Crawler-type forest fire fighting truck based on machine learning, control method and system

Abstract

The invention discloses a crawler-type forest fire fighting truck based on machine learning and a control method and system thereof, and the method comprises the steps: obtaining fire position data, the current position data of the fire fighting truck and the current state data of the fire fighting truck, and carrying out the data preprocessing, and obtaining standard position data; performing path calculation operation according to the standard position data and the current state data of the fire fighting truck to obtain path data; performing state conversion operation according to the path data and the current state data of the fire fighting truck to obtain state data; performing correction coefficient calculation operation by adopting a decision tree according to the state data to obtain a correction coefficient; calculating the control command data according to the correction coefficient to obtain the control command data; and according to the control command data, generating a control command and sending the control command to the forest fire fighting truck to realize accurate control of the forest fire fighting truck. The method can improve adaptability and control precision of fire fighting truck control.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire fighting vehicle control, and in particular to a tracked forest fire fighting vehicle, a control method and a system based on machine learning. Background Art

[0002] With the frequent occurrence of forest fires and the increasing complexity of the fire environment, tracked forest fire trucks play an increasingly important role in fire fighting and rescue. However, traditional fire truck control methods show obvious shortcomings when facing large-scale, complex and changeable fire environments. Forest fires are usually accompanied by problems such as dense smoke, soft ground, and limited vision, which poses a huge challenge to the precise control and efficient fire fighting of fire trucks. Therefore, how to improve the autonomous control capability of tracked forest fire trucks in complex fire environments has become a key issue that needs to be urgently solved in the current forest fire fighting field.

[0003] In order to solve the above problems, the existing technology mainly uses automatic control methods based on machine learning to improve the fire-fighting efficiency of fire trucks. For example, the smoke image of the fire scene is captured by an infrared detector, and the smoke concentration is analyzed by a machine learning algorithm to determine the location of the fire source and plan the fire-fighting path. In addition, some technologies also monitor the status information of the fire truck in real time through sensors, and combine machine learning models for path planning and speed control to improve the response speed and fire-fighting accuracy of the fire truck. However, the smoke image analysis method based on infrared detectors is easily affected by environmental climate changes, resulting in inaccurate positioning of the fire source. The positioning method of the existing technology is complex and has low accuracy. It cannot effectively solve problems such as smoke obstruction and limited field of view, which affects the overall fire-fighting efficiency of the fire truck.

[0004] In summary, the existing technology lacks flexibility and adaptability when dealing with complex fire environments, and it is difficult to meet the needs of tracked forest fire trucks for precise control. Summary of the invention

[0005] The present invention provides a tracked forest fire truck, a control method and a system based on machine learning to improve the adaptability and control accuracy of the fire truck control.

[0006] In the first aspect, in order to solve the above technical problems, the present invention provides a control method of a crawler-type forest fire truck based on machine learning, comprising:

[0007] Obtain fire location data, fire truck current location data, and fire truck current status data;

[0008] Performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data;

[0009] Performing a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data;

[0010] Performing a state conversion operation according to the path data and the current state data of the fire truck to obtain state data;

[0011] According to the state data, a correction coefficient calculation operation is performed using a decision tree to obtain a correction coefficient;

[0012] Calculating the control command data according to the correction coefficient to obtain the control command data;

[0013] According to the control command data, a control command is generated and sent to the forest fire truck to achieve precise control of the forest fire truck.

[0014] As an optional implementation manner, performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data includes:

[0015] Performing a data cleaning operation according to the fire location data and the current location data of the fire truck to obtain clean data;

[0016] Performing a coordinate conversion operation according to the cleaning data to obtain converted coordinate data;

[0017] Performing a time synchronization operation according to the coordinate data to obtain synchronization data;

[0018] A data fusion operation is performed according to the synchronization data to obtain standard position data.

[0019] As an optional implementation manner, performing a path calculation operation according to the standard position data and the current state data of the fire truck to obtain the path data includes:

[0020] According to the standard position data and the current state data of the fire truck, a state representation operation is performed to obtain position state data and fire truck state data;

[0021] Perform a DQN decision operation according to the position state data and the fire truck state data to obtain action data;

[0022] A path generation operation is performed according to the action data to obtain path data.

[0023] As an optional implementation, performing a DQN decision operation according to the position state data and the fire truck state data to obtain action data includes:

[0024] Performing a Q value prediction operation according to the position state data and the fire truck state data to obtain Q value data;

[0025] According to the Q value data, an ε-greedy strategy operation is performed to obtain action selection data;

[0026] According to the action selection data, an action execution operation is performed to obtain action data.

[0027] As an optional implementation manner, performing a Q value prediction operation according to the position state data and the fire truck state data to obtain the Q value data includes:

[0028] Input the position state data and the fire truck state data into a pre-trained DQN model, perform forward propagation calculation, and obtain network output data;

[0029] Performing a Q value extraction operation according to the network output data to obtain Q value data;

[0030] The training process of the DQN model is as follows:

[0031] Initialize the initial parameters and experience replay buffer of the main network and target network;

[0032] Performing environmental interaction and data collection according to the initial parameters to obtain empirical data;

[0033] According to the empirical data, data sampling and back propagation training are performed to obtain a trained DQN model.

[0034] As an optional implementation manner, performing a state conversion operation according to the path data and the current state data of the fire truck to obtain the state data includes:

[0035] Performing a data normalization operation according to the path data and the current state data of the fire truck to obtain normalized state data;

[0036] Performing feature extraction according to the normalized state data to obtain state feature data;

[0037] Performing one-hot encoding processing on the state feature data to obtain state data;

[0038] The state data includes current vehicle speed data, deflection angle data and target position data.

[0039] As an optional implementation manner, the correction coefficient calculation operation is performed using a decision tree according to the state data to obtain the correction coefficient, including:

[0040] Inputting the state data into a decision tree model to obtain an initial classification result;

[0041] According to the initial classification result, a feature selection operation is performed to obtain key feature data;

[0042] According to the key feature data, a decision tree classification operation is performed to obtain classification result data;

[0043] A correction coefficient calculation operation is performed according to the classification result data to obtain the correction coefficient.

[0044] As an optional implementation manner, the correction coefficient calculation operation is performed according to the classification result data to obtain the correction coefficient. The calculation formula of the correction coefficient calculation is as follows:

[0045]

[0046] Among them, K represents the correction coefficient; α, β and γ are the characteristic weight coefficients; w i is the weight of the feature at the ith position; f i is the i-th position feature; Δd ​​is the distance difference between the current position of the fire truck and the target position; Δθ is the deflection angle difference between the current direction of the fire truck and the target direction.

[0047] As an optional implementation manner, the calculating the control command data according to the correction coefficient to obtain the control command data includes:

[0048] According to the correction coefficient, a deflection angle correction calculation is performed to obtain a corrected deflection angle;

[0049] According to the correction coefficient, a speed correction calculation is performed to obtain a corrected speed;

[0050] Performing a data integration operation according to the corrected deflection angle and the corrected speed to obtain control command data;

[0051] The calculation formula for the deflection angle correction calculation is as follows:

[0052] θ new =K·Δθ max

[0053] The calculation formula for the speed correction calculation is as follows:

[0054] v new =v current +K·Δv max

[0055] Among them, θ new is the correction deflection angle; K is the correction coefficient; Δθ max is the maximum allowable deflection angle; vnew is the correction speed; v current is the current speed; Δv max is the maximum allowable speed change.

[0056] In a second aspect, the present invention provides a control system for a tracked forest fire truck based on machine learning, comprising:

[0057] A data acquisition module is used to acquire fire location data, fire truck current location data and fire truck current status data;

[0058] A data preprocessing module, used to perform data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data;

[0059] A path planning module, used to perform a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data;

[0060] A state conversion module, used to perform a state conversion operation according to the path data and the current state data of the fire truck to obtain state data;

[0061] A correction calculation module, used to perform a correction coefficient calculation operation using a decision tree according to the state data to obtain a correction coefficient;

[0062] A control correction module calculates the control command data according to the correction coefficient to obtain the control command data;

[0063] The control module generates a control command according to the control command data and sends it to the forest fire truck to achieve precise control of the forest fire truck.

[0064] In a third aspect, in order to solve the above technical problems, the present invention provides a tracked forest fire fighting vehicle based on machine learning, comprising: a vehicle head, and a trailer body modularly assembled with the vehicle head;

[0065] The end of the front of the vehicle is equipped with a cabin, and at the same time, a water gun for pressurized water spraying is installed at the front side of the end of the cabin;

[0066] At the same time, a detector is installed on the left side of the cabin by bolt fastening, and the detector is used to detect the external environmental conditions, while a drone is installed on the right side of the front end of the vehicle, and the drone is used to manually detect the surrounding environment.

[0067] As an optional embodiment, the bottom of the vehicle head is driven by a driving source and is equipped with multiple sets of front drive tracks, while the bottom of the trailer body is equipped with rear drive tracks for auxiliary support of the trailer body; the coordinated use of the front drive tracks and the rear drive tracks enables the vehicle head and the trailer body to adapt to complex terrain and provide traction for the vehicle head and the trailer body.

[0068] As an optional implementation, an unmanned aircraft cabin is embedded and installed on the right side of the cabin end, and the unmanned aircraft cabin is used in conjunction with a drone. Two sets of mechanical grippers are symmetrically arranged along the axis at the bottom of the drone, and the two sets of mechanical grippers are used in conjunction to grab objects.

[0069] The right side of the end of the trailer body is equipped with a function box, and there are multiple function boxes.

[0070] As an optional embodiment, a fire plow is installed on the left side of the bottom of the front of the vehicle, which is designed to be retractable and is used to open up fire isolation zones and clear obstacles on the route;

[0071] Specifically, the fire plow also includes a driving part for driving it and a hydraulic cylinder 2 for adjusting its position;

[0072] A ladder is arranged at the end of the towing body, wherein the bottom of the ladder is mounted to the end of the towing body through a clamping piece.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] The present invention provides a control method for a tracked forest fire truck based on machine learning, comprising: acquiring fire location data, current location data of a fire truck and current state data of the fire truck; performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data; performing a path calculation operation according to the standard location data and the current state data of the fire truck to obtain path data; performing a state conversion operation according to the path data and the current state data of the fire truck to obtain state data; performing a correction coefficient calculation operation according to the state data using a decision tree to obtain a correction coefficient; performing control command data calculation according to the correction coefficient to obtain control command data; generating a control command according to the control command data and sending it to a forest fire truck to achieve precise control of the forest fire truck.

[0075] In the present invention, by acquiring the fire location data and the current location data of the fire truck, the DQN algorithm is used for path planning, and then the fire truck travels according to the planned path and collects the current state data in real time, and transmits the state data to the main control module for state conversion to obtain state data. Next, the decision tree algorithm is used to calculate the correction coefficient of the state data to obtain the correction coefficient, and the control command data is generated according to the correction coefficient. Finally, the control command data is sent to the fire truck actuator to achieve precise control of the fire truck. Among them, the correction coefficient is dynamically calculated by the machine learning module according to the real-time state of the fire truck and the fire environment, which is more adaptable to the complex and changeable fire environment, and the path data and state data are used simultaneously in the control process to make decisions, which can achieve precise adjustment of the motion state while the fire truck is driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a flow chart of a method for controlling a crawler-type forest fire truck based on machine learning provided by an embodiment of the present invention;

[0077] Figure 2 It is a structural schematic diagram of a control system of a crawler-type forest fire truck based on machine learning provided by an embodiment of the present invention;

[0078] Figure 3 This is one of the structural assembly diagrams of a crawler-type forest fire fighting vehicle provided by an embodiment of the present invention;

[0079] Figure 4 This is a schematic diagram of the structure decomposition of a crawler-type forest fire fighting vehicle provided by an embodiment of the present invention;

[0080] Figure 5 is a top view schematic diagram of the connection between the protection frame and the adjustment shaft structure provided by an embodiment of the present invention;

[0081] Figure 6 This is the second schematic diagram of the structure decomposition of the crawler-type forest fire fighting vehicle provided by the embodiment of the present invention;

[0082] Figure 7 It is a bottom view schematic diagram of a partial structure of a trailer body provided by an embodiment of the present invention;

[0083] Figure 8 is a bottom view schematic diagram of a vehicle head structure provided by an embodiment of the present invention;

[0084] Fig. 9 This is the second schematic diagram of the structure assembly of a crawler-type forest fire fighting vehicle provided by an embodiment of the present invention;

[0085] Fig.10 It is a schematic diagram of the exploded structure of the aerial ladder and the towing machine body provided by an embodiment of the present invention;

[0086] Fig.11is a schematic cross-sectional view of a local structure of a trailer body provided by an embodiment of the present invention;

[0087] Fig.12 It is a schematic front view of the ladder structure provided by an embodiment of the present invention.

[0088] Figure numerals: 100, front of the vehicle; 101, locking assembly; 110, vehicle cabin; 120, front drive track; 130, unmanned vehicle cabin; 200, trailer body; 201, limit slot one; 202, transfer slot one; 203, snap-in slot one; 204, limit slot two; 205, transfer slot two; 206, snap-in slot two; 210, rear drive track; 220, function box; 230, protective frame; 231, connecting shaft; 240, adjusting shaft; 241, connecting shaft tube; 242, plug-in bolt; 250, hydraulic cylinder one; 251, connecting sleeve; 300, water gun; 400, fire plow; 410, hydraulic cylinder two; 420, driving unit; 500, detector; 510, lighting unit; 600, unmanned vehicle; 610, mechanical gripper; 700, ladder; 710, snap-in piece. DETAILED DESCRIPTION

[0089] With the frequent occurrence of forest fires and the increasing complexity of the fire environment, tracked forest fire trucks play an increasingly important role in fire fighting and rescue. However, traditional fire truck control methods show obvious shortcomings when facing large-scale, complex and changeable fire environments. Forest fires are usually accompanied by problems such as dense smoke, soft ground, and limited vision, which poses a huge challenge to the precise control and efficient fire fighting of fire trucks. Therefore, how to improve the autonomous control capability of tracked forest fire trucks in complex fire environments has become a key issue that needs to be urgently solved in the current forest fire fighting field.

[0090] In order to solve the above problems, the existing technology mainly uses automatic control methods based on machine learning to improve the fire-fighting efficiency of fire trucks. For example, the smoke image of the fire scene is captured by an infrared detector, and the smoke concentration is analyzed by a machine learning algorithm to determine the location of the fire source and plan the fire-fighting path. In addition, some technologies also monitor the status information of the fire truck in real time through sensors, and combine machine learning models for path planning and speed control to improve the response speed and fire-fighting accuracy of the fire truck. However, the smoke image analysis method based on infrared detectors is easily affected by environmental climate changes, resulting in inaccurate positioning of the fire source. The positioning method of the existing technology is complex and has low accuracy. It cannot effectively solve problems such as smoke obstruction and limited field of view, which affects the overall fire-fighting efficiency of the fire truck.

[0091] In summary, the existing technology lacks flexibility and adaptability when dealing with complex fire environments, and it is difficult to meet the needs of tracked forest fire trucks for precise control.

[0092] In order to solve the above-mentioned existing problems, the present invention provides a tracked forest fire truck, a control method and a system based on machine learning to improve the adaptability and control accuracy of the fire truck control.

[0093] 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.

[0094] Reference Figure 1 The first embodiment of the present invention provides a control method for a tracked forest fire truck based on machine learning, comprising the following steps:

[0095] S11, obtaining fire location data, fire truck current location data and fire truck current status data;

[0096] S12, performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data;

[0097] S13, performing a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data;

[0098] S14, performing a state conversion operation according to the path data and the current state data of the fire truck to obtain state data;

[0099] S15, performing a correction coefficient calculation operation using a decision tree according to the state data to obtain a correction coefficient;

[0100] S16, calculating the control command data according to the correction coefficient to obtain the control command data;

[0101] S17, generating a control command based on the control command data and sending it to the forest fire truck to achieve precise control of the forest fire truck.

[0102] In step S11, the fire location data, the current location data of the fire truck and the current status data of the fire truck are acquired.

[0103] It should be noted that the fire location data is collected through multi-source sensors such as infrared detectors, smoke sensors and temperature sensors to locate the fire source and ensure the accuracy and real-time nature of the data; the current location data of the fire truck is obtained through the GPS positioning system and the inertial navigation system to provide the precise location information of the fire truck in the fire scene; the current status data of the fire truck includes speed, direction, deflection angle, acceleration, etc., which are collected in real time through the on-board sensors and transmitted to the main control module. The acquisition of these data provides a reliable basis for subsequent path planning and control decisions, ensuring that the fire truck can efficiently and safely perform firefighting tasks in complex fire environments.

[0104] In step S12, data preprocessing is performed according to the fire location data and the current location data of the fire truck to obtain standard location data, including:

[0105] Performing a data cleaning operation according to the fire location data and the current location data of the fire truck to obtain clean data;

[0106] Performing a coordinate conversion operation according to the cleaning data to obtain converted coordinate data;

[0107] Performing a time synchronization operation according to the coordinate data to obtain synchronization data;

[0108] A data fusion operation is performed according to the synchronization data to obtain standard position data.

[0109] It is worth noting that the data preprocessing operation refers to the cleaning, conversion and fusion of the original data through a series of technical means to ensure the accuracy, consistency and availability of the data. This operation minimizes noise and errors in the data processing process to ensure that the important features of the data are retained and suitable for subsequent analysis. In the embodiment of the present invention, the data preprocessing operation includes four steps: data cleaning, coordinate conversion, time synchronization and data fusion. Data cleaning removes outliers and noise through filtering algorithms and anomaly detection technologies to obtain clean data; coordinate conversion uses a standard geographic coordinate conversion algorithm to convert the geographic coordinate system (such as longitude and latitude) into a local coordinate system (such as UTM coordinate system) to improve the accuracy of position calculation; time synchronization uses timestamp alignment and interpolation methods to ensure the time consistency of the fire location data and the current location data of the fire truck; data fusion uses Kalman filtering to integrate multi-source sensor information to generate high-precision standard location data. In the control of crawler forest fire trucks, the use of data preprocessing operations can effectively convert complex fire environment and fire truck status data into standardized data to reduce the computational complexity of subsequent path planning and control decisions, while ensuring the accuracy and reliability of the data.

[0110] In step S13, the path calculation operation is performed according to the standard position data and the current state data of the fire truck to obtain path data, including:

[0111] According to the standard position data and the current state data of the fire truck, a state representation operation is performed to obtain position state data and fire truck state data;

[0112] Perform a DQN decision operation according to the position state data and the fire truck state data to obtain action data;

[0113] A path generation operation is performed according to the action data to obtain path data.

[0114] It is worth noting that the state representation operation is a step in reinforcement learning, which aims to convert the raw data in the environment (such as sensor data, location information, etc.) into a format suitable for processing by machine learning models (such as DQN models). The goal of the state representation operation is to extract key features in the environment and organize them into a structured, low-dimensional and information-rich representation so that the model can learn and make decisions efficiently. For example, in the control of a tracked forest fire truck, the state representation operation helps the model to efficiently understand the state of the environment and provide a reliable basis for subsequent path planning and decision-making.

[0115] Among them, the path generation operation aims to generate specific path points based on the action data, so as to guide the fire truck to reach the target location safely and efficiently from the current position. The path generation operation converts abstract decision actions (such as left turn, right turn, acceleration, deceleration) into executable path planning to ensure that the fire truck can smoothly perform firefighting tasks in complex fire environments. For example, in the control of a tracked forest fire truck, the specific process of path generation is: according to the action data output by the DQN model, the specific control instructions are parsed; according to the parsed action data, combined with the current position and state of the fire truck, the path points at the next moment are calculated; the generated path points are smoothed to avoid sudden changes or discontinuities in the path; the generated path is verified to ensure the feasibility and safety of the path; the generated path points are output as executable path data for use by the control system of the fire truck.

[0116] It is worth noting that the DQN decision operation is performed according to the position state data and the fire truck state data to obtain action data, including:

[0117] Performing a Q value prediction operation according to the position state data and the fire truck state data to obtain Q value data;

[0118] According to the Q value data, an ε-greedy strategy operation is performed to obtain action selection data;

[0119] According to the action selection data, an action execution operation is performed to obtain action data.

[0120] It should be noted that the DQN decision operation refers to mapping the current state to the optimal action through a deep reinforcement learning model, while ensuring the intelligence and adaptability of the decision-making process, that is, the optimal action can be accurately selected from the complex environmental state to achieve the goal. This operation minimizes the decision error in the path planning process and ensures that the fire truck can reach the fire source efficiently and safely. Exemplarily, in an embodiment of the present invention, the DQN decision operation adopts DQN technology and is implemented by continuous state-action value function (Q function) calculation and strategy optimization (such as ε-greedy strategy). In the control of crawler forest fire trucks, the use of DQN decision operations can effectively map the complex fire environment state and fire truck state data to the optimal action to reduce the complexity of path planning, and adjust the decision in real time in a dynamic environment to perform more accurate path planning and motion control. Among them, the Q function is a core concept in reinforcement learning, which is used to evaluate the long-term benefits of taking an action in a given state. Specifically, the Q function represents the expected value of the future cumulative reward after taking action a in state s. The calculation of the Q function is usually based on the Bellman equation. In the DQN model (DQN), the Q function is approximated by a neural network, the network input is the state s, and the output is the Q value of each possible action. By training the neural network, the Q value of its output gradually approaches the real Q value, thereby realizing the selection of the optimal action. The ε-greedy strategy is a strategy used in reinforcement learning to balance exploration and exploitation. The ε-greedy strategy is: in most cases, the action that is currently considered to be optimal (exploitation) is selected, but other actions (exploration) are randomly selected with a certain probability to avoid falling into the local optimum. In the control of crawler forest fire trucks, Q function calculation and ε-greedy strategy work together to help fire trucks make intelligent decisions in complex fire environments and achieve efficient and safe firefighting tasks. The action execution operation converts the generated control commands (such as steering, acceleration, and deceleration) into the actions actually executed by the fire truck to ensure that the fire truck can travel efficiently and safely according to the planned path. The action execution operation transmits the control command to the actuator (such as the motor, steering system, and braking system) of the fire truck, and monitors the execution effect in real time to ensure the accuracy and stability of the action.

[0121] Wherein, performing a Q value prediction operation according to the position state data and the fire truck state data to obtain the Q value data includes:

[0122] Input the position state data and the fire truck state data into a pre-trained DQN model, perform forward propagation calculation, and obtain network output data;

[0123] Performing a Q value extraction operation according to the network output data to obtain Q value data;

[0124] The training process of the DQN model is as follows:

[0125] Initialize the initial parameters and experience replay buffer of the main network and target network;

[0126] Performing environmental interaction and data collection according to the initial parameters to obtain empirical data;

[0127] According to the empirical data, data sampling and back propagation training are performed to obtain a trained DQN model.

[0128] It should be noted that the Q-value prediction operation of the DQN model refers to mapping the current state to the Q-value of each action through a neural network, while ensuring the accuracy and stability of the prediction process, that is, the long-term benefits of each action can be accurately evaluated from the complex environmental state. This operation minimizes the prediction error in the reinforcement learning process and ensures that the agent can select the optimal action to achieve the goal. Exemplarily, in an embodiment of the present invention, the Q-value prediction operation adopts DQN technology and is implemented by forward propagation calculation (inputting state data into the network and outputting the Q-value of each action). In the control of crawler forest fire trucks, the use of Q-value prediction operations can effectively map the complex fire environment state and fire truck state data to the Q-value of each action to reduce the complexity of decision-making, and adjust the strategy in real time in a dynamic environment to perform more accurate path planning and motion control. The training process of DQN aims to gradually optimize the parameters of the neural network through interaction with the environment, so that it can accurately predict the Q-value (i.e., long-term benefits) of each action, thereby guiding the agent to make the best decision. In an embodiment of the present invention, the training process of DQN includes: initializing the main network, the target network and the experience replay buffer; the fire truck selects actions (such as turning left, turning right, accelerating, decelerating) according to the ε-greedy strategy in the fire environment, and observes the new state and reward; the experience data is stored in the experience replay buffer; a batch of data is randomly sampled from the buffer, the target Q value and the predicted Q value are calculated, and the parameters of the main network are updated; the parameters of the target network are updated regularly, and the exploration rate ∈ is gradually reduced; when the training converges, a trained DQN model is obtained for the intelligent control of the fire truck. The Q value extraction operation is to extract the Q value corresponding to each action from the output of the neural network so that the intelligent agent can select the optimal action based on these Q values. In the control of the tracked forest fire truck, the Q value extraction operation ensures that the fire truck can select the optimal action based on the Q value, thereby efficiently and safely performing the fire extinguishing task in a complex fire environment.

[0129] In step S14, the state conversion operation is performed according to the path data and the current state data of the fire truck to obtain the state data, including:

[0130] Performing a data normalization operation according to the path data and the current state data of the fire truck to obtain normalized state data;

[0131] Performing feature extraction according to the normalized state data to obtain state feature data;

[0132] Performing one-hot encoding processing on the state feature data to obtain state data;

[0133] The state data includes current vehicle speed data, deflection angle data and target position data.

[0134] It should be noted that the state conversion operation refers to converting the original path data and fire truck state data into state data suitable for the input of the decision tree model through a series of data processing steps, while ensuring the integrity and consistency of the conversion process, that is, the core information of the original data can be accurately reflected from the state data. This operation minimizes information loss during the data conversion process and ensures that the important features of the original data are retained. Exemplarily, in an embodiment of the present invention, the state conversion operation adopts data normalization, feature extraction and unique hot encoding technology, and is implemented through continuous data processing steps (normalization is used to unify the data range, feature extraction is used to filter key information, and unique hot encoding is used to process categorical variables). In the control of crawler forest fire trucks, the use of state conversion operations can effectively convert complex path data and fire truck state data into structured, low-dimensional state data to reduce the computational complexity of the model, and at the same time, when necessary, it can be combined with the original data for more accurate path correction and control decisions. Among them, the data normalization operation is implemented through the minimum-maximum scaling method to ensure that the data of different features are at the same level and avoid some features from having too much influence on the model; the feature extraction operation is implemented through automatic feature selection technology to ensure that the state feature data contains the most important information for decision-making; the one-hot encoding processing converts the categorical variables into binary vectors to ensure that the model can correctly process the categorical data.

[0135] In step S15, the correction coefficient calculation operation is performed according to the state data using a decision tree to obtain the correction coefficient, including:

[0136] Inputting the state data into a decision tree model to obtain an initial classification result;

[0137] According to the initial classification result, a feature selection operation is performed to obtain key feature data;

[0138] According to the key feature data, a decision tree classification operation is performed to obtain classification result data;

[0139] A correction coefficient calculation operation is performed according to the classification result data to obtain the correction coefficient.

[0140] It should be noted that the correction coefficient calculation operation refers to mapping the state data into the correction coefficient through the decision tree model, while ensuring the intelligence and adaptability of the calculation process, that is, the most relevant information for path planning and control decisions can be accurately extracted from the complex state data. This operation minimizes the calculation error in the correction coefficient calculation process to ensure the accuracy and rationality of the correction coefficient. Exemplarily, in an embodiment of the present invention, the correction coefficient calculation operation adopts decision tree technology and is implemented through continuous data processing steps (state data input, feature selection, classification and correction coefficient calculation). In the control of crawler forest fire trucks, the use of correction coefficient calculation operations can effectively convert complex fire environment states and fire truck state data into correction coefficients to optimize path planning and control decisions, and can be combined with original data for more precise adjustments when necessary. This operation ensures the efficiency and accuracy of data processing and provides reliable support for the intelligent control of fire trucks in complex fire environments. Among them, the decision tree model can efficiently classify the state data by recursively selecting the optimal features for node splitting; the feature selection operation selects the most important features for classification through indicators such as information gain or Gini coefficient, ensuring that the key feature data contains the most relevant information for the calculation of the correction coefficient; the correction coefficient calculation operation ensures the accuracy and rationality of the correction coefficient by combining the classification results and preset correction rules (such as adjusting the weight according to the classification results).

[0141] It is worth noting that the correction coefficient calculation operation is performed according to the classification result data to obtain the correction coefficient. The calculation formula of the correction coefficient calculation is as follows:

[0142]

[0143] Among them, K represents the correction coefficient; α, β and γ are the characteristic weight coefficients; w i is the weight of the feature at the ith position; f i is the i-th position feature; Δd ​​is the distance difference between the current position of the fire truck and the target position; Δθ is the deflection angle difference between the current direction of the fire truck and the target direction.

[0144] It should be noted that the correction coefficient calculation formula comprehensively considers the influence of position features, distance difference and deflection angle difference on the correction coefficient by weighted summation, ensuring that the calculation of the correction coefficient is both comprehensive and accurate. The feature weight coefficients α, β and γ are determined by training data or empirical values, and can dynamically adjust the contribution of different features. In the embodiment of the present invention, α, β and γ are set to 0.2, 0.3 and 0.5 respectively. Of course, according to the actual application scenario and user needs, α, β and γ can also be set to other values, which is not limited by the present invention. Position feature f i Including the relative position of the fire truck and the fire source, terrain features, obstacle distance, etc., extracted through feature selection operations; the distance difference Δd and the deflection angle difference Δθ directly reflect the deviation between the current state of the fire truck and the target state. This calculation formula can effectively combine multi-source data to generate high-precision correction coefficients, providing reliable support for subsequent path planning and control decisions.

[0145] In step S16, the control command data is calculated according to the correction coefficient to obtain the control command data, including:

[0146] According to the correction coefficient, a deflection angle correction calculation is performed to obtain a corrected deflection angle;

[0147] According to the correction coefficient, a speed correction calculation is performed to obtain a corrected speed;

[0148] Performing a data integration operation according to the corrected deflection angle and the corrected speed to obtain control command data;

[0149] The calculation formula for the deflection angle correction calculation is as follows:

[0150] θ new =K·Δθ max

[0151] The calculation formula for the speed correction calculation is as follows:

[0152] v new =v current +K·Δv max

[0153] Among them, θ new is the correction deflection angle; K is the correction coefficient; Δθ max is the maximum allowable deflection angle; v new is the correction speed; v current is the current speed; Δv max is the maximum allowable speed change.

[0154] It should be noted that the control command data calculation refers to the generation of specific control instructions through correction coefficients, while ensuring the accuracy and rationality of the calculation process, that is, the deflection angle and speed adjustment required by the fire truck can be accurately derived from the correction coefficients. This operation minimizes calculation errors in the control command generation process to ensure the accuracy and executability of the control command. In an embodiment of the present invention, the control command data calculation adopts a correction coefficient formula and is implemented through continuous calculation steps (deflection angle correction calculation, speed correction calculation, and data integration operation). In the control of tracked forest fire trucks, the use of control command data calculation can effectively convert the correction coefficient into a specific control instruction to optimize the motion state of the fire truck. At the same time, it can be dynamically adjusted in combination with real-time data when necessary to perform more accurate path tracking and fire extinguishing operations. Among them, the deflection angle correction calculation is performed by combining the correction coefficient K with the maximum allowable deflection angle Δθ max Multiply to ensure that the corrected deflection angle is within a reasonable range to avoid oversteering; the speed correction calculation is calculated by multiplying the correction coefficient K and the maximum allowable speed change Δv max Multiply and combine with the current speed v current , ensure that the correction speed is within a safe range and avoid sudden speed changes; the data integration operation ensures the integrity and executability of the control command by integrating the correction deflection angle and the correction speed into unified control command data.

[0155] In step S17, a control command is generated based on the control command data and sent to the forest fire truck to achieve precise control of the forest fire truck.

[0156] It should be noted that the control command generation operation converts the control command data (such as the corrected deflection angle and the corrected speed) into specific execution instructions to ensure that the fire truck can accurately perform the required actions. The control commands include steering instructions and speed instructions, where the steering instructions are implemented by adjusting the speed difference of the tracks, and the speed instructions are implemented by adjusting the output power of the motor. The control command sending operation transmits the instructions to the actuator of the fire truck in real time through a communication module (such as wireless communication or CAN bus) to ensure the timeliness and reliability of the instructions. In addition, during the control command generation and sending process, real-time monitoring data (such as current speed, deflection angle, position, etc.) will be combined for dynamic adjustment to ensure that the fire truck can travel efficiently and safely according to the planned path.

[0157] Reference Figure 2 The second embodiment of the present invention provides a control system for a tracked forest fire truck based on machine learning, comprising:

[0158] A data acquisition module is used to acquire fire location data, fire truck current location data and fire truck current status data;

[0159] A data preprocessing module, used to perform data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data;

[0160] A path planning module, used to perform a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data;

[0161] A state conversion module, used to perform a state conversion operation according to the path data and the current state data of the fire truck to obtain state data;

[0162] A correction calculation module, used to perform a correction coefficient calculation operation using a decision tree according to the state data to obtain a correction coefficient;

[0163] A control correction module calculates the control command data according to the correction coefficient to obtain the control command data;

[0164] The control module generates a control command according to the control command data and sends it to the forest fire truck to achieve precise control of the forest fire truck.

[0165] It should be noted that the tracked forest fire truck control system based on machine learning provided in an embodiment of the present invention is used to execute all the process steps of the tracked forest fire truck control method based on machine learning in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0166] In summary, the present invention provides a control method for a tracked forest fire truck based on machine learning, including: obtaining fire location data, current location data of a fire truck and current status data of a fire truck; performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data; performing path calculation operations according to the standard location data and the current status data of the fire truck to obtain path data; performing state conversion operations according to the path data and the current status data of the fire truck to obtain state data; performing correction coefficient calculation operations according to the state data using a decision tree to obtain correction coefficients; performing control command data calculation according to the correction coefficients to obtain control command data; generating control commands according to the control command data and sending them to the forest fire truck to achieve precise control of the forest fire truck.

[0167] In the present invention, by acquiring the fire location data and the current location data of the fire truck, the DQN algorithm is used for path planning, and then the fire truck travels according to the planned path and collects the current state data in real time, and transmits the state data to the main control module for state conversion to obtain state data. Next, the decision tree algorithm is used to calculate the correction coefficient of the state data to obtain the correction coefficient, and the control command data is generated according to the correction coefficient. Finally, the control command data is sent to the fire truck actuator to achieve precise control of the fire truck. Among them, the correction coefficient is dynamically calculated by the machine learning module according to the real-time state of the fire truck and the fire environment, which is more adaptable to the complex and changeable fire environment, and the path data and state data are used simultaneously in the control process to make decisions, which can achieve precise adjustment of the motion state while the fire truck is driving.

[0168] The embodiment of the present invention further provides a terminal device. The terminal device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a crawler-type forest fire truck control program based on machine learning. When the processor executes the computer program, the steps in the above-mentioned crawler-type forest fire truck control method based on machine learning are implemented, such as Figure 1 Alternatively, the processor implements the functions of each module / unit in the above-mentioned system embodiments when executing the computer program.

[0169] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0170] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of terminal devices and do not constitute a limitation on the terminal device. The terminal device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0171] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0172] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0173] Wherein, if the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or system capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0174] Reference Figures 3 to 12 As shown, the third embodiment of the present invention provides a crawler-type forest fire fighting vehicle based on machine learning, including a vehicle head 100, and a plurality of front drive crawlers 120 are installed at the bottom of the vehicle head 100 through a driving source;

[0175] In order to flexibly connect the front of the vehicle 100 and the towing body 200, a locking assembly 101 is installed on the right side of the end of the front of the vehicle 100, and the locking assembly 101 is composed of a driving motor, a rotating rod and a limiting plate, wherein the driving motor is installed at the bottom of the front of the vehicle 100, and one end of the rotating rod is connected to the output end of the driving motor. At the same time, the top of the rotating rod passes through the end of the front of the vehicle 100 and is connected to the limiting plate, wherein the connection between the rotating rod and the front of the vehicle 100 is flexibly connected through a bearing, thereby increasing the stability of the rotating rod.

[0176] A trailer body 200 that is modularly assembled with the front of the vehicle 100 is provided on one side; a rear drive track 210 is installed at the bottom of the trailer body 200 to assist in supporting the trailer body 200; the coordinated use of the front drive track 120 and the rear drive track 210 enables the front of the vehicle 100 and the trailer body 200 to adapt to complex terrain and provide traction for the front of the vehicle 100 and the trailer body 200.

[0177] Specifically, the front drive crawler 120 and the rear drive crawler 210 use crawlers instead of traditional tires to adapt to complex terrains, such as mountains, jungles, and muddy roads. They improve off-road capabilities and reduce the risk of slipping and getting stuck; at the same time, they can disperse the weight of the vehicle, reduce the pressure on the ground, and protect vegetation. In addition, the front drive crawler 120 and the rear drive crawler 210 support on-site steering to enhance mobility.

[0178] The left side of the bottom of the trailer body 200 is respectively provided with a limit groove 201, a transfer groove 202 and a snap-in groove 203 from bottom to top, wherein the groove cavities of the limit groove 201 and the snap-in groove 203 are offset by forty-five degrees, and the inner cavities of the limit groove 201, the transfer groove 202 and the snap-in groove 203 are interconnected, and the inner cavity of the transfer groove 202 is annular, and the shape of the limit plate of the locking assembly 101 is the same as the groove cavities of the limit groove 201 and the snap-in groove 203.

[0179] In actual use, the front of the vehicle 100 and the trailer body 200 are in a horizontal alignment state, and the limit plate is driven by the driving motor to pass through the limit slot 201 and enter the transfer slot 202. After rotating forty-five degrees clockwise in the transfer slot 202, the limit plate is inserted into the snap-in slot 203 until it is vertically aligned with the groove cavity of the snap-in slot 203, thereby completing the connection between the front of the vehicle 100 and the trailer body 200.

[0180] A function box 220 is installed on the right side of the end of the trailer body 200, wherein the function box 220 and the trailer body 200 are connected in a detachable manner, such as by snap connection and bolt positioning and fastening connection, etc., and there are multiple function boxes 220, wherein there are three function boxes 220 in this solution, and the three function boxes 220 correspond to the water tank, the foam dry powder box and the tool box from left to right, respectively, and the upper boxes of the function boxes 220 have their own functions and attributes. Firefighters can choose any combination according to task requirements to form a professional and multi-purpose fire truck, which greatly improves work efficiency and fire safety; wherein the input end of the water gun 300 is connected to the output end of the function box 220 through a pipeline, specifically, the water gun 300 is connected to the water tank in the function box 220;

[0181] As a further optimization of this solution, each functional box 220 has a built-in RFID tag to mark the box type, such as a water tank, a foam dry powder box, a tool box, and capacity parameters.

[0182] When the function box 220 is installed on the trailer body 200, the reader / writer provided on the trailer body 200 automatically identifies the box information and updates the fire extinguishing strategy through the central control system, such as giving priority to calling the water tank or the foam dry powder tank.

[0183] Among them, liquid level sensors and weight sensors are installed on the inner wall of the function box 220, and the data is transmitted to the central control system in real time. Combined with machine learning, the remaining fire extinguishing agent usage time is predicted and the fire extinguishing path is dynamically optimized.

[0184] In order to protect the function box 220, a protective frame 230 is movably provided on the outer side of the function box 220, wherein the shaft ends at the front and rear ends of the right side of the protective frame 230 are connected with connecting shafts 231, and an adjusting shaft 240 is installed on the right side of the bottom of the inner cavity of the trailer body 200, and the connection between the two ends of the adjusting shaft 240 and the inner cavity of the trailer body 200 is connected by a bearing movable connection, which assists in supporting the adjusting shaft 240, wherein the front and rear shaft ends of the adjusting shaft 240 are connected with connecting shaft cylinders 241, and the connecting shaft cylinders 241 penetrate the outside of the trailer body 200;

[0185] The end of the connecting shaft 231 away from the protective frame 230 extends to the inner cavity of the connecting shaft cylinder 241. At the same time, in order to connect the connecting shaft cylinder 241 and the connecting shaft 231, a plurality of plug bolts 242 are provided through the outer ring of the surface of the connecting shaft cylinder 241. The plug bolts 242 penetrate the connecting shaft cylinder 241 and extend to the inner cavity of the connecting shaft 231. In this way, the connecting shaft cylinder 241 and the connecting shaft 231 are fastened and connected by the plug bolts 242.

[0186] Specifically, in order to adjust the position of the protection frame 230 and facilitate the user to adjust the function box 220, a hydraulic cylinder 250 is provided at the bottom of the inner cavity of the trailer body 200, wherein a rotating shaft is installed at the fixed end of the hydraulic cylinder 250, the movable end of the rotating shaft is connected to the hydraulic cylinder 250, and the fixed end of the rotating shaft is connected to the bottom wall of the inner cavity of the trailer body 200, and the output end of the hydraulic cylinder 250 is connected to a connecting sleeve 251, and the connecting sleeve 251 is sleeved on the surface of the adjusting shaft 240;

[0187] In actual use, the hydraulic cylinder 250 can drive the adjusting shaft 240 to rotate, and then drive the connecting shaft 231 to rotate through the plug-in bolt 242 and the connecting shaft cylinder 241, and finally drive the protective frame 230 to flip, and its flipping state is as follows: Figure 2 , Figure 4 , Figure 8 and Fig. 9 shown.

[0188] The end of the front of the vehicle 100 is equipped with a cabin 110, which is equipped with a four-person cockpit and ergonomically designed to facilitate the actual needs of firefighters during the rescue process. The surface of the cabin 110 is equipped with a red warning paint, an LED lighting system and a warning flash light.

[0189] Specifically, the vehicle cabin 110 body adopts a fire-resistant coating and a heat-insulating layer to withstand high temperature environments.

[0190] At the same time, a water gun 300 for pressurized water spraying is installed at the front end of the cabin 110. The water gun 300 has the characteristics of high water flow, high pressure, high precision and wide-angle spray, and is used to extinguish the fire.

[0191] A detector 500 is fastened with bolts on the left side of the cabin 110, and the detector 500 is used to detect the external environment. A lighting unit 510 is installed at the front and rear ends of the left side of the cabin 110 to assist in lighting the surrounding environment when extinguishing a fire.

[0192] Specifically, the detector 500 integrates environmental monitoring sensors, such as temperature, oxygen, and smoke sensors to monitor the fire environment in real time; the wind direction sensor assists in predicting the direction of fire spread, and is combined with Example 1 and Example 2 to achieve precise control of the tracked fire truck.

[0193] A drone 600 is mounted on the right side of the end of the front of the vehicle 100, and the drone 600 is used to manually detect the surrounding environment; an unmanned aircraft cabin 130 is embedded and installed on the right side of the end of the cabin 110, and the unmanned aircraft cabin 130 is used in conjunction with the drone 600 to assist in positioning the drone 600; at the same time, two groups of mechanical grippers 610 are symmetrically arranged along the axis at the bottom of the drone 600, and the two groups of mechanical grippers 610 are used in conjunction to grab objects; wherein, the drone 600 and the mechanical gripper 610 are connected in a detachable manner, such as threaded connection or snap-on connection, and the drone 600 body is connected to the detachable mechanical gripper 610, so that when needed, the gripper can carry and deliver a certain amount of fire extinguishing agent or emergency relief materials.

[0194] A fire-proof plow 400 is installed on the left side of the bottom of the front of the vehicle 100. The fire-proof plow 400 is located below the front of the vehicle 100 and is designed to be retractable. It is used to open up fire-proof isolation zones and clear obstacles on the route, such as shrubs and dead branches. The fire-proof plow 400 is electrically driven to avoid fuel pollution and supports adjustment.

[0195] like Figure 6 As shown, specifically, the fire plow 400 also includes a driving part 420 for driving it and a hydraulic cylinder 410 for adjusting its own position; wherein, the output end of the hydraulic cylinder 410 is connected to the driving part 420 through a connecting plate, and the driving part 420 is fastened to the surface of the connecting plate by bolts, and the connecting plate is slidably installed on the bottom wall of the vehicle head 100, and the slide plate and the vehicle head 100 are connected by a slide groove opened obliquely in the vehicle head 100.

[0196] The driving part 420 is composed of a motor and a synchronous pulley, the motor is installed on the surface of the connecting plate, and the synchronous pulleys are multiple and are respectively installed on the output end of the fire plow 400;

[0197] Among them, there are two groups of synchronous pulleys, a single group is two synchronous pulleys used together, specifically, a single fire plow 400 and the synchronous pulley on the motor are a group, and two adjacent synchronous pulleys are connected by a synchronous belt drive, so that when the motor on the driving part 420 is working, it can drive the corresponding fire plow 400 to rotate through the corresponding synchronous pulley.

[0198] As a further optimization of this solution, the driving unit 420 of the fire plow 400 and the hydraulic cylinder 2 410 are integrated with an environmental adaptive control system, specifically including: installing a lidar sensor at the front end of the fire plow 400 to scan the density and height of vegetation in front in real time: the machine learning model predicts the downward force and rotation speed required for the fire plow 400 based on vegetation data, automatically adjusts the position of the fire plow 400 through the hydraulic cylinder 2 410, and synchronously controls the motor output power of the driving unit 420.

[0199] At the same time, combined with the wind direction data of the detector 500, an isolation zone is automatically generated to open up a path, giving priority to blocking the direction of fire spread.

[0200] To provide high altitude rescue, such as Figures 9 to 12 As shown, a ladder 700 is set up at the end of the towing body 200, wherein the bottom of the ladder 700 is mounted to the end of the towing body 200 through a clamping member 710;

[0201] Specifically, in order to flexibly install the aerial ladder 700 on the trailer body 200, the left side of the end of the trailer body 200 is respectively provided with a second limiting groove 204, a second transfer groove 205 and a second clamping groove 206 from top to bottom, wherein the slots of the second limiting groove 204 and the second clamping groove 206 are staggered at 45 degrees, and the inner cavities of the second limiting groove 204, the second transfer groove 205 and the second clamping groove 206 are interconnected, and the shape of the clamping member 710 is the same as the groove shape of the second limiting groove 204 and the second clamping groove 206, and the inner cavity of the second transfer groove 205 is annular disc-shaped;

[0202] When installing the aerial ladder 700 on the towing body 200, when the clamping piece 710 is inserted into the second limiting groove 204, the clamping piece 710 is inserted into the second limiting groove 204 and enters into the second transfer groove 205 under the force of gravity, and then the clamping piece 710 rotates 45 degrees clockwise in the second transfer groove 205 until it is vertically corresponding to the inner cavity of the second clamping groove 206, and then is inserted into the second clamping groove 206 under the force of gravity, thereby completing the installation of the aerial ladder 700;

[0203] In this way, the water gun 300 or the ladder 700 can be selected according to actual needs.

[0204] In actual use, the ladder 700 is used for fire fighting and rescue, rescue of high-altitude accidents, etc., in order to support the firefighters' operations at height;

[0205] At the same time, the Ladder 700 has automatic extension, recovery system, basket rotation function and the flexibility of free combination to facilitate firefighting tasks in different situations.

[0206] The working principle of the crawler forest fire truck based on machine learning is as follows: the front of the vehicle 100 is driven in coordination with the rear driving crawler 210 of the trailer body 200 through the front driving crawler 120 to adapt to complex terrain and provide traction. The two are connected in a modular manner through the locking assembly 101, wherein the driving motor drives the rotating rod to drive the limit plate to pass through the limit slot 201 and the transfer slot 202 of the trailer body 200 in sequence and rotate to snap into the snap-in slot 203 to complete the locking; the protective frame 230 drives the adjusting shaft 240 through the hydraulic cylinder 250 to drive the connecting shaft cylinder 241 and the connecting shaft 231 to rotate to achieve angle adjustment; the front of the vehicle 100 is equipped with a protective The fire plow 400 cooperates with the driving unit 420 through the hydraulic cylinder 410 to simultaneously open up a fire isolation zone; the cabin 110 integrates the detector 500 and the drone 600 to form a multi-dimensional monitoring system, and the drone 600 performs the material delivery task through the mechanical gripper 610; the towing body 200 cooperates with the limit slot 204, the transfer slot 205, and the clamping slot 206 to quickly install the ladder 700 to achieve high-altitude rescue function. Each system integrates environmental monitoring data, fire extinguishing agent status and terrain information through the central control system, and combines the machine learning model to adjust the fire extinguishing strategy and travel path in real time to complete efficient firefighting operations.

[0207] It should be noted that the system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the system embodiment provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0208] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A control method for a tracked forest fire truck based on machine learning, characterized in that: Executed by the server, including: Obtain fire location data, fire truck current location data, and fire truck current status data; Performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data; Performing a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data; Performing a state conversion operation according to the path data and the current state data of the fire truck to obtain state data; According to the state data, a correction coefficient calculation operation is performed using a decision tree to obtain a correction coefficient; Calculating the control command data according to the correction coefficient to obtain the control command data; According to the control command data, a control command is generated and sent to the forest fire truck to achieve precise control of the forest fire truck.

2. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 1 is characterized in that: The method of performing data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data includes: Performing a data cleaning operation according to the fire location data and the current location data of the fire truck to obtain clean data; Performing a coordinate conversion operation according to the cleaning data to obtain converted coordinate data; Performing a time synchronization operation according to the coordinate data to obtain synchronization data; A data fusion operation is performed according to the synchronization data to obtain standard position data.

3. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 1 is characterized in that: The step of performing a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data includes: According to the standard position data and the current state data of the fire truck, a state representation operation is performed to obtain position state data and fire truck state data; Perform a DQN decision operation according to the position state data and the fire truck state data to obtain action data; A path generation operation is performed according to the action data to obtain path data.

4. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 3 is characterized in that: The step of performing a DQN decision operation according to the position state data and the fire truck state data to obtain action data includes: Performing a Q value prediction operation according to the position state data and the fire truck state data to obtain Q value data; According to the Q value data, an ε-greedy strategy operation is performed to obtain action selection data; According to the action selection data, an action execution operation is performed to obtain action data.

5. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 4 is characterized in that: The step of performing a Q value prediction operation according to the position state data and the fire truck state data to obtain the Q value data includes: Input the position state data and the fire truck state data into a pre-trained DQN model, perform forward propagation calculation, and obtain network output data; Performing a Q value extraction operation according to the network output data to obtain Q value data; The training process of the DQN model is as follows: Initialize the initial parameters and experience replay buffer of the main network and target network; Performing environmental interaction and data collection according to the initial parameters to obtain empirical data; According to the empirical data, data sampling and back propagation training are performed to obtain a trained DQN model.

6. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 1, characterized in that: The state conversion operation is performed according to the path data and the current state data of the fire truck to obtain the state data, including: Performing a data normalization operation according to the path data and the current state data of the fire truck to obtain normalized state data; Performing feature extraction according to the normalized state data to obtain state feature data; Performing one-hot encoding processing on the state feature data to obtain state data; The state data includes current vehicle speed data, deflection angle data and target position data.

7. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 1 is characterized in that: The step of calculating the correction coefficient using a decision tree according to the state data to obtain the correction coefficient includes: Inputting the state data into a decision tree model to obtain an initial classification result; According to the initial classification result, a feature selection operation is performed to obtain key feature data; According to the key feature data, a decision tree classification operation is performed to obtain classification result data; A correction coefficient calculation operation is performed according to the classification result data to obtain the correction coefficient.

8. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 7 is characterized in that: According to the classification result data, a correction coefficient calculation operation is performed to obtain a correction coefficient. The calculation formula for the correction coefficient calculation is as follows: Among them, K represents the correction coefficient; α, β and γ are the characteristic weight coefficients; w i is the weight of the feature at the ith position; f i is the i-th position feature; Δd ​​is the distance difference between the current position of the fire truck and the target position; Δθ is the deflection angle difference between the current direction of the fire truck and the target direction.

9. The method for controlling a crawler-type forest fire truck based on machine learning according to claim 1, characterized in that: The step of calculating the control command data according to the correction coefficient to obtain the control command data comprises: According to the correction coefficient, a deflection angle correction calculation is performed to obtain a corrected deflection angle; According to the correction coefficient, a speed correction calculation is performed to obtain a corrected speed; Performing a data integration operation according to the corrected deflection angle and the corrected speed to obtain control command data; The calculation formula for the deflection angle correction calculation is as follows: i new =K·Δθ max The calculation formula for the speed correction calculation is as follows: v new =v current +K·Δv max Among them, θ new is the correction deflection angle; K is the correction coefficient; Δθ max is the maximum allowable deflection angle; v new is the correction speed; v current is the current speed; Δv max is the maximum allowable speed change.

10. A crawler forest fire truck control system based on machine learning, characterized in that: include: A data acquisition module is used to acquire fire location data, fire truck current location data and fire truck current status data; A data preprocessing module, used to perform data preprocessing according to the fire location data and the current location data of the fire truck to obtain standard location data; A path planning module, used to perform a path calculation operation according to the standard position data and the current state data of the fire truck to obtain path data; A state conversion module, used to perform a state conversion operation according to the path data and the current state data of the fire truck to obtain state data; A correction calculation module, used to perform a correction coefficient calculation operation using a decision tree according to the state data to obtain a correction coefficient; A control correction module calculates the control command data according to the correction coefficient to obtain the control command data; The control module generates a control command according to the control command data and sends it to the forest fire truck to achieve precise control of the forest fire truck.

11. A crawler forest fire truck based on machine learning, characterized in that: include: A vehicle head (100), and a trailer body (200) modularly assembled with the vehicle head (100); The end of the vehicle head (100) is equipped with a vehicle cabin (110), and at the same time, a water gun (300) for pressurized water spraying is installed at the front side of the end of the vehicle cabin (110); At the same time, a detector (500) is mounted on the left side of the vehicle cabin (110) by bolt fastening, and the detector (500) is used to detect external environmental conditions, while a drone (600) is mounted on the right side of the end of the vehicle front (100), and the drone (600) is used to manually detect the surrounding environment.

12. The crawler-type forest fire fighting vehicle based on machine learning according to claim 11, characterized in that: The bottom of the vehicle head (100) is driven by a driving source and is equipped with a plurality of front drive tracks (120), while the bottom of the trailer body (200) is equipped with a rear drive track (210) for assisting in supporting the trailer body (200); the coordinated use of the front drive tracks (120) and the rear drive tracks (210) enables the vehicle head (100) and the trailer body (200) to adapt to complex terrain and provide traction for the vehicle head (100) and the trailer body (200).

13. The crawler-type forest fire fighting vehicle based on machine learning according to claim 12, characterized in that: An unmanned aircraft cabin (130) is embedded and installed on the right side of the end of the vehicle cabin (110), and the unmanned aircraft cabin (130) is used in conjunction with the unmanned aircraft (600). Two sets of mechanical grippers (610) are symmetrically arranged along the axis at the bottom of the unmanned aircraft (600), and the two sets of mechanical grippers (610) are used in conjunction with each other to grab objects. The right side of the end of the trailer body (200) is equipped with a function box (220), and there are multiple function boxes (220).

14. The crawler-type forest fire fighting vehicle based on machine learning according to claim 13, characterized in that: The left side of the bottom of the vehicle head (100) is equipped with a fire plow (400), which is designed to be retractable and is used to open up fire isolation zones and clear obstacles on the travel route; Specifically, the fire plow (400) further includes a driving unit (420) for driving the plow and a second hydraulic cylinder (410) for adjusting its position; A ladder (700) is mounted on the end of the towing body (200), wherein the bottom of the ladder (700) is mounted to the end of the towing body (200) via a clamping member (710).

Images