Equipment chassis data automatic acquisition and analysis system based on CAN bus

Abstract

The invention relates to the technical field of engineering equipment, and particularly discloses an equipment chassis data automatic acquisition and analysis system based on a CAN (Controller Area Network) bus, which is used for solving the problem of real-time optimization of chassis control and equipment operation of two types of engineering equipment vehicles in a complex environment. The system comprises a bridging engineering module, a two-type engineering equipment vehicle chassis, a mine clearance engineering module and a vehicle CPU (Central Processing Unit), and the two-type engineering equipment vehicle chassis is connected with a data analysis module; the data analysis module is connected with the bridging engineering module, the two-type engineering equipment vehicle chassis and the mine sweeping engineering module, and the bridging engineering module, the two-type engineering equipment vehicle chassis and the mine sweeping engineering module are connected with the vehicle CPU; according to the method, chassis and equipment operation strategies are optimized by separating chassis data and combining a deep learning model and a knowledge graph, the flexibility and dynamic adaptability of multi-task scheduling are ensured, and the efficiency of bridging and mine sweeping tasks is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data acquisition, and more specifically, to an automatic acquisition and analysis system for equipment chassis data based on a CAN bus. Background Art

[0002] When two types of engineering equipment vehicles are carrying out bridge-building and mine-sweeping tasks, the stability of the vehicle chassis and the accuracy of equipment operation are the key to successfully completing the tasks. However, during the tasks, they will face complex environments such as rugged terrain, high temperature, low temperature and obstacles, which bring huge challenges to the operation of vehicles and equipment. It is necessary to monitor the chassis status in real time through a variety of chassis data, and also to perceive external conditions through external environmental sensors. It is also necessary to coordinate the data collected in real time by the two types of engineering and the data collected by the on-board system itself to ensure that the bridge-building engineering tasks and mine-sweeping engineering tasks can be smoothly coordinated with the chassis data, external perception data, two types of engineering work data and on-board information data. However, these data have high latitude and high frequency. How to extract the key features that affect the chassis control and equipment operation without sacrificing real-time performance, and analyze the key features to improve the flexibility and dynamic adaptability of the two types of engineering equipment vehicles, while responding quickly to changes in the external environment and optimizing multi-task scheduling is the main problem that needs to be solved urgently. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides an automatic equipment chassis data collection and analysis system based on CAN bus, which optimizes the chassis and equipment operation strategies in real time by separating the chassis data and combining the deep learning model and knowledge graph, ensures the flexibility and dynamic adaptability of multi-task scheduling, and improves the efficiency and safety of bridge-building and mine-sweeping tasks.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A CAN bus-based equipment chassis data automatic collection and analysis system includes a bridge engineering module, two types of engineering equipment vehicle chassis, a minesweeping engineering module and a vehicle CPU. The two types of engineering equipment vehicle chassis are connected to a data analysis module, and the data analysis module is connected to the bridge engineering module, the two types of engineering equipment vehicle chassis and the minesweeping engineering module. The bridge engineering module, the two types of engineering equipment vehicle chassis and the minesweeping engineering module are connected to the vehicle CPU. The two types of engineering equipment vehicle chassis identify chassis data as chassis sensor direct acquisition data and CAN bus data, use a first deep learning model to obtain a first target feature from the CAN bus data, use a second deep learning module to divide the chassis sensor direct acquisition data into a first type of feature and a second type of feature according to a trigger condition, and use a first knowledge graph and a second knowledge graph to obtain pre-selected features and auxiliary screening features therefrom, respectively. Through the third knowledge graph, auxiliary screening features are used to obtain the second target features from the pre-selected features, and then the chassis target features are obtained through feature fusion, and the chassis target features are transmitted to the data analysis module. The data analysis module extracts peripheral features from peripheral conditions and vehicle-borne information through the peripheral feature extraction module, and combines the peripheral features with the chassis target features for data analysis to obtain the coordinated optimization strategy of the chassis, bridge-building engineering, and mine-sweeping engineering, and transmits it to the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles, and coordinately optimizes the relevant parameters of the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles. At the same time, the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles feed back their respective status data to the vehicle CPU, and feedback and control the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles.

[0006] As a further solution of the present invention, the chassis of two types of engineering equipment vehicles are connected to a chassis data acquisition module for collecting chassis data. The chassis data acquisition module classifies the data into chassis sensor direct data and CAN bus data through the embedded source information in the metadata of the collected data. The chassis sensor direct data is connected to ADC1 / DAC1 through APB1. ADC1 / DAC1 is connected to a data source label generation module to generate data with data source label codes of "chassis, vehicle related" and "external environment, accessory equipment related", and transmit the labeled data to the second deep learning model. The features acquired by the second deep learning model trigger the classification of the first type of features and the second type of features by judging the numerical value of the data source label code and the preset threshold. The first type of features are chassis and vehicle related features, and the second type of features are external environment and accessory equipment related features. The first type of features are input into the first knowledge graph, and pre-selected features are retrieved. The second type of features are input into the second knowledge graph to obtain auxiliary screening features. The pre-selected features are input into the third knowledge graph to obtain the second target features under the constraints of the auxiliary screening features.

[0007] As a further solution of the present invention, the data source label generation module and the chassis data acquisition module are both connected to the data analysis module through a feature fusion module. The feature fusion module fuses the first target feature and the second target feature through feature engineering to obtain the chassis target feature.

[0008] As a further scheme of the present invention, in the chassis data acquisition module, the first target feature obtained by the collected CAN bus data through the first deep learning model includes data 1 and data 2. The length of data 1 and data 2 is 8. Data 1 includes but is not limited to brake light status, left turn light status, right turn light status, left turn switch, right turn switch, oil pressure, boost pressure, engine water temperature and engine oil temperature. Data 2 includes but is not limited to engine speed, equipment speed, current mileage, current motorcycle hours. The chassis sensor direct data includes but is not limited to vehicle acceleration, vehicle angular velocity, chassis height, hydraulic cylinder working status, chassis position, tire pressure, and suspension system status.

[0009] As a further solution of the present invention, the first knowledge graph is constructed based on chassis status and vehicle-related data, and is used to identify and retrieve a knowledge base of vehicle chassis status and control pre-selected features. The pre-selected features are chassis height, bridge pushing position and pressure, hydraulic cylinder status, vehicle acceleration and angular velocity. The second knowledge graph is constructed based on the external environment and auxiliary equipment status, and is used to identify and retrieve a knowledge base of auxiliary features that affect task operations. The auxiliary screening features are the degree of influence of external temperature on minesweeping and bridge erection working conditions, the degree of association of obstacle distance with path planning, and the degree of influence of wind speed on working stability. The third knowledge graph combines pre-selected features and auxiliary features to generate a knowledge base of second target features for chassis control and two-type equipment operation, which can adapt to the actual working environment and task requirements to the greatest extent. The second target features are the adjustment amount of bridge pushing pressure and environmental adaptability, the path optimization adjustment amount, the chassis posture and vehicle speed that maintain chassis stability and the best bridge erection and minesweeping working conditions, and the chassis target features are the chassis posture adjustment amount, the bridge pushing and hydraulic system optimization amount, and the chassis speed and stability control amount.

[0010] As a further solution of the present invention, the peripheral feature extraction module is connected to the DWA controller memory, network adapter, and hard disk controller through AHB. The peripheral feature extraction module communicates information with the bus matrix through AHB to obtain vehicle CPU data. The peripheral feature extraction module is connected to APB2 through ADC2 / DAC2. APB2 is connected to USRT, UART, GPIO, PMW output, timer, and keyboard.

[0011] As a further solution of the present invention, the features extracted by the peripheral feature extraction module include but are not limited to environmental perception data, obstacle detection data, task priority data, and vehicle-mounted information data. The environmental data includes ambient temperature, ambient humidity, wind speed, wind direction, rainfall, lighting, and ground conditions. The obstacle detection data includes the location, size, number, and shape of obstacles. The vehicle-mounted information data is performance data related to the vehicle acquired by the vehicle-mounted CPU.

[0012] As a further scheme of the present invention, the bridge-building engineering module and the mine-sweeping engineering module are connected with two types of engineering work data collection modules for collecting the working status data of the bridge-building engineering and the mine-sweeping engineering. The data collected by the two types of engineering work data collection modules include bridge-building work data and mine-sweeping work data. The bridge-building work data include bridge-opening cylinder data, pin-threading cylinder data, fixer data, front cantilever data, support shovel data, rear swing frame data, C and D data, security data, auxiliary arm data, continuous dismantling position data, bridge-retracting motor data, and operation mode. The mine-sweeping work data include small flip box status, launcher status, rocket status, one rope status, two rope status, three rope status, return oil blockage information, insurance status, delivery distance, standard operation mode, cover opening cylinder status, standard oil cylinder status, emergency power status, mine-sweeping plow status, and auxiliary plow status.

[0013] As a further solution of the present invention, the chassis data acquisition module and the two-type engineering work data acquisition module are connected to the vehicle CPU through a feedback module.

[0014] As a further solution of the present invention, the first deep learning model and the second deep learning model are both incremental conditional variational autoencoder models. The first deep learning model uses directly generated data instead of historical data for training. The replay strategy focuses on the combination of new tasks and historical tasks and emphasizes classification accuracy. The second machine learning model combines different incremental learning strategies to maintain the use of historical knowledge. The replay strategy makes more detailed comparisons and adjustments to new tasks and historical tasks to maintain the effectiveness of knowledge transfer.

[0015] In order to solve the technical problems existing in the prior art, the technical effect of the system proposed in the present invention is as follows: the system proposed in the present invention divides vehicle chassis data into sensor direct acquisition data and CAN bus data, and extracts key features therefrom through deep learning models and knowledge graphs to generate target features for chassis control and equipment operation. Through the data analysis module, the system combines chassis data, external perception data and on-board system information, extracts peripheral features from environmental conditions and task requirements, and performs real-time optimization in combination with chassis features. It can cope with the challenges of complex environments, and automatically coordinates and optimizes the operating strategies of vehicle chassis and auxiliary equipment while ensuring real-time and high efficiency, realizing the flexibility and dynamic adaptability of multi-task scheduling, and significantly improving the operating efficiency and safety of two types of engineering equipment vehicles in bridge-building and mine-sweeping tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of data analysis of the system proposed in the present invention;

[0017] Figure 2 It is a schematic diagram of the model of the physical prototype (instrument testing mechanism);

[0018] Figure 3 The file information page for importing the data collected for the bridging working status of the system proposed by the present invention;

[0019] Figure 4 This is a page showing the properties of the data collected for the minesweeping working status of the system proposed in the present invention;

[0020] Figure 5 This is a page for collecting the bridging working status of the system proposed in the present invention;

[0021] Figure 6 This is a minesweeping working status collection page of the system proposed by the present invention;

[0022] Figure 7 It is a display platform interface for the minesweeping operation data of the system proposed by the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] like Figures 1 to 7As shown, the present invention proposes an automatic equipment chassis data collection and analysis system based on CAN bus, including a bridge engineering module, two types of engineering equipment vehicle chassis, a minesweeping engineering module and a vehicle CPU, the two types of engineering equipment vehicle chassis are connected to the data analysis module, the data analysis module is connected to the bridge engineering module, the two types of engineering equipment vehicle chassis and the minesweeping engineering module, the bridge engineering module, the two types of engineering equipment vehicle chassis and the minesweeping engineering module are connected to the vehicle CPU, the two types of engineering equipment vehicle chassis identify the chassis data as chassis sensor direct acquisition data and CAN bus data, use the first deep learning model to obtain the first target feature from the CAN bus data, use the second deep learning module to divide the chassis sensor direct acquisition data into the first type of feature and the second type of feature according to the trigger condition, and use the first knowledge graph and the second knowledge graph to obtain pre-selected features and auxiliary features therefrom respectively. Filter features, through the third knowledge graph, use auxiliary filtering features to obtain the second target features from the pre-selected features, and then obtain the chassis target features through feature fusion, and transmit the chassis target features to the data analysis module. The data analysis module extracts peripheral features from peripheral conditions and vehicle-mounted information through the peripheral feature extraction module, and combines peripheral features with chassis target features for data analysis to obtain the coordinated optimization strategy of chassis, bridge-building engineering, and mine-sweeping engineering, and transmit it to the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles, and coordinately optimize the relevant parameters of the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles. At the same time, the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles feed back their respective status data to the vehicle CPU, and feedback and control the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles.

[0025] The system proposed in the present invention divides vehicle chassis data into sensor direct acquisition data and CAN bus data, and extracts key features from them through deep learning models and knowledge graphs to generate target features for chassis control and equipment operation. Through the data analysis module, the system combines chassis data, external perception data and on-board system information, extracts peripheral features from environmental conditions and task requirements, and performs real-time optimization based on chassis features. It can cope with the challenges of complex environments, and automatically coordinate and optimize the operating strategies of vehicle chassis and auxiliary equipment while ensuring real-time and high efficiency, thus realizing the flexibility and dynamic adaptability of multi-task scheduling, and significantly improving the operating efficiency and safety of two types of engineering equipment vehicles in bridge-building and mine-sweeping tasks.

[0026] It should be noted that the chassis of the two types of engineering equipment vehicles are connected to a chassis data acquisition module for collecting chassis data. The chassis data acquisition module classifies the data into chassis sensor direct data and CAN bus data through the embedded source information in the metadata of the collected data. The chassis sensor direct data is connected to ADC1 / DAC1 through APB1, and ADC1 / DAC1 is connected to a data source label generation module to generate data with data source label codes of "chassis, vehicle related" and "external environment, accessory equipment related", and transmit the labeled data to the second deep learning model. The features acquired by the second deep learning model trigger the classification of the first and second features by judging the numerical value of the data source label code and the preset threshold. The first type of features are chassis and vehicle related features, and the second type of features are external environment and accessory equipment related features. The first type of features are input into the first knowledge graph, and pre-selected features are retrieved. The second type of features are input into the second knowledge graph to obtain auxiliary screening features. The pre-selected features are input into the third knowledge graph, and the second target features are obtained under the constraints of the auxiliary screening features.

[0027] By introducing the data source label generation module into the chassis data acquisition module, the system can clearly divide the chassis data into chassis sensor direct data and CAN bus data, and further generate label codes for "chassis, vehicle related" and "external environment, accessory equipment related". This classification mechanism enables the data to be quickly located in terms of its source and purpose during transmission and processing, reduces unnecessary data confusion, and ensures a clear distinction between the first type of features (chassis, vehicle related) and the second type of features (external environment, accessory equipment related), thereby improving the accuracy and efficiency of feature extraction and classification.

[0028] It should be noted that the data source label generation module and the chassis data acquisition module are connected to the data analysis module through the feature fusion module. The feature fusion module fuses the first target feature and the second target feature through feature engineering to obtain the chassis target feature.

[0029] It should be noted that, as shown in Table 1, in the chassis data acquisition module, the first target features obtained by the first deep learning model of the collected CAN bus data include data 1 and data 2. The lengths of data 1 and data 2 are both 8. Data 1 includes but is not limited to brake light status, left turn light status, right turn light status, left turn switch, right turn switch, oil pressure, power assist pressure, engine water temperature and engine oil temperature. Data 2 includes but is not limited to engine speed, equipment speed, mileage this time, motorcycle hours this time. The chassis sensor direct data includes but is not limited to vehicle acceleration, vehicle angular velocity, chassis height, hydraulic cylinder working status, chassis position, tire pressure, and suspension system status.

[0030] Table 1

[0031]

[0032]

[0033] It should be noted that the first knowledge graph is constructed based on chassis status and vehicle-related data, and is used to identify and retrieve the knowledge base of vehicle chassis status and control pre-selected features. The pre-selected features are chassis height, bridge pushing position and pressure, hydraulic cylinder status, vehicle acceleration and angular velocity. The second knowledge graph is constructed based on the external environment and auxiliary equipment status, and is used to identify and retrieve the knowledge base of auxiliary features that affect task operations. The auxiliary screening features are the degree of influence of external temperature on minesweeping and bridge erection working conditions, the degree of association of obstacle distance with path planning, and the degree of influence of wind speed on working stability. The third knowledge graph combines pre-selected features and auxiliary features to generate a knowledge base of second target features for chassis control and two-type equipment operation, which can adapt to the actual working environment and task requirements to the greatest extent. The second target features are the adjustment amount of bridge pushing pressure and environmental adaptability, the path optimization adjustment amount, the chassis posture and vehicle speed that maintain chassis stability and the best working conditions for bridge erection and minesweeping. The chassis target features are the chassis posture adjustment amount, the bridge pushing and hydraulic system optimization amount, and the chassis speed and stability control amount.

[0034] Through three layers of knowledge graphs and feature fusion, the system achieves accurate processing and optimization of multi-dimensional data such as chassis status, external environment, and equipment operation. First, the first knowledge graph focuses on chassis status and vehicle control, and provides accurate chassis operation references through pre-selected features (such as chassis height, bridge pusher position, hydraulic cylinder status, etc.), ensuring that the system can quickly respond to changes in chassis status. Secondly, the second knowledge graph focuses on the external environment and auxiliary equipment, and helps the system evaluate the impact of the external environment on task operations through auxiliary screening features (such as external temperature, obstacle distance, wind speed, etc.), thereby improving the system's adaptability in complex operating environments. The introduction of the third knowledge graph realizes the organic combination of pre-selected features and auxiliary features, and the generated second target features (such as bridge pusher pressure adjustment, path optimization adjustment, chassis posture, and vehicle speed control) can optimize chassis control and equipment operation to the greatest extent. This knowledge graph fusion mechanism enables the system to not only accurately adjust the chassis status, but also adaptively adjust the operation of bridge erection and minesweeping equipment according to the dynamic changes of the external environment, ensuring efficient operation in various complex task scenarios. Through this system, multi-dimensional data can be integrated and the chassis and equipment can be adjusted in real time to ensure the stability and flexibility of the vehicle in complex environments. At the same time, the efficiency of multi-task coordination is improved, and the dynamic adaptability and operating accuracy of the two types of engineering equipment vehicles are significantly enhanced, thereby improving the overall success rate of mission execution.

[0035] It should be noted that the peripheral feature extraction module is connected to the DWA controller memory, network adapter, and hard disk controller through AHB. The peripheral feature extraction module communicates with the bus matrix through AHB to obtain vehicle CPU data. The peripheral feature extraction module is connected to APB2 through ADC2 / DAC2. APB2 is connected to USRT, UART, GPIO, PMW output, timer, and keyboard.

[0036] The benefit of this design is that the peripheral feature extraction module is connected to key devices such as the DWA controller memory, network adapter, and hard disk controller through the AHB bus, and can extract data from multiple system components efficiently and quickly. At the same time, the module can obtain the operation and status data of the vehicle CPU in real time by exchanging information with the bus matrix through AHB, ensuring that the system has sufficient real-time and responsiveness. Through the connection between ADC2 / DAC2 and APB2, the peripheral feature extraction module can directly interact with the underlying hardware (such as USRT, UART, GPIO, PWM output, timer, keyboard, etc.) to obtain external input, control signals, and device feedback. This multi-level data acquisition and intercommunication design enables the system to integrate multi-source data, optimize the extraction and analysis of peripheral features, improve the data processing efficiency and dynamic response capability of the entire system, and ensure real-time adjustment and multi-dimensional data fusion of vehicle chassis and equipment in complex tasks.

[0037] It should be noted that the features extracted by the peripheral feature extraction module include but are not limited to environmental perception data, obstacle detection data, task priority data, and vehicle information data. The environmental data includes ambient temperature, ambient humidity, wind speed, wind direction, rainfall, lighting, and ground conditions. The obstacle detection data includes the location, size, number, and shape of obstacles. The vehicle information data is vehicle-related performance data obtained through the on-board CPU.

[0038] The system can fully monitor and analyze the external environment and vehicle performance in real time by extracting environmental perception data, obstacle detection data, task priority data and vehicle information data through the peripheral feature extraction module. Environmental data (such as temperature, humidity, wind speed, wind direction, rainfall, etc.) enable the system to adaptively adjust the operation of vehicles and equipment according to external climate conditions to ensure operational stability; obstacle detection data (such as location, size, quantity, and shape) can help the system plan paths in real time, avoid collisions, and optimize the work efficiency and safety of equipment; task priority data is used to dynamically adjust task scheduling to ensure that important tasks are executed first; vehicle information data (such as vehicle performance and operating status) is provided by the vehicle CPU to ensure that the operation and working status of the vehicle are within the optimal range. This fusion and extraction of multi-dimensional data enables the system to quickly perceive external changes in complex task environments, flexibly respond to task requirements, improve real-time and dynamic adaptability, and thus solve technical problems such as complex environmental adaptability, vehicle stability, and task scheduling optimization during multi-task execution.

[0039] It should be noted that the bridge engineering module and the minesweeping engineering module are connected to two types of engineering work data collection modules for collecting the working status data of the bridge engineering and the minesweeping engineering, such as Figures 3 to 7 As shown, the data collected by the two types of engineering work data collection modules include bridge-building work data and mine-sweeping work data. The bridge-building work data include bridge-opening cylinder data, pin-threading cylinder data, fixer data, front cantilever data, support shovel data, rear swing frame data, C and D data, security data, auxiliary arm data, continuous dismantling position data, bridge-retracting motor data, and operation mode. The mine-sweeping work data include small flip box status, launcher status, rocket status, first rope status, second rope status, third rope status, oil return blockage information, insurance status, release distance, standard operation mode, cover opening cylinder status, standard oil cylinder status, emergency power status, mine-sweeping plow status, and auxiliary plow status.

[0040] Both bridge erection and minesweeping tasks are highly complex engineering operations involving multiple mechanical components and processes. The data of the bridge opening cylinder, the data of the fixture, the data of the support shovel, etc. in the bridge erection work directly affect the stability of the bridge and the accuracy of the installation. If the status of these components is not monitored and adjusted in real time, it may lead to operational errors, unstable bridges or installation failures; similarly, the status of the launcher, the status of the rocket, the release distance, etc. in the minesweeping work are also crucial, and any operational errors may lead to mission failure or safety risks. Therefore, real-time monitoring of these data can ensure that all operations meet the mission requirements and ensure the safety and success rate of the operation. In bridge erection and minesweeping tasks, the coordination between multiple components is crucial. The front cantilever and the support shovel in the bridge erection operation need to work precisely in coordination to ensure that the bridge can be safely placed; in the minesweeping operation, the linkage relationship between the minesweeping plow status, the standard operation mode and the launcher status must also be closely coordinated. Monitoring the status data of these components can ensure that each part operates normally according to the predetermined working mode and improve the efficiency of collaborative work between equipment. Complex operating environments and mission requirements require the system to have real-time adjustment capabilities. By monitoring the specific working status of bridge-building and mine-clearing equipment (such as bridge-retracting motor data, mine-clearing plow status, etc.), the system can dynamically adjust the working mode of the equipment according to environmental changes or task requirements, ensuring that the system can respond quickly and complete task optimization in tasks with high real-time requirements. Monitoring this data can help the system identify potential problems in advance. When abnormal conditions are detected in hydraulic cylinders or robotic arms, the system can quickly take countermeasures to avoid equipment damage or mission failure. This preventive monitoring can significantly improve the reliability of operations and the service life of equipment.

[0041] It should be noted that the chassis data acquisition module and the two-type engineering work data acquisition module are connected to the vehicle CPU through the feedback module.

[0042] It should be noted that both the first deep learning model and the second deep learning model are incremental conditional variational autoencoder models. The first deep learning model uses directly generated data instead of historical data for training. The replay strategy focuses on the combination of new tasks and historical tasks and emphasizes classification accuracy. The second machine learning model combines different incremental learning strategies to maintain the use of historical knowledge. The replay strategy makes more detailed comparisons and adjustments to new tasks and historical tasks to maintain the effectiveness of knowledge transfer.

[0043] The first deep learning model is trained by replacing historical data with directly generated data, and adopts a replay strategy to combine new tasks with historical tasks. This approach enables the model to quickly adapt to new tasks while retaining knowledge of historical tasks. Because the model focuses on classification accuracy, it can ensure that the system can efficiently classify and accurately judge in changing task scenarios while processing new tasks. This is particularly suitable for real-time feature extraction in vehicle chassis control and equipment operation, ensuring that the system can quickly adjust and optimize operations when processing new chassis data. The second deep learning model uses a replay mechanism combined with different incremental learning strategies, which can effectively maintain the knowledge of historical tasks and make detailed comparisons and adjustments between new and historical tasks. In this way, the model can transfer and reuse effective knowledge in historical tasks while performing new tasks, avoiding "catastrophic forgetting". This mechanism helps ensure that the system can learn and optimize for a long time in complex environments, improving the overall operational efficiency and intelligence of the system. Through the incremental learning strategy, the replay strategy of the second model can make more detailed comparisons and adjustments between new and historical tasks, enabling the system to adapt to changes in multi-task scenarios. Especially when dealing with complex bridge-building and mine-sweeping tasks, the model can dynamically adjust the knowledge of historical tasks according to the needs of new tasks, making knowledge transfer more effective, thereby improving the flexibility and dynamic adaptability of the system in a multi-task environment. Since both models adopt incremental learning and knowledge transfer strategies, the system has long-term learning capabilities and can continuously improve its understanding and optimization capabilities of tasks in multiple task iterations. This approach ensures that the system can maintain efficient operating performance and task execution stability even in the face of ever-changing tasks and environments, greatly improving the system's intelligence level and adaptability to application scenarios.

[0044] In order to clearly explain the difference between the first deep learning model and the second deep learning model, a detailed description is given through embodiments.

[0045] Example

[0046] The first deep learning model uses an incremental conditional variational autoencoder. The replay strategy is to directly generate new data to replace historical data. It focuses on combining new tasks with historical figures data. The goal is to improve classification accuracy, with the focus on optimizing the processing of new task data. The specific operation process is as follows:

[0047] Step 1: The model obtains the bridge pushing position and hydraulic cylinder pressure data of the current bridge erection task, and generates corresponding simulated historical data for combined training;

[0048] Step 2: Generate new data to replace the historical data stored in the system and directly apply it to the classification and judgment of the current task;

[0049] Step 3: By optimizing the classification algorithm, the model improves the classification accuracy of the new task data and determines in real time whether the bridge pushing state, position and chassis are suitable for the current terrain.

[0050] The second deep learning model adopts an incremental conditional variational autoencoder model, combined with a variety of incremental learning strategies, to maintain the use of historical knowledge, to make more detailed comparisons and adjustments between new tasks and historical task data, to maintain the effectiveness of knowledge transfer, and to optimize the stability of long-term learning.

[0051] When performing minesweeping tasks, the status data of the vehicle chassis and minesweeping equipment needs to be compared and adjusted with the relevant data in historical figures. The specific operation process is as follows:

[0052] Step 1: The model extracts data from the current minesweeping mission (depth of the minesweeping plow, distance to obstacles, etc.) and compares it with historical data accumulated from previous minesweeping missions.

[0053] Step 2: Based on the task-based knowledge transfer strategy, the model retains and reuses valid data from historical tasks to avoid forgetting historical experience;

[0054] Step 3, fine-tune the new tasks and historical tasks through the replay strategy to ensure that the current task is heavy and the stability of the minesweeping equipment operation can be optimized based on historical experience.

[0055] The task-based knowledge transfer strategy adopts a strategy based on knowledge distillation, selective knowledge transfer and cumulative learning. The replay strategy selects a no-forgetting generation replay strategy and a no-storage generation replay strategy. By refining and migrating the knowledge in the complex model to a lighter model, it is possible to reduce the computational burden while maintaining efficient task execution. For this technical solution, when processing bridge-building or mine-sweeping tasks, by distilling the complex knowledge of historical tasks, the system can quickly adapt to new tasks without the need for a complete historical model, thereby reducing training time and computing resource consumption. By screening and migrating historical task knowledge, "negative migration" (i.e., interference from irrelevant or erroneous knowledge) is avoided. In this embodiment, in two types of engineering tasks, in different environments of mine-sweeping or bridge-building tasks, the system can selectively migrate effective feature data (such as bridge pushing position, obstacle distance) in historical tasks to optimize the accuracy and efficiency of current task execution. By generating simulated data for historical tasks, it is ensured that the model will not forget previously learned knowledge when training for new tasks. For this technical solution, in the two types of engineering equipment tasks, this strategy can generate key data (such as hydraulic cylinder pressure, chassis posture) in historical tasks for replay, and even in new complex task environments, it can maintain the memory of important historical knowledge. In the cumulative learning strategy, as the task continues to be executed, the model can continuously learn and accumulate experience from historical tasks, and then perform better knowledge transfer in future tasks. This helps this solution to improve the adaptability and long-term performance of the system in complex and multi-task scenarios through cumulative learning, making the system more intelligent and flexible when facing changing operating conditions. The storage-free generation replay strategy reduces data storage requirements by generating historical data instead of storing all historical task data. In this solution, when processing bridge-building and mine-sweeping tasks, the system does not need to save a large amount of historical data, but only needs to reproduce the key data of historical tasks by generating models. This not only saves storage space, but also reduces the overhead of computing resources. The combination of selective knowledge transfer and the forget-free generation replay strategy ensures that the system can respond flexibly in complex and changing task environments. When processing minesweeping tasks, the system can prioritize replaying historical knowledge related to the current task (such as obstacle detection and path planning) according to environmental changes, and use selective knowledge transfer to extract previous experience to further optimize the execution of new tasks. The approach adopted in this embodiment can achieve efficient learning, knowledge accumulation and task adaptation for two types of engineering tasks. This combined strategy improves the system's learning speed for new tasks and memory retention of historical tasks, reduces dependence on historical data storage, and ensures that the system has dynamic adaptability, long-term stability and efficient resource utilization in complex task environments.

[0056] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0057] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A CAN bus-based equipment chassis data automatic collection and analysis system, including a bridge engineering module, two types of engineering equipment vehicle chassis, a minesweeping engineering module and a vehicle CPU, characterized in that: The chassis of the two types of engineering equipment vehicles are connected to a data analysis module, which is connected to the bridge engineering module, the chassis of the two types of engineering equipment vehicles and the mine-sweeping engineering module. The bridge engineering module, the chassis of the two types of engineering equipment vehicles and the mine-sweeping engineering module are connected to the vehicle CPU. The chassis of the two types of engineering equipment vehicles recognizes the chassis data as chassis sensor direct acquisition data and CAN bus data, and uses the first deep learning model to obtain the first target feature from the CAN bus data. The second deep learning module is used to divide the chassis sensor direct acquisition data into the first category of features and the second category of features according to the trigger condition, and the first knowledge graph and the second knowledge graph are used to obtain pre-selected features and auxiliary screening features therefrom respectively. Through the third knowledge graph, the auxiliary screening features are used to obtain the second target from the pre-selected features. Features, and then chassis target features are obtained through feature fusion, and the chassis target features are transmitted to the data analysis module. The data analysis module extracts peripheral features from peripheral conditions and vehicle-borne information through the peripheral feature extraction module, and performs data analysis by combining peripheral features and chassis target features to obtain the coordinated optimization strategy of chassis, bridge-building engineering, and mine-sweeping engineering, and transmits it to the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles, and coordinately optimizes the relevant parameters of the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles. At the same time, the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles feed back their respective status data to the vehicle CPU, and feedback and control are performed on the chassis, bridge-building engineering module, and mine-sweeping engineering module of the two types of engineering equipment vehicles.

2. The automatic equipment chassis data acquisition and analysis system based on CAN bus according to claim 1 is characterized in that: The chassis of the two types of engineering equipment vehicles are connected to a chassis data acquisition module for collecting chassis data. The chassis data acquisition module classifies the data into chassis sensor direct data and CAN bus data through the embedded source information in the metadata of the collected data. The chassis sensor direct data is connected to ADC1 / DAC1 through APB1. ADC1 / DAC1 is connected to a data source label generation module to generate data with data source label codes of "chassis, vehicle related" and "external environment, auxiliary equipment related", and transmit the labeled data to the second deep learning model. The features acquired by the second deep learning model trigger the classification of the first type of features and the second type of features by judging the numerical value of the data source label code and the preset threshold. The first type of features are chassis and vehicle related features, and the second type of features are external environment and auxiliary equipment related features. The first type of features are input into the first knowledge graph, and pre-selected features are retrieved. The second type of features are input into the second knowledge graph to obtain auxiliary screening features. The pre-selected features are input into the third knowledge graph, and the second target features are obtained under the constraints of the auxiliary screening features.

3. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 2 is characterized in that: The data source label generation module and the chassis data acquisition module are both connected to the data analysis module through the feature fusion module. The feature fusion module fuses the first target feature and the second target feature through feature engineering to obtain the chassis target feature.

4. The automatic equipment chassis data acquisition and analysis system based on CAN bus according to claim 1 is characterized in that: In the chassis data acquisition module, the first target features of the collected CAN bus data obtained through the first deep learning model include data 1 and data 2. The lengths of data 1 and data 2 are both 8. Data 1 includes but is not limited to brake light status, left turn signal status, right turn signal status, left turn switch, right turn switch, oil pressure, power assist pressure, engine water temperature and engine oil temperature. Data 2 includes but is not limited to engine speed, equipment speed, mileage, motorcycle hours. The chassis sensor direct data includes but is not limited to vehicle acceleration, vehicle angular velocity, chassis height, hydraulic cylinder working status, chassis position, tire pressure, and suspension system status.

5. The automatic equipment chassis data acquisition and analysis system based on CAN bus according to claim 4 is characterized in that: The first knowledge graph is constructed based on chassis status and vehicle-related data, and is used to identify and retrieve a knowledge base for vehicle chassis status and control pre-selected features. The pre-selected features are chassis height, bridge pushing position and pressure, hydraulic cylinder status, vehicle acceleration and angular velocity. The second knowledge graph is constructed based on the external environment and auxiliary equipment status, and is used to identify and retrieve a knowledge base for auxiliary features that affect task operations. The auxiliary screening features are the degree of influence of external temperature on minesweeping and bridge erection working conditions, the degree of association of obstacle distance with path planning, and the degree of influence of wind speed on working stability. The third knowledge graph combines pre-selected features and auxiliary features to generate a knowledge base for second target features for chassis control and two-type equipment operation that can adapt to the actual working environment and task requirements to the greatest extent. The second target features are the adjustment amount of bridge pushing pressure and environmental adaptability, the path optimization adjustment amount, and the chassis posture and vehicle speed that maintain chassis stability and the best working conditions for bridge erection and minesweeping. The chassis target features are the chassis posture adjustment amount, bridge pushing and hydraulic system optimization amount, and chassis speed and stability control amount.

6. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 1 is characterized in that: The peripheral feature extraction module is connected to the DWA controller memory, network adapter, and hard disk controller through AHB. The peripheral feature extraction module communicates with the bus matrix through AHB to obtain vehicle CPU data. The peripheral feature extraction module is connected to APB2 through ADC2 / DAC2. APB2 is connected to USRT, UART, GPIO, PMW output, timer, and keyboard.

7. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 6 is characterized in that: The features extracted by the peripheral feature extraction module include but are not limited to environmental perception data, obstacle detection data, task priority data, and vehicle information data. The environmental data includes ambient temperature, ambient humidity, wind speed, wind direction, rainfall, lighting, and ground conditions. The obstacle detection data includes the location, size, number, and shape of obstacles. The vehicle information data is vehicle-related performance data obtained through the on-board CPU.

8. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 1 is characterized in that: The bridge-building engineering module and the mine-sweeping engineering module are connected with two types of engineering work data collection modules for collecting the working status data of the bridge-building engineering and the mine-sweeping engineering. The data collected by the two types of engineering work data collection modules include bridge-building work data and mine-sweeping work data. The bridge-building work data includes bridge-opening cylinder data, pin-threading cylinder data, fixer data, front cantilever data, support shovel data, rear swing frame data, C and D data, security data, auxiliary arm data, continuous dismantling position data, bridge-retracting motor data, and operation mode. The mine-sweeping work data includes small flip box status, launcher status, rocket status, first rope status, second rope status, third rope status, oil return blockage information, insurance status, release distance, standard operation mode, cover opening cylinder status, standard oil cylinder status, emergency power status, mine-sweeping plow status, and auxiliary plow status.

9. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 8 is characterized in that: The chassis data acquisition module and the two-type engineering work data acquisition module are connected to the vehicle CPU through the feedback module.

10. The automatic equipment chassis data collection and analysis system based on CAN bus according to claim 1, characterized in that: Both the first deep learning model and the second deep learning model are incremental conditional variational autoencoder models. The first deep learning model uses directly generated data instead of historical data for training. The replay strategy focuses on the combination of new tasks and historical tasks and emphasizes classification accuracy. The second machine learning model combines different incremental learning strategies to maintain the use of historical knowledge. The replay strategy makes more detailed comparisons and adjustments to new tasks and historical tasks to maintain the effectiveness of knowledge transfer.

Images