推理请求至云端推理服务器，之前还包括：
向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
具体地，自动驾驶车辆为了能够通过云端推理服务器对其推理请求进行推理处理，在云端推理服务器接收自动驾驶车辆发送的推理请求并进行处理之前，自动驾驶车辆需要发送注册推理引擎服务请求至云端推理服务器；注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求 进行推理处理的要约信息。云端推理服务器需要判断自动驾驶车辆是否满足预设的注册条件，在自动驾驶车辆满足预设的注册条件，同意自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至自动驾驶车辆。否则，发送拒绝注册的返回信息至自动驾驶车辆。
本实施例中，自动驾驶车辆向云端推理服务器发送推理请求并请求云端推理服务器对之进行处理之前，还需要自动驾驶车辆发送注册推理引擎服务请求至云端推理服务器，并判断自动驾驶车辆是否满足预设的注册条件，在自动驾驶车辆满足预设的注册条件，云端推理服务器同意自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至自动驾驶车辆，提高了云端推理服务器的安全性。
在一个实施例中，所述将所述推理请求发送至所述云端推理服务器，之前还包括：
通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
具体地，自动驾驶车辆中设置有车载路面信息获取装置，用于获取数据包，包括路况图像信息和路况点云信息。车载路面信息获取装置具体包括：车载路面图像采集装置和车载雷达系统。其中，车载路面图像采集装置用于采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；路况图像信息是自动驾驶车辆周围的路况信息，主要包括周围路面信息，周围车辆信息和周围障碍物、行人信息等。车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端；车载雷达系统不仅仅能够探测周围范围内的路况信息，还能够通过雷达探测较远区域内的路况信息，有利于自动驾驶车辆的安全行驶。
本实施例中，通过车载路面信息获取装置获取数据包，数据包包括路况图像信息和路况点云信息，使得云端推理服务器能够根据路况图像信息和路况点云信息进行推理。
应该理解的是，虽然图5-图6的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，图5-图6中的至少一部分步骤可以包括多个步骤或者多个阶段，这些步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤中的步骤或者阶段的至少一部分轮流或者交替地执行。
在一个实施例中，如图7所示，提供了一种车辆自动驾驶装置，包括：推理请求接收模块701和推理结果生成模块702，其中：
推理请求接收模块701，用于接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
推理结果生成模块702，用于根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，所述推理结果生成模块702，还用于：对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；根据所述深度学习模型对数据包进行推理处理，获取推理结果；将所述推理结果发送至所述自动驾驶车辆。
在一个实施例中，所述推理结果生成模块702，还用于：对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理处理时使用的深度学习模型。
在一个实施例中，所述推理结果生成模块702，还用于：若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
在一个实施例中，所述推理请求接收模块701，还用于：接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
上述车辆自动驾驶系统装置，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，如图8所示，提供了一种车辆自动驾驶装置，包括：推理请求生成模块801和推理结果接收模块802，其中：
推理请求生成模块801，用于生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
推理结果接收模块802，用于接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，所述推理请求生成模块801，还用于：当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；将所述推理请求发送至所述云端推理服务器。
在一个实施例中，所述推理结果接收模块802，还用于：接收所述云端推理服务器发送的推理结果；对所述推理结果进行解析处理，生成驾驶指令；根据所述驾驶指令控制所述自动驾驶车辆的行驶。
在一个实施例中，所述推理请求生成模块801，还用于：向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
在一个实施例中，所述推理请求生成模块801，还用于：通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
在一个实施例中，所述车辆自动驾驶系统装置还包括：车载路面信息获取装置、车载行驶控制装置和车载智能处理终端；
所述车载路面信息获取装置用于获取数据包，所述数据包包括路况图像信息和路况点云信息；
所述车载智能处理终端设置于所述推理请求生成模块内，用于生成推理请求，并将推理请求发送至云端推理服务器，以及接收所述云端推理服务器生成的推理结果；
所述车载行驶控制装置与所述推理结果接收模块连接，用于根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶
所述车载路面信息获取装置用于获取数据包，所述数据包包括路况图像信息和路况点云信息；
所述车载智能处理终端用于生成推理请求，并将推理请求发送至云端推理服务器，以及接收所述云端推理服务器生成的推理结果；
所述车载行驶控制装置用于根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶。
具体地，自动驾驶车辆包括车载路面信息获取装置、车载行驶控制装置和车载智能处理终端。其中，车载路面信息获取装置用于获取数据包，所述数据包包括路况图像信息和路况点云信息。数据包不仅仅包括路况图像信息和路况点云信息，不同的推理请求数据包也不同，例如还包括车辆的行驶状态信息等。车载智能处理终端用于生成推理请求，并将推理请求发送至云端推理服务器，以及接收云端推理服务器生成的推理结果。
图3为一个实施例中车辆自动驾驶系统中车载智能处理终端的信息处理流程图，如图3所示，车载智能处理终端是安装与车辆内部的能够操作的终端，能够接收外部的指令和云端推理服务器发送的信息，包括人为操作的指令或预设的由于中断操作所引发的指令；还能够根据接收的指令发送信息至云端推理服务器。车载行驶控制装置用于根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶，车载行驶控制装置与车载智能处理终端相连接，能够通过车载智能处理终端接收云端车辆服务器发送的推理结果，并根据推理结果对自动驾驶车辆的行驶状态进行控制，执行对应的控制动作。
本实施例中，通过车载路面信息获取装置获取数据包，并通过车载智能处理终端生成推理请求，并将 推理请求发送至云端推理服务器，以及接收所述云端推理服务器生成的推理结果。还通过车载行驶控制装置根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶，实现了通过云端推理服务器生成推理结果以及对自动驾驶车辆进行控制。
在一个实施例中，所述车载路面信息获取装置包括车载路面图像采集装置和车载雷达系统；
所述车载路面图像采集装置用于采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；
所述车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端。
具体地，车载路面信息获取装置包括车载路面图像采集装置和车载雷达系统。其中，车载路面图像采集装置用于采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；路况图像信息是自动驾驶车辆周围的路况信息，主要包括周围路面信息，周围车辆信息和周围障碍物、行人信息等。车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端；例如，通过激光雷达系统采集路况激光雷达点云信息，并将路况激光雷达点云信息发送至所述车载智能处理终端。车载雷达系统不仅仅能够探测周围范围内的路况信息，还能够通过雷达探测较远区域内的路况信息，有利于自动驾驶车辆的安全行驶。
本实施例中，通过车载路面图像采集装置采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；通过车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端，使得云端推理信息能够根据路况图像信息和路况点云信息进行推理。
在一个实施例中，所述车载智能处理终端包括：处理器，用于接收所述路况图像信息和路况点云信息，以及生成推理请求；通信装置，用于实现车载智能处理和云端推理服务器之间的无线通信。
具体地，图4为一个实施例中车辆自动驾驶系统中推理请求结构图，如图4所示，推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。车载智能处理终端包括处理器和通信装置，处理器连接所述车载路面信息获取装置和车载行驶控制装置，处理器用于对信息进行处理，例如对路面信息获取装置获取的信息进行处理；所述通信装置连接所述处理器，用于根据处理器的指令与云端推理服务器进行通信，例如将处理器生成的推理请求或车载路面信息获取装置获取的路面信息发送至云端推理服务器。处理器还连接车载行驶控制装置，以使得能够根据推理结果控制车载行驶控制装置执行对应的行驶指令。
本实施例中，通过车载智能处理终端中的处理器连接车载路面信息获取装置、车载行驶控制装置和通信装置，实现了对路况信息的获取、处理和与云端推理服务器之间通信。
在一个实施例中，所述车载雷达系统包括激光雷达系统、超声波雷达系统、毫米波雷达系统和微波雷达系统中的至少一种。
具体地，当自动驾驶车辆处在高速行驶的状态，可能存在因为天气或者视野的原因，无法及时看到前方的断裂或台阶路段。通过车载雷达系统，能够扫描到前方的异常路况，并及时将路面异常情况发送到自动驾驶车辆的车载智能处理终端，通过通信装置将路面异常信息发送至云端推理服务器，经过云端推理服务器的推理生成推理结果发送至自动驾驶车辆。在此种路况下，自动驾驶车辆获取路面异常信息主要靠能够探测较远距离的车载雷达系统，车载雷达系统的类型包括激光雷达系统、超声波雷达系统和微波雷达系统中的至少一种。例如，在车载雷达系统为激光雷达系统时，车载雷达系统生成路况激光雷达点云信息。
本实施例中，通过车载雷达系统对自动驾驶车辆前方的异常路段进行探测，并将异常路面信息发送至云端推理服务器，通过云端推理服务器的处理，生成推理结果发送至自动驾驶车辆以供自动驾驶车载执行对应动作，提高了自动驾驶车辆的行驶安全性。
在一个实施例中，所述通信装置包括连接所述处理器的WIFI装置、蓝牙装置、GPRS装置、3G、4G和5G模块中的的至少一种。
具体地，可以根据实际需要选择合适的无线通信装置，适用性广泛。例如，车载智能处理终端的无线通信装置使用与处理器连接的WIFI装置、蓝牙装置、GPRS装置、3G、4G和5G模块中的的至少一种；WIFI装置、蓝牙装置适合无线通信距离较近的情形，例如自动驾驶车辆在园区内等较为封闭、狭小的区域内自动行驶；GPRS装置、3G、4G和5G模块适用于较远距离的通信，例如在城际之间的公路上行驶。
本实施例中，自动驾驶车辆与云端推理服务器之间通过WIFI装置、蓝牙装置、GPRS装置、3G、4G和5G模块中的的至少一种模块进行无线连接，实现了自动驾驶车辆与云端推理服务器之间的稳定无线通信，避免了运算平台的核心算法存放自动驾驶车辆中所导致的核心算法被泄露的风险。
关于车辆自动驾驶装置的具体限定可以参见上文中对于车辆自动驾驶方法的限定，在此不再赘述。上述车辆自动驾驶装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
上述车辆自动驾驶装置，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图9所示。该计算机设备包括通过系统总线连接的处理器、存储器和网络接口。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种视频客服分配方法。
本领域技术人员可以理解，图9中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
在一个实施例中，提供了一种计算机设备，包括存储器和处理器，存储器中存储有计算机程序，该处理器执行计算机程序时实现以下步骤：
接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；根据所述深度学习模型对数据包进行推理处理，获取推理结果；将所述推理结果发送至所述自动驾驶车辆。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理处理时使用的深度学习模型。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
上述计算机设备，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算 法被泄露。
在一个实施例中，提供了一种计算机设备，包括存储器和处理器，存储器中存储有计算机程序，该处理器执行计算机程序时实现以下步骤：
生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；将所述推理请求发送至所述云端推理服务器。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：接收所述云端推理服务器发送的推理结果；对所述推理结果进行解析处理，生成驾驶指令；根据所述驾驶指令控制所述自动驾驶车辆的行驶。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
上述计算机设备，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现以下步骤：
接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；根据所述深度学习模型对数据包进行推理处理，获取推理结果；将所述推理结果发送至所述自动驾驶车辆。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理处理时使用的深度学习模型。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
上述存储介质，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将 推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现以下步骤：
生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；将所述推理请求发送至所述云端推理服务器。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：接收所述云端推理服务器发送的推理结果；对所述推理结果进行解析处理，生成驾驶指令；根据所述驾驶指令控制所述自动驾驶车辆的行驶。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
上述存储介质，通过自动驾驶车辆生成和发送推理请求至与自动驾驶车辆无线连接的云端推理服务器，并通过云端推理服务器接收自动驾驶车辆发送的推理请求，根据推理请求生成对应的推理结果，并将推理结果发送至自动驾驶车辆；最后，自动驾驶车辆接收云端推理服务器发送的推理结果，并根据推理结果执行对应驾驶指令。由于核心算法布置于与自动驾驶车辆无线连接的云端推理服务器上，避免了核心算法被泄露。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和易失性存储器中的至少一种。非易失性存储器可包括只读存储器(Read-Only Memory，ROM)、磁带、软盘、闪存或光存储器等。易失性存储器可包括随机存取存储器(Random Access Memory，RAM)或外部高速缓冲存储器。作为说明而非局限，RAM可以是多种形式，比如静态随机存取存储器(Static Random Access Memory，SRAM)或动态随机存取存储器(Dynamic Random Access Memory，DRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种车辆自动驾驶系统，其特征在于，所述系统包括：云端推理服务器和自动驾驶车辆；所述自动驾驶车辆和所述云端推理服务器之间通过无线通信连接；
    所述自动驾驶车辆用于生成和发送推理请求至所述云端推理服务器，接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；
    所述云端推理服务器用于接收所述自动驾驶车辆发送的推理请求，根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；
    其中，所述推理结果是所述云端推理服务器根据所述推理请求生成的指令；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据。
  2. 一种车辆自动驾驶方法，其特征在于，所述方法包括：
    接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
    根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型处理生成的用于控制所述自动驾驶车辆行驶的指令。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆包括：
    对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型；
    根据所述深度学习模型对数据包进行推理处理，获取推理结果；
    将所述推理结果发送至所述自动驾驶车辆。
  4. 根据权利要求3所述的方法，其特征在于，所述对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型包括：
    对所述推理请求进行解析，判断所述推理请求中是否已经选择所需的深度学习模型；
    若所述推理请求中已选择所需的深度学习模型，选择所述推理请求中已选择的深度学习模型作为推理处理时使用的深度学习模型。
  5. 根据权利要求4所述的方法，其特征在于，所述对所述推理请求进行解析，根据所述推理请求选择对应的深度学习模型还包括：
    若所述推理请求中未选择所需的深度学习模型，根据所述推理请求获取推理目的；
    根据所述推理目的的内容，选择能够实现所述推理目的的预测准确率最高的深度学习模型。
  6. 根据权利要求2所述的方法，其特征在于，所述接收自动驾驶车辆发送的推理请求，之前还包括：
    接收自动驾驶车辆发送的注册推理引擎服务请求；所述注册推理引擎服务请求是所述自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
    判断所述自动驾驶车辆是否满足预设的注册条件，若所述自动驾驶车辆满足预设的注册条件，同意所述自动驾驶车辆的注册推理引擎服务请求，发送同意注册的返回信息至所述自动驾驶车辆。
  7. 一种车辆自动驾驶方法，其特征在于，所述方法包括：
    生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
    接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
  8. 根据权利要求7所述的方法，其特征在于，所述生成并发送推理请求至云端推理服务器包括：
    当接收到推理请求的指令和/或自动驾驶车辆的运行状态符合推理请求生成条件时，生成对应的推理请求；
    将所述推理请求发送至所述云端推理服务器。
  9. 根据权利要求7所述的方法，其特征在于，所述接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令包括：
    接收所述云端推理服务器发送的推理结果；
    对所述推理结果进行解析处理，生成驾驶指令；
    根据所述驾驶指令控制所述自动驾驶车辆的行驶。
  10. 根据权利要求7所述的方法，其特征在于，所述生成并发送推理请求至云端推理服务器，之前还包括：
    向所述云端推理服务器发送注册推理引擎服务请求；所述注册推理引擎服务请求是自动驾驶车辆发送的用于申请云端推理服务器对其发送的推理请求进行推理处理的要约信息；
    接收所述云端推理服务器发送的返回信息；所述返回信息包括同意注册的返回信息和拒绝注册的返回信息。
  11. 根据权利要求8所述的方法，其特征在于，所述将所述推理请求发送至所述云端推理服务器，之前还包括：
    通过车载路面信息获取装置获取数据包，所述数据包包括路况图像信息和路况点云信息。
  12. 一种车辆自动驾驶装置，其特征在于，所述装置包括：
    推理请求接收模块，用于接收自动驾驶车辆发送的推理请求；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
    推理结果生成模块，用于根据所述推理请求生成对应的推理结果，并将所述推理结果发送至所述自动驾驶车辆；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
  13. 一种车辆自动驾驶装置，其特征在于，所述装置包括：
    推理请求生成模块，用于生成并发送推理请求至云端推理服务器；所述推理请求包括请求类型和/或数据包；所述请求类型是需要云端推理服务器进行推理服务的类型；所述数据包是根据请求类型进行推理时所需要的推理数据；
    推理结果接收模块，用于接收所述云端推理服务器发送的推理结果，并根据所述推理结果执行对应驾驶指令；所述推理结果是云端推理服务器根据所述推理请求通过深度学习模型生成的用于控制所述自动驾驶车辆行驶的指令。
  14. 根据权利要求13所述的装置，其特征在于，还包括车载路面信息获取装置、车载行驶控制装置和车载智能处理终端；
    所述车载路面信息获取装置用于获取数据包，所述数据包包括路况图像信息和路况点云信息；
    所述车载智能处理终端设置于所述推理请求生成模块内，用于生成推理请求，并将推理请求发送至云端推理服务器，以及接收所述云端推理服务器生成的推理结果；
    所述车载行驶控制装置与所述推理结果接收模块连接，用于根据所述云端推理服务器生成的推理结果控制所述自动驾驶车辆的行驶。
  15. 根据权利要求14所述的系统，其特征在于，所述车载路面信息获取装置包括车载路面图像采集装置和车载雷达系统；
    所述车载路面图像采集装置用于采集路况图像信息，并将所述路况图像信息发送至所述车载智能处理终端；
    所述车载雷达系统用于采集路况点云信息，并将路况点云信息发送至所述车载智能处理终端。
  16. 根据权利要求15所述的系统，其特征在于，所述车载智能处理终端包括：
    处理器，用于接收所述路况图像信息和路况点云信息，以及生成推理请求；
    通信装置，用于实现车载智能处理和云端推理服务器之间的无线通信。
  17. 根据权利要求15所述的系统，其特征在于，所述车载雷达系统包括激光雷达系统、超声波雷达系统、毫米波雷达系统和微波雷达系统中的至少一种。
  18. 根据权利要求16所述的系统，其特征在于，所述通信装置包括连接所述处理器的WIFI装置、蓝牙装置、GPRS装置、3G、4G和5G模块中的的至少一种。
  19. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求2至11中任一项所述的方法的步骤。
  20. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求2至11中任一项所述的方法的步骤。
PCT/CN2021/099280 2021-06-10 2021-06-10 车辆自动驾驶系统、方法、装置、计算机设备和存储介质 WO2022257056A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2021/099280 WO2022257056A1 (zh) 2021-06-10 2021-06-10 车辆自动驾驶系统、方法、装置、计算机设备和存储介质
CN202180050153.5A CN115956045A (zh) 2021-06-10 2021-06-10 车辆自动驾驶系统、方法、装置、计算机设备和存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/099280 WO2022257056A1 (zh) 2021-06-10 2021-06-10 车辆自动驾驶系统、方法、装置、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022257056A1 true WO2022257056A1 (zh) 2022-12-15

Family

ID=84424756

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/099280 WO2022257056A1 (zh) 2021-06-10 2021-06-10 车辆自动驾驶系统、方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN115956045A (zh)
WO (1) WO2022257056A1 (zh)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106713430A (zh) * 2016-12-07 2017-05-24 深圳市元征科技股份有限公司 一种驾驶控制方法及装置
CN108931971A (zh) * 2018-05-24 2018-12-04 奇瑞汽车股份有限公司 用于无人驾驶的移动终端、车辆、服务器及无人驾驶系统
CN109358614A (zh) * 2018-08-30 2019-02-19 深圳市易成自动驾驶技术有限公司 自动驾驶方法、系统、装置及可读存储介质
US20190096238A1 (en) * 2017-06-20 2019-03-28 Cavh Llc Intelligent road infrastructure system (iris): systems and methods
CN110850711A (zh) * 2019-12-06 2020-02-28 中国科学院自动化研究所 基于云端的辅助驾驶控制系统及方法
CN112650220A (zh) * 2020-12-04 2021-04-13 东风汽车集团有限公司 一种车辆自动驾驶方法、车载控制器及系统
CN112700668A (zh) * 2020-12-22 2021-04-23 北京百度网讯科技有限公司 自动驾驶的远程控制方法、自动驾驶车辆及云端设备
CN112925657A (zh) * 2021-01-18 2021-06-08 国汽智控(北京)科技有限公司 一种车路云协同处理系统及方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106713430A (zh) * 2016-12-07 2017-05-24 深圳市元征科技股份有限公司 一种驾驶控制方法及装置
US20190096238A1 (en) * 2017-06-20 2019-03-28 Cavh Llc Intelligent road infrastructure system (iris): systems and methods
CN108931971A (zh) * 2018-05-24 2018-12-04 奇瑞汽车股份有限公司 用于无人驾驶的移动终端、车辆、服务器及无人驾驶系统
CN109358614A (zh) * 2018-08-30 2019-02-19 深圳市易成自动驾驶技术有限公司 自动驾驶方法、系统、装置及可读存储介质
CN110850711A (zh) * 2019-12-06 2020-02-28 中国科学院自动化研究所 基于云端的辅助驾驶控制系统及方法
CN112650220A (zh) * 2020-12-04 2021-04-13 东风汽车集团有限公司 一种车辆自动驾驶方法、车载控制器及系统
CN112700668A (zh) * 2020-12-22 2021-04-23 北京百度网讯科技有限公司 自动驾驶的远程控制方法、自动驾驶车辆及云端设备
CN112925657A (zh) * 2021-01-18 2021-06-08 国汽智控(北京)科技有限公司 一种车路云协同处理系统及方法

Also Published As

Publication number Publication date
CN115956045A (zh) 2023-04-11

Similar Documents

Publication Publication Date Title
US20200356100A1 (en) Generation of autonomy map for autonomous vehicle
EP4067821A1 (en) Path planning method for vehicle and path planning apparatus for vehicle
JP2022511161A (ja) 無人運転車両を制御するための方法及び装置
CN107123295A (zh) 拥堵路段预测方法、装置、服务器以及存储介质
CN107015550B (zh) 诊断测试执行控制系统和方法
CN110949406B (zh) 一种智能驾驶系统及方法
US11675367B2 (en) Normalization of intelligent transport system handling characteristics
US20230249660A1 (en) Electronic Mechanical Braking Method and Electronic Mechanical Braking Apparatus
CN109087409A (zh) 一种数据处理、信息获取方法、电子设备及存储介质
CN113741440A (zh) 一种切入车辆的危险目标识别方法、装置、介质及设备
Lu et al. The emergence of vehicle computing
US20230048680A1 (en) Method and apparatus for passing through barrier gate crossbar by vehicle
CN113442948B (zh) 基于云端推理服务的自动驾驶方法、装置和计算机设备
US11703874B2 (en) System and method for collection of performance data by a vehicle
US11074108B2 (en) Caching electronic control unit mapping solutions for connected vehicles
CN110789515B (zh) 机动车辆中用于硬件验证的系统和方法
WO2022257056A1 (zh) 车辆自动驾驶系统、方法、装置、计算机设备和存储介质
US20230415753A1 (en) On-Vehicle Driving Behavior Modelling
EP3872786A1 (en) Automatic vehicle and method for operating the same
CN107562056A (zh) 一种基于云计算的机动车自动驾驶方法及系统
EP4074562A1 (en) Control system and control method for intelligent connected vehicle
US20220207348A1 (en) Real-time neural network retraining
CN115107779A (zh) 用于使用机器学习和人工智能的节能交通的系统和方法
US11613269B2 (en) Learning safety and human-centered constraints in autonomous vehicles
CN111640330A (zh) 一种基于边缘计算的防碰撞方法及相关装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21944566

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE