Intelligent control system of all-in-one machine based on cloud computing

Abstract

The invention discloses an all-in-one machine intelligent control system based on cloud computing, relates to the technical field of cloud computing, and solves the problems that firstly, it is difficult to preprocess unmanned vehicle data and preliminarily detect and analyze obstacles by means of an edge computing technology; secondly, a deep learning model is difficult to predict an obstacle behavior mode and an action intention; then, the operation state of the all-in-one machine is difficult to automatically adjust by using the safety evaluation index, and the safety of the adjusted unmanned vehicle is difficult to ensure; and finally, path planning and dynamic control instruction generation are difficult to perform by using an artificial intelligence technology, and intelligent control of the unmanned vehicle by the all-in-one machine is difficult to realize. According to the invention, unmanned vehicle data and obstacle detection are preprocessed by using an edge computing technology; a dynamic control instruction is generated by automatically adjusting the running state of the all-in-one machine and planning a path, and intelligent control is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud computing, and in particular is an all-in-one intelligent control system based on cloud computing. Background Art

[0002] With the development of cloud computing technology, cloud computing provides powerful computing power and storage resources for processing massive data. The powerful computing power and storage capacity of the cloud computing platform are used to analyze and process the large amount of data collected by the all-in-one machine, thereby realizing intelligent control and management. The all-in-one machine integrates various sensors and actuators of the unmanned vehicle and is the core equipment for the unmanned vehicle to achieve autonomous operation. However, there are many challenges in relying solely on cloud computing. Therefore, edge computing technology has emerged. By preprocessing and preliminarily analyzing the data of unmanned vehicles at the edge node, the burden of cloud computing can be effectively reduced. In addition, the deep learning model, with its powerful data processing and analysis capabilities, can efficiently perform pattern recognition and prediction, providing strong support for the intelligent control of unmanned vehicles. Artificial intelligence technology is used to plan paths and generate dynamic control instructions to realize the intelligent control of the all-in-one machine.

[0003] The following problems exist in the existing technology: first, it is difficult to use edge computing technology to pre-process unmanned vehicle data and preliminarily detect and analyze obstacles; second, it is difficult to use deep learning models to predict obstacle behavior patterns and action intentions; then, it is difficult to use the safety assessment index to automatically adjust the operating status of the all-in-one machine and ensure the safety of the adjusted unmanned vehicle; finally, it is difficult to use artificial intelligence technology to plan paths and generate dynamic control instructions, and it is difficult to realize the intelligent control of the unmanned vehicle by the all-in-one machine. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an all-in-one intelligent control system based on cloud computing, comprising the following modules:

[0005] Data collection module: used to collect real-time data of the unmanned vehicle from various sensors and actuators of the integrated machine, and transmit the real-time data to the edge node module; the real-time data of the unmanned vehicle includes: traffic flow, monitoring images, location and speed of the unmanned vehicle;

[0006] Edge node module: Using edge computing technology, edge nodes are deployed between the data collection module and the cloud computing center module; real-time data of the unmanned vehicle is preprocessed, including: denoising, data cleaning and image enhancement; preliminary analysis is performed based on the real-time data of the unmanned vehicle to detect obstacles;

[0007] Cloud computing center module: used to receive real-time data transmitted by the edge node module; segment the environment around the unmanned vehicle based on the monitoring image and predict the real-time obstacle behavior pattern and action intention; automatically adjust the operating status of the integrated machine based on the real-time data and re-evaluate the safety of the adjusted operating status of the unmanned vehicle; plan the path and generate dynamic control instructions based on the segmentation results of the environment around the unmanned vehicle; send the generated control instructions to the instruction issuing module;

[0008] Instruction issuing module: receives the control instructions generated by the cloud computing center module, and sends the control instructions to the actuator of the all-in-one machine to realize intelligent control of the all-in-one machine.

[0009] Preferably, performing preliminary analysis based on the real-time data of the unmanned vehicle to detect obstacles includes the following steps:

[0010] By using the convolutional neural network model in the computer vision algorithm, obstacles in the monitoring image are detected;

[0011] Collecting a historical monitoring image data set, wherein the historical monitoring image data set has obstacle labels; inputting the historical monitoring image data set into a convolutional neural network model for training; performing obstacle detection on a real-time input monitoring image according to the trained convolutional neural network model, and identifying and locating obstacles in the monitoring image;

[0012] When an obstacle is detected, the edge node uses a classification algorithm to classify the obstacle and distinguish the type of obstacle, including pedestrians, vehicles, animals, and road signs;

[0013] The edge node tracks and detects the detected obstacles through Kalman filter, particle filter or multi-target tracking algorithm to determine the position and speed of the obstacles. Based on the obstacle recognition result and the current state of the unmanned vehicle, the edge node predicts the distance between the unmanned vehicle and the obstacle. The potential collision risk index is obtained by calculating the ratio of the predicted distance between the unmanned vehicle and the obstacle and the speed of the unmanned vehicle.

[0014] When the potential collision risk index is lower than the preset safety threshold, the edge node will consider an emergency to have occurred and immediately send an emergency braking command to the braking system of the unmanned vehicle, causing the drone to stop moving forward, otherwise it will continue to detect obstacles; when the edge node confirms that the emergency braking command has been executed and monitors the real-time status of the unmanned vehicle, it will synchronously transmit the emergency handling results and the real-time data of the unmanned vehicle to the cloud computing center module.

[0015] Preferably, segmenting the surrounding environment of the unmanned vehicle according to the monitoring image and predicting the real-time obstacle behavior pattern and action intention includes the following steps:

[0016] Use image segmentation technology to divide different areas in the surveillance image into road areas, sidewalk areas and building areas;

[0017] According to the obstacle detection results, the behavior pattern of the detected obstacles is analyzed to determine whether pedestrians are walking, vehicles are changing lanes, and animals are on the road;

[0018] Use deep learning models to predict the behavior patterns and action intentions of obstacles based on their behavior; annotate the collected historical surveillance image data sets, including: road areas, sidewalk areas, building areas, and the location, category, and behavior patterns of obstacles;

[0019] Using a deep learning model that combines a convolutional neural network and a recurrent neural network, the annotated historical surveillance image dataset is input into the deep learning model for training;

[0020] The trained deep learning model is deployed to the cloud computing center module to perform real-time obstacle behavior pattern and action intention prediction.

[0021] Preferably, automatically adjusting the operating state of the integrated machine according to the real-time data and re-evaluating the safety of the adjusted operating state of the unmanned vehicle comprises the following steps:

[0022] By monitoring the real-time data of the unmanned vehicle in real time, the real-time operating status of the unmanned vehicle is evaluated for safety, and the safety evaluation index formula is calculated:

[0023]

[0024] Get the safety assessment index Q, where V a ,θ a , D a and T a They represent the speed, steering angle, following distance and traffic flow of the unmanned vehicle, V max ,θ max , D max and T max They represent the maximum permissible speed, maximum permissible steering angle, safe following distance and maximum permissible traffic flow of the unmanned vehicle respectively; α, β, γ and δ represent weight coefficients respectively;

[0025] According to the results of the safety assessment index, the operating status of the integrated machine is automatically adjusted, including: adjusting the driving speed, steering angle, and following distance of the unmanned vehicle; the adjusted operating status of the unmanned vehicle is safety assessed again, and the process ends when the safety assessment index reaches the safety threshold.

[0026] Preferably, performing path planning and generating dynamic control instructions based on the segmentation results of the environment surrounding the unmanned vehicle includes the following steps:

[0027] Using artificial intelligence technology, including A* algorithm, Dijkstra algorithm or deep learning model, to plan the initial global path from the current position to the target position for the unmanned vehicle; dynamically update the grid map and path planning according to the real-time status of the unmanned vehicle and changes in the surrounding environment;

[0028] Based on the global path, local path planning is performed, and obstacle avoidance operations are performed through control instructions according to the identified and detected obstacles; based on the path planning results, control instructions are generated including: acceleration, deceleration and steering.

[0029] Preferably, receiving the control instruction generated by the cloud computing center module and sending the control instruction to the actuator of the all-in-one machine, and intelligently controlling the all-in-one machine includes the following steps:

[0030] Receive the control instructions generated by the cloud computing center module and send them to the actuator of the all-in-one machine; the actuator controls the driving direction, speed and steering action of the unmanned vehicle according to the received control instructions; based on the execution results, the actuator then feeds back to the cloud computing center module, and the cloud computing center module continuously monitors the status of the unmanned vehicle.

[0031] Compared with the prior art, the present invention has at least the following beneficial effects:

[0032] The present invention deploys edge nodes between the data collection module and the cloud computing center module, and uses the edge nodes to pre-process the unmanned vehicle data and preliminarily analyze obstacles, thereby reducing the burden on the cloud computing center, and judging whether to perform emergency braking to react quickly based on the obstacle detection results, and monitoring the real-time status of the unmanned vehicle;

[0033] The present invention uses monitoring images to segment the environment around the unmanned vehicle, divides it into different areas, and uses a deep learning model to predict the behavior pattern and action intention of obstacles, so as to provide more accurate information support for the path planning and obstacle avoidance of the unmanned vehicle; the safety evaluation index is analyzed according to the real-time data of the real-time state of the unmanned vehicle to automatically adjust the operating state of the integrated machine, and the operating state of the unmanned vehicle is further evaluated according to the adjusted operating state of the integrated machine to ensure the safety of the operating state of the unmanned vehicle;

[0034] The present invention utilizes artificial intelligence technology to plan the path of the unmanned vehicle and generate dynamic control instructions; the control instructions are executed by the actuator in the all-in-one machine, thereby controlling the driving direction, speed and steering action of the unmanned vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION

[0037] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0038] In one embodiment, see Figure 1 As shown, an all-in-one intelligent control system based on cloud computing is provided, including the following modules:

[0039] Data collection module: used to collect real-time data of the unmanned vehicle from various sensors and actuators of the integrated machine, and transmit the real-time data to the edge node module; the real-time data of the unmanned vehicle includes: traffic flow, monitoring images, location and speed of the unmanned vehicle;

[0040] Edge node module: Using edge computing technology, edge nodes are deployed between the data collection module and the cloud computing center module; real-time data of the unmanned vehicle is preprocessed, including: denoising, data cleaning and image enhancement; preliminary analysis is performed based on the real-time data of the unmanned vehicle to detect obstacles;

[0041] Cloud computing center module: used to receive real-time data transmitted by the edge node module; segment the environment around the unmanned vehicle based on the monitoring image and predict the real-time obstacle behavior pattern and action intention; automatically adjust the operating status of the integrated machine based on the real-time data and re-evaluate the safety of the adjusted operating status of the unmanned vehicle; plan the path and generate dynamic control instructions based on the segmentation results of the environment around the unmanned vehicle; send the generated control instructions to the instruction issuing module;

[0042] Instruction issuing module: receives the control instructions generated by the cloud computing center module, and sends the control instructions to the actuator of the all-in-one machine to realize intelligent control of the all-in-one machine.

[0043] Specifically, real-time data is collected from various sensors and actuators on the integrated machine, and the real-time data includes: traffic flow, monitoring images, the position and speed of the unmanned vehicle and other key information; the real-time data is transmitted to the edge node module for further processing. Using edge computing technology, the edge node is deployed between the data collection module and the cloud computing center module. The real-time data of the unmanned vehicle is preprocessed, including: denoising and data cleaning of traffic flow, the position and speed of the unmanned vehicle, denoising of the monitoring image, and enhancing the image comparison and edge information to improve data quality and analysis efficiency. Preliminary analysis is performed based on the real-time data of the unmanned vehicle, and obstacles are detected to provide a basis for subsequent decision-making. The real-time data transmitted by the edge node module is received, and the surrounding environment of the unmanned vehicle is segmented according to the monitoring image, and the real-time obstacle behavior mode and action intention are predicted. The operating state of the unmanned vehicle is automatically adjusted according to the real-time data, and the operating state of the unmanned vehicle is re-evaluated for safety. Path planning and dynamic control instructions are generated based on the segmentation results of the surrounding environment to optimize the driving path and driving efficiency of the unmanned vehicle. The generated control instructions are sent to the instruction issuing module. Receive the control instructions generated by the cloud computing center module and send them to the actuator of the all-in-one machine to realize intelligent control of the unmanned vehicle.

[0044] In one embodiment, performing preliminary analysis based on the real-time data of the unmanned vehicle to detect obstacles includes the following steps:

[0045] By using the convolutional neural network model in the computer vision algorithm, obstacles in the monitoring image are detected;

[0046] Collecting a historical monitoring image data set, wherein the historical monitoring image data set has obstacle labels; inputting the historical monitoring image data set into a convolutional neural network model for training; performing obstacle detection on a real-time input monitoring image according to the trained convolutional neural network model, and identifying and locating obstacles in the monitoring image;

[0047] When an obstacle is detected, the edge node uses a classification algorithm to classify the obstacle and distinguish the type of obstacle, including pedestrians, vehicles, animals, and road signs;

[0048] The edge node tracks and detects the detected obstacles through Kalman filter, particle filter or multi-target tracking algorithm to determine the position and speed of the obstacles. Based on the obstacle recognition result and the current state of the unmanned vehicle, the edge node predicts the distance between the unmanned vehicle and the obstacle. The potential collision risk index is obtained by calculating the ratio of the predicted distance between the unmanned vehicle and the obstacle and the speed of the unmanned vehicle.

[0049] When the potential collision risk index is lower than the preset safety threshold, the edge node will consider an emergency to have occurred and immediately send an emergency braking command to the braking system of the unmanned vehicle, causing the drone to stop moving forward, otherwise it will continue to detect obstacles; when the edge node confirms that the emergency braking command has been executed and monitors the real-time status of the unmanned vehicle, it will synchronously transmit the emergency handling results and the real-time data of the unmanned vehicle to the cloud computing center module.

[0050] Specifically, a large number of historical surveillance image data sets are collected. These data sets should have detailed obstacle labels, such as pedestrians, vehicles, animals, road signs, etc. These historical surveillance image data sets are input into the convolutional neural network model for training; during the training process, the model will learn how to extract features from the surveillance images and classify and identify obstacles based on these features. The trained convolutional neural network model is used to detect obstacles in the real-time input surveillance images. The convolutional neural network model automatically identifies and locates obstacles in the image. When an obstacle is detected, the edge node further uses a classification algorithm to classify the obstacle to distinguish the specific type of the obstacle. The edge node tracks and detects the detected obstacles through a Kalman filter, a particle filter, or a multi-target tracking algorithm. These algorithms can predict the future position and speed of the obstacle based on its historical position and speed information. Through the tracking algorithm, the dynamic information of the obstacle can be obtained in real time, providing basic data for subsequent collision prediction and obstacle avoidance path generation. Based on the recognition results of the obstacle and the current state of the unmanned vehicle, such as speed, direction, etc., the edge node predicts the distance between the unmanned vehicle and the obstacle. The potential collision risk index is obtained by calculating the ratio of the predicted distance between the unmanned vehicle and the obstacle to the speed of the unmanned vehicle. This index reflects the probability of a collision between the unmanned vehicle and the obstacle. When the potential collision risk index is lower than the preset safety threshold, the edge node will consider that an emergency has occurred and immediately send an emergency braking command to the braking system of the unmanned vehicle. This safety threshold is set and adjusted based on a variety of factors such as the performance of the unmanned vehicle, road conditions, and traffic rules to ensure the timeliness and effectiveness of emergency braking. In this embodiment, the preset threshold is 1.2. After receiving the emergency braking command, the unmanned vehicle will immediately perform the braking operation and stop moving forward. The edge node will continue to monitor the real-time status of the unmanned vehicle to ensure that the emergency braking command has been correctly executed. When the edge node confirms that the emergency braking command has been executed, it will synchronously transmit the emergency processing results and the real-time data of the unmanned vehicle to the cloud computing center module, including the status information of the unmanned vehicle, the detection and identification results of the obstacle, the execution of the emergency braking command, etc.

[0051] In one embodiment, segmenting the surrounding environment of the unmanned vehicle based on the monitoring image and predicting the real-time obstacle behavior pattern and action intention includes the following steps:

[0052] Use image segmentation technology to divide different areas in the surveillance image into road areas, sidewalk areas and building areas;

[0053] According to the obstacle detection results, the behavior pattern of the detected obstacles is analyzed to determine whether pedestrians are walking, vehicles are changing lanes, and animals are on the road;

[0054] Use deep learning models to predict the behavior patterns and action intentions of obstacles based on their behavior; annotate the collected historical surveillance image data sets, including: road areas, sidewalk areas, building areas, and the location, category, and behavior patterns of obstacles;

[0055] Using a deep learning model that combines a convolutional neural network and a recurrent neural network, the annotated historical surveillance image dataset is input into the deep learning model for training;

[0056] The trained deep learning model is deployed to the cloud computing center module to perform real-time obstacle behavior pattern and action intention prediction.

[0057] Specifically, image segmentation technology, including semantic segmentation or instance segmentation, is used to accurately divide the surveillance image. Through image segmentation, different areas in the surveillance image are divided into road areas, sidewalk areas, and building areas. The obstacles in the surveillance image are detected using a convolutional neural network model, and the detected obstacles include pedestrians, vehicles, animals, etc. The behavior pattern of the detected obstacles is analyzed to determine the walking state of pedestrians, including walking, stillness, running, etc., the driving state of vehicles, including straight, changing lanes, turning, etc., and the activity state of animals, including walking on the road, stillness, etc. A large number of historical surveillance image data sets are collected and these data sets are annotated in detail. The annotation content includes: road area, sidewalk area, building area, and the location, category and behavior pattern of obstacles. A deep learning model combining convolutional neural network CNN and recurrent neural network RNN ​​is constructed, wherein CNN is used to extract feature information in the image, while RNN is used to process sequence data and capture the behavior pattern of obstacles; the annotated historical surveillance image data set is input into the deep learning model for training. Through training, the deep learning model can learn the characteristics, behavior patterns and relationships between obstacles. The trained deep learning model is deployed to the cloud computing center module, and the powerful computing power of the cloud computing center is used to predict the behavior patterns and action intentions of obstacles in the real-time input monitoring images. When the unmanned vehicle is driving, the cloud computing center module will receive the real-time transmitted monitoring image data, and use the deep learning model deployed in the cloud computing center to predict the behavior patterns and action intentions of obstacles in the monitoring images in real time.

[0058] In one embodiment, automatically adjusting the operating state of the integrated machine according to the real-time data and re-evaluating the safety of the adjusted operating state of the unmanned vehicle includes the following steps:

[0059] By monitoring the real-time data of the unmanned vehicle in real time, the real-time operating status of the unmanned vehicle is evaluated for safety, and the safety evaluation index formula is calculated:

[0060]

[0061] Get the safety assessment index Q, where V a ,θ a , D a and T a They represent the speed, steering angle, following distance and traffic flow of the unmanned vehicle, V max ,θ max , D max and T max They represent the maximum permissible speed, maximum permissible steering angle, safe following distance and maximum permissible traffic flow of the unmanned vehicle respectively; α, β, γ and δ represent weight coefficients respectively;

[0062] According to the results of the safety assessment index, the operating status of the integrated machine is automatically adjusted, including: adjusting the driving speed, steering angle, and following distance of the unmanned vehicle; the adjusted operating status of the unmanned vehicle is safety assessed again, and the process ends when the safety assessment index reaches the safety threshold.

[0063] Specifically, sensors are used to monitor and collect real-time data of unmanned vehicles in real time. The maximum permissible speed, maximum permissible steering angle, safe following distance and maximum permissible traffic flow of unmanned vehicles need to be dynamically set and adjusted according to factors such as the performance of the unmanned vehicle, road conditions, traffic rules, and actual conditions. The maximum permissible speed, maximum permissible steering angle, safe following distance and maximum permissible traffic flow of unmanned vehicles should ensure that unmanned vehicles can operate safely in most cases. α, β, γ and δ are weight coefficients, whose values ​​are 0.3, 0.25, 0.25 and 0.2 respectively in this implementation, and the weight coefficients are dynamically adjusted according to historical data and actual conditions. The real-time monitored unmanned vehicle status, traffic flow data and weight coefficients are substituted into the safety assessment index formula for calculation to obtain the safety assessment index Q. According to the result of the safety assessment index Q, the operating status of the integrated machine is automatically adjusted. When the Q value is higher than 0.6, it means that there are safety hazards in the current operation state of the unmanned vehicle, and it is necessary to reduce the driving speed, adjust the steering angle or increase the following distance; when the Q value is between 0.3 and 0.6, it means that there may be safety hazards and vigilance is required; when the Q value is lower than 0.3, it means that the current operation state of the unmanned vehicle is safe, and the driving speed can be maintained or appropriately increased. After adjusting the operating state of the all-in-one machine, it is necessary to conduct a safety assessment of the operating state of the unmanned vehicle again to ensure that the adjusted state of the unmanned vehicle is safer. If the safety assessment index of the adjusted unmanned vehicle still does not meet the safety threshold, it is necessary to continue to adjust until the safety standard is met. At the same time, the results of the safety assessment and the data of the adjustment process are fed back to the cloud computing center.

[0064] In one embodiment, performing path planning and generating dynamic control instructions based on the segmentation results of the environment around the unmanned vehicle includes the following steps:

[0065] Using artificial intelligence technology, including A* algorithm, Dijkstra algorithm or deep learning model, to plan the initial global path from the current position to the target position for the unmanned vehicle; dynamically update the grid map and path planning according to the real-time status of the unmanned vehicle and changes in the surrounding environment;

[0066] Based on the global path, local path planning is performed, and obstacle avoidance operations are performed through control instructions according to the identified and detected obstacles; based on the path planning results, control instructions are generated including: acceleration, deceleration and steering.

[0067] Specifically, the A* algorithm, Dijkstra algorithm or deep learning model in artificial intelligence technology are used to plan the initial global path from the current position to the target position for the unmanned vehicle. According to the starting position, target position and high-precision map information of the unmanned vehicle, an optimal path is generated using the selected algorithm; the path should be as short and safe as possible to avoid collisions with obstacles and violations of traffic rules. According to the real-time status of the unmanned vehicle and changes in the surrounding environment, the grid map is dynamically updated. The grid map is a representation method that divides the environment into multiple small grids and marks whether each grid is occupied by an obstacle. The surrounding environment is perceived in real time through sensors, and the obstacle information in the grid map is updated; based on the dynamically updated grid map, the path of the unmanned vehicle is replanned; if new obstacles are found or traffic conditions change, the path is adjusted in time to avoid collisions or optimize the driving route. Based on the global path, local path planning is performed. Local path planning mainly focuses on the driving path of the unmanned vehicle near the current position to deal with emergencies and obstacles. By using the convolutional neural network model on the edge node to identify and detect obstacles in real time and judge their impact on the driving of the unmanned vehicle based on information such as the location, speed and direction of movement of the obstacles; according to the obstacle information identified and detected, obstacle avoidance operations are performed through control instructions, which include: changing the driving direction, accelerating or decelerating, etc., to ensure that the unmanned vehicle can safely bypass the obstacles. Based on the path planning results, control instructions are generated including: acceleration, deceleration and steering, etc., and the control instructions are directly used to control the driving state of the unmanned vehicle. According to the current position of the unmanned vehicle and the planned path, the required control instructions are calculated and sent to the actuator of the unmanned vehicle to control the acceleration, deceleration and steering of the unmanned vehicle.

[0068] In one embodiment, receiving a control instruction generated by a cloud computing center module and sending the control instruction to an actuator of the all-in-one machine, and intelligently controlling the all-in-one machine includes the following steps:

[0069] Receive the control instructions generated by the cloud computing center module and send them to the actuator of the all-in-one machine; the actuator controls the driving direction, speed and steering action of the unmanned vehicle according to the received control instructions; based on the execution results, the actuator then feeds back to the cloud computing center module, and the cloud computing center module continuously monitors the status of the unmanned vehicle.

[0070] Specifically, the corresponding control instructions are generated according to the current state, target position, and surrounding environment of the unmanned vehicle; the control instructions are transmitted to the integrated machine through the communication network, and the integrated machine has a built-in receiving module for receiving the control instructions from the cloud computing center module. After receiving the control instructions, the integrated machine sends the instructions to the actuator through the internal communication bus or interface. The actuator is a hardware device responsible for the specific execution of the control instructions, such as a steering controller. According to the received control instructions, the driving direction, speed, and steering action of the unmanned vehicle are precisely controlled. If the control instruction requires the unmanned vehicle to accelerate and turn right, the actuator will adjust the motor speed and the angle of the steering mechanism accordingly, and the execution result feedback actuator will collect the status information of the unmanned vehicle such as the current speed, driving direction, steering angle, etc. after completing the control action. These status information is fed back to the cloud computing center module through the communication bus or interface inside the integrated machine. The cloud computing center module continuously monitors, and after receiving the status information fed back by the actuator, the cloud computing center module will perform real-time processing and analysis. By comparing the actual state of the unmanned vehicle with the expected state, the cloud computing center module can judge the execution effect of the control instruction.

[0071] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. An all-in-one intelligent control system based on cloud computing, characterized in that: Includes the following modules: Data collection module: used to collect real-time data of the unmanned vehicle from the sensors and actuators of the integrated machine, and transmit the real-time data to the edge node module; The real-time data of the unmanned vehicle includes: traffic flow, monitoring images, location and speed of the unmanned vehicle; Edge node module: Using edge computing technology, edge nodes are deployed between the data collection module and the cloud computing center module; real-time data of the unmanned vehicle is preprocessed, including: denoising, data cleaning and image enhancement; preliminary analysis is performed based on the real-time data of the unmanned vehicle to detect obstacles; Cloud computing center module: used to receive real-time data transmitted by the edge node module; segment the environment around the unmanned vehicle based on the monitoring image and predict the real-time obstacle behavior pattern and action intention; automatically adjust the operating status of the integrated machine based on the real-time data and re-evaluate the safety of the adjusted operating status of the unmanned vehicle; plan the path and generate dynamic control instructions based on the segmentation results of the environment around the unmanned vehicle; send the generated control instructions to the instruction issuing module; Instruction issuing module: receives the control instructions generated by the cloud computing center module, and sends the control instructions to the actuator of the all-in-one machine to realize intelligent control of the all-in-one machine.

2. The cloud computing-based all-in-one intelligent control system according to claim 1, characterized in that: Perform a preliminary analysis based on the real-time data of the unmanned vehicle to detect obstacles, including the following steps: By using the convolutional neural network model in the computer vision algorithm, obstacles in the monitoring image are detected; Collecting a historical monitoring image data set, wherein the historical monitoring image data set has obstacle labels; inputting the historical monitoring image data set into a convolutional neural network model for training; performing obstacle detection on a real-time input monitoring image according to the trained convolutional neural network model, and identifying and locating obstacles in the monitoring image; When an obstacle is detected, the edge node uses a classification algorithm to classify the obstacle and distinguish the type of obstacle, including pedestrians, vehicles, animals, and road signs; The edge node tracks and detects the detected obstacles through Kalman filter, particle filter or multi-target tracking algorithm to determine the position and speed of the obstacles. Based on the obstacle recognition result and the current state of the unmanned vehicle, the edge node predicts the distance between the unmanned vehicle and the obstacle. The potential collision risk index is obtained by calculating the ratio of the predicted distance between the unmanned vehicle and the obstacle and the speed of the unmanned vehicle. When the potential collision risk index is lower than the preset safety threshold, the edge node will consider an emergency to have occurred and immediately send an emergency braking command to the braking system of the unmanned vehicle, causing the drone to stop moving forward, otherwise it will continue to detect obstacles; when the edge node confirms that the emergency braking command has been executed and monitors the real-time status of the unmanned vehicle, it will synchronously transmit the emergency handling results and the real-time data of the unmanned vehicle to the cloud computing center module.

3. The cloud computing-based all-in-one intelligent control system according to claim 1, characterized in that: The process of segmenting the surrounding environment of the unmanned vehicle based on the monitoring image and predicting the real-time obstacle behavior pattern and action intention includes the following steps: Use image segmentation technology to divide different areas in the surveillance image into road areas, sidewalk areas and building areas; According to the obstacle detection results, the behavior pattern of the detected obstacles is analyzed to determine whether pedestrians are walking, vehicles are changing lanes, and animals are on the road; Use deep learning models to predict the behavior patterns and action intentions of obstacles based on their behavior; annotate the collected historical surveillance image data sets, including: road areas, sidewalk areas, building areas, and the location, category, and behavior patterns of obstacles; Using a deep learning model that combines a convolutional neural network and a recurrent neural network, the annotated historical surveillance image dataset is input into the deep learning model for training; The trained deep learning model is deployed to the cloud computing center module to perform real-time obstacle behavior pattern and action intention prediction.

4. The cloud computing-based all-in-one intelligent control system according to claim 1, characterized in that: Automatically adjust the operating status of the integrated machine according to real-time data and re-evaluate the safety of the adjusted operating status of the unmanned vehicle, including the following steps: By monitoring the real-time data of the unmanned vehicle in real time, the real-time operating status of the unmanned vehicle is evaluated for safety, and the safety evaluation index formula is calculated: Get the safety assessment index Q, where V a ,θ a , D a and T a They represent the speed, steering angle, following distance and traffic flow of the unmanned vehicle, V max ,θ max , D max and T max They represent the maximum permissible speed, maximum permissible steering angle, safe following distance and maximum permissible traffic flow of the unmanned vehicle respectively; α, β, γ and δ represent weight coefficients respectively; According to the results of the safety assessment index, the operating status of the integrated machine is automatically adjusted, including: adjusting the driving speed, steering angle, and following distance of the unmanned vehicle; the adjusted operating status of the unmanned vehicle is safety assessed again, and the process ends when the safety assessment index reaches the safety threshold.

5. The cloud computing-based all-in-one intelligent control system according to claim 1, characterized in that: Path planning and dynamic control instructions are generated based on the segmentation results of the environment around the unmanned vehicle, including the following steps: Using artificial intelligence technology, including A* algorithm, Dijkstra algorithm or deep learning model, to plan the initial global path from the current position to the target position for the unmanned vehicle; dynamically update the grid map and path planning according to the real-time status of the unmanned vehicle and changes in the surrounding environment; Based on the global path, local path planning is performed, and obstacle avoidance operations are performed through control instructions according to the identified and detected obstacles; based on the path planning results, control instructions are generated including: acceleration, deceleration and steering.

6. The cloud computing-based all-in-one intelligent control system according to claim 1, characterized in that: Receiving the control instructions generated by the cloud computing center module and sending the control instructions to the actuator of the all-in-one machine, the intelligent control of the all-in-one machine includes the following steps: Receive the control instructions generated by the cloud computing center module and send them to the actuator of the all-in-one machine; the actuator controls the driving direction, speed and steering action of the unmanned vehicle according to the received control instructions; based on the execution results, the actuator then feeds back to the cloud computing center module, and the cloud computing center module continuously monitors the status of the unmanned vehicle.

Images