A method and system for remote control of a mine truck with hierarchical collision warning

Abstract

The present application relates to a kind of mine truck remote control grading collision early warning method and system, belong to remote control early warning technical field.Therein, the method includes: constructing collision feature correlation window, the state data of mine truck and driver is mapped as risk time series state vector, obtain mine truck control parameter set;The mine truck control parameter set is used as constraint condition, and the direction sensitivity of early warning level is energy level reorganization according to risk time series state vector, generates remote control collision early warning scheme;Potential collision interaction is used as the main driving variable of grading collision early warning, and is intervened in the regulation of early warning level trigger threshold;Build early warning feedback optimization mechanism, according to the weight coefficient in the coupling parameter set of the collision early warning of the feedback data collected is iteratively corrected, and the response priority of grading early warning signal is controlled.

Description

Technical Field

[0001] This invention belongs to the field of remote control and early warning technology, specifically relating to a graded collision early warning method and system for remote control of mining trucks. Background Technology

[0002] With the continuous expansion of production scale in open-pit and underground mines, mining trucks are increasingly widely used in mining transportation systems. Mining trucks typically undertake long-distance transportation of materials such as ore and waste rock, operating in environments characterized by steep slopes, numerous curves, high dust concentrations, and poor visibility. Under these complex conditions, mining trucks are susceptible to various factors, including environmental changes, vehicle load, and driver behavior, increasing the risk of collisions. Especially in mining transportation scenarios involving multiple vehicles working together, the lack of effective safety warning mechanisms can easily lead to vehicle collisions, rollovers, or equipment damage, severely impacting mine production efficiency and operational safety.

[0003] In recent years, with the advancement of smart mine construction, remote control technology has been gradually applied to mining truck transportation systems. Controlling mining trucks through remote control terminals can reduce the risks to personnel operating in hazardous environments to some extent. However, in remote control mode, drivers cannot directly obtain complete information about the vehicle's surrounding environment, and there may be signal delays or information loss during remote communication. This makes it difficult for drivers to make timely and accurate judgments when facing sudden risks, thereby further increasing the potential collision risk.

[0004] In existing technologies, some mining vehicle safety systems primarily rely on simple distance detection or single sensor data for collision warnings. Their warning mechanisms typically judge based solely on the distance or relative speed between the vehicle and obstacles, lacking a comprehensive analysis of driver status, vehicle handling behavior, and multi-dimensional risk factors. Furthermore, in remote control scenarios, these warning systems often fail to adequately consider the coupling relationship between driver fatigue, human-machine interaction, and vehicle handling stability, resulting in limited accuracy and reliability of the warning results. Moreover, in complex mining environments where risks change rapidly, existing warning systems struggle to dynamically adjust warning levels based on risk changes in different directions and lack mechanisms for continuous optimization of the warning model based on real-time feedback data.

[0005] Therefore, how to comprehensively utilize vehicle operation status data, driver behavior data, and environmental perception information in the remote control environment of mining trucks to construct a graded collision warning method that can dynamically reflect risk changes, and improve the response efficiency and reliability of the warning system through human-machine interaction control and feedback optimization mechanisms, has become an important technical problem that urgently needs to be solved in the field of intelligent transportation in mines. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a graded collision warning method for remote control of mining trucks. The objective of this invention can be achieved through the following technical solutions: S1: Obtain the status data of the mining truck and the driver, construct a collision feature association window, map the status data of the mining truck and the driver into a risk time-series state vector, and perform feature analysis and action semantic recognition on the human-machine interaction during operation to obtain the mining truck control parameter set. S2: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. S3: Execute remote control actions according to the remote control collision warning scheme, and dynamically adjust the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the regulation of the warning level trigger threshold; S4: Construct an early warning feedback optimization mechanism, iteratively correct the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and control the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

[0007] Specifically, the method for constructing the collision feature association window is as follows: Based on the sampling period of different data sources, the acquired mining truck and driver status data are timestamped. The timestamps are time-aligned with the risk features, a sliding time window is constructed with a preset time step, and the multi-dimensional risk features obtained by continuous sampling are aggregated to obtain the collision risk feature aggregation interval. Based on the collision risk feature aggregation interval, the vehicle operating status and driver control behavior features are correlated, and the collision feature correlation window is constructed by adaptively adjusting the window length to match the risk change rate under different operating conditions.

[0008] Specifically, the human-computer interaction includes control command interaction and status feedback interaction: The control command interaction is used to receive control commands input by the driver and to parse and process the control commands to obtain vehicle input signals. The vehicle input signals include vehicle steering control signals, acceleration control signals, and braking control signals; The status feedback interaction is used to feed back the operating status information, environmental perception information and collision warning information of the mining truck to the collision warning interaction logic network in a visual or signaling manner, and to evaluate the driver's control response time and operational stability.

[0009] Specifically, the process of establishing the graded collision warning model is as follows: Extract collision factor feature parameters within the collision feature association window; The collision factor characteristic parameters include vehicle speed change rate, steering angle fluctuation, and driver control response time. Set multi-level warning trigger thresholds, quantify the contribution of different risk characteristics, and generate a risk index sequence that characterizes the overall collision risk level; The risk index sequence is mapped to different levels of collision warning status, and combined with the risk guidance control framework in a remote control environment, a hierarchical collision warning model with adaptive update capability is established.

[0010] Specifically, the process by which the risk level fluctuation algorithm establishes the quantitative relationship between driver fatigue state and handling stability is as follows: Extract driver state features, perform statistical analysis on driver stability and operation response time, and construct a set of fatigue feature vectors characterizing driver fatigue level; Stability analysis was performed on the handling behavior data of mining trucks to extract braking response deviation and handling trajectory offset, and a set of handling stability feature vectors characterizing the smoothness of vehicle handling was constructed. The risk energy level fluctuation algorithm is used to couple and model the driver fatigue feature vector set and the handling stability feature vector set, and to evaluate the collision risk factors based on the fluctuation amplitude of energy level changes within a continuous time window, thereby generating a risk energy level sequence. A risk volatility relationship curve is constructed based on the risk energy level sequence; The risk fluctuation relationship curve is used to reflect the degree of coupling between changes in driver state and vehicle handling behavior.

[0011] Specifically, the method for generating the remote-controlled collision warning scheme is as follows: A collision risk assessment function is constructed based on the risk time-series state vector to calculate the collision risk index of the vehicle at the current moment. Based on the directional sensitivity of different warning levels, the risk contribution of the potential collision direction of the vehicle is reorganized, and multi-level warning trigger thresholds are set. It also combines the collision risk index with preset multi-level warning trigger thresholds to match different risk levels with corresponding warning trigger strategies, generating a remote-controlled collision warning scheme. The remote-controlled collision warning scheme includes warning prompt methods, frequency of interactive information output, and level of remote control intervention.

[0012] Specifically, the execution process of the remote control action is as follows: Receive vehicle control commands sent by a remote control terminal, separate and reconstruct the control fields of the vehicle control commands, and obtain the vehicle remote control signal; The vehicle remote control signals include steering control signals, speed control signals, and braking control signals; The remote control signal of the vehicle is matched and verified with the real-time operating status data of the mining truck to obtain the deviation of the control command execution. Based on the deviation of the control command execution and the current collision risk level, the risk constraint adjustment is performed on the remote control action; The risk constraint adjustment includes the execution range of remote control actions and the execution priority of control commands.

[0013] Specifically, the collision warning interaction logic network includes a risk information interaction layer and a control feedback interaction layer: The risk information interaction, etc.: The warning information corresponding to different collision risk levels is structured and encoded, and the risk warning signal is mapped into a multi-level warning warning instruction according to the preset information hierarchy rules. The warning information with differentiated warning intensity is output to the driver through the remote control interface. The control feedback interaction layer receives the driver's control response signals during remote control, performs timing verification on the execution results of control commands and the driver's response delay, generates control feedback scheduling parameters, and dynamically adjusts the warning trigger level, interaction information refresh frequency, and execution priority of remote control commands based on the control feedback scheduling parameters.

[0014] Specifically, the process of adjusting the interaction intensity is as follows: Quantitatively model the interactive behavior in the remote control system to generate an initial vector of interaction intensity; A fatigue threshold mapping function is introduced, and the initial interaction intensity vector is used as an input parameter to dynamically correct the initial interaction intensity vector, thereby obtaining the interaction intensity adjustment coefficient. The interaction intensity adjustment coefficient is used to control the interaction intensity at different risk stages.

[0015] Specifically, the execution process of the early warning feedback optimization mechanism is as follows: Collect feedback data generated during remote control; The feedback data includes the execution status of remote control commands, collision warning trigger records, and driver response time. Based on the deviation between the early warning triggering results and the actual risk evolution process, the contribution of risk impact factors is redistributed, and the collision projection change corresponding to the short-term running trajectory of mining trucks is verified. When the change in the collision projection approaches the warning level trigger threshold, the priority of the warning signal response in the high-risk direction is increased, and the execution frequency and intervention level of the remote control command are dynamically adjusted.

[0016] Specifically, the graded collision warning model also includes a risk-guided control framework: The risk guidance control framework is used to establish a dynamic guidance relationship between collision risk assessment results and remote control decisions; The dynamic guidance relationship is formed by continuously evolving the risk time-series state vector and calculating the risk gain coefficient in different motion directions to create a risk guidance vector with directional constraint characteristics. The risk guidance vector is converted into an executable control guidance quantity, and the execution amplitude and response priority of the remote control command are adjusted in real time according to the collision risk level and the driver fatigue threshold, so as to complete the control intervention in advance when the collision risk increases. The risk guidance vector includes a steering correction coefficient, a speed inhibition coefficient, and a braking intervention coefficient.

[0017] Specifically, a graded collision warning system for remotely controlled mining trucks includes: Multi-source data perception and analysis module: acquires mining truck and driver status data, constructs a collision feature association window, maps the mining truck and driver status data into a risk time-series state vector, and performs feature analysis and action semantic recognition on human-machine interaction during operation to obtain a mining truck control parameter set; Graded collision warning modeling module: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. Remote control interaction control module: Executes remote control actions according to the remote control collision warning scheme, and dynamically adjusts the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the control of the warning level trigger threshold; Early warning feedback optimization control module: Constructs an early warning feedback optimization mechanism, iteratively corrects the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and controls the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

[0018] The beneficial effects of this invention are as follows: This invention provides a graded collision warning method for remote control of mining trucks. By comprehensively analyzing the operating status data of mining trucks, driver status data, and human-machine interaction information, a collision feature association window is constructed, and multi-source data is mapped into a risk time-series state vector. This enables continuous monitoring and dynamic analysis of potential collision risks during the operation of mining trucks, improving the real-time performance and accuracy of risk identification.

[0019] By establishing a graded collision warning model and introducing a risk energy level fluctuation algorithm to quantitatively analyze the relationship between driver fatigue and vehicle handling stability, the warning model can not only reflect changes in vehicle operating status but also comprehensively consider driver behavior factors, thereby improving the comprehensiveness and reliability of collision risk assessment. Furthermore, by re-normalizing the directional sensitivity of the warning levels, the system can dynamically adjust the warning level according to risk changes in different directions, thus improving the adaptability of the warning results in complex mining environments.

[0020] During remote control, the intensity of human-machine interaction is dynamically adjusted in conjunction with the driver's fatigue threshold, and potential collision interaction is used as the main driving variable for graded collision warning. This enables the warning information to intervene in remote control decision-making in a timely manner according to changes in risk, thereby effectively reducing safety hazards caused by remote control delays or insufficient driver reaction.

[0021] By constructing an early warning feedback optimization mechanism, the coupling parameters in the collision early warning model are iteratively corrected based on the feedback data collected during actual operation. This enables the early warning system to continuously optimize its risk assessment capabilities and prioritize increasing the response level of the early warning signal when the change in collision projection approaches the early warning trigger threshold, thereby further improving the safety and stability of remote-controlled transportation of mining trucks. Attached Figure Description

[0022] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0023] Figure 1 This is a schematic diagram of the framework of a graded collision warning method for remote control of mining trucks according to the present invention.

[0024] Figure 2 This is a schematic diagram of the collision warning interactive logic network in the graded collision warning method for remote control of mining trucks of the present invention. Detailed Implementation

[0025] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0026] Please see Figure 1 A graded collision warning method for remote control of mining trucks: S1: Obtain the status data of the mining truck and the driver, construct a collision feature association window, map the status data of the mining truck and the driver into a risk time-series state vector, and perform feature analysis and action semantic recognition on the human-machine interaction during operation to obtain the mining truck control parameter set. S2: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. S3: Execute remote control actions according to the remote control collision warning scheme, and dynamically adjust the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the regulation of the warning level trigger threshold; S4: Construct an early warning feedback optimization mechanism, iteratively correct the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and control the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

[0027] In this embodiment, the method for constructing the collision feature association window is as follows: Based on the sampling period of different data sources, the acquired mining truck and driver status data are timestamped. The timestamps are time-aligned with the risk features, a sliding time window is constructed with a preset time step, and the multi-dimensional risk features obtained by continuous sampling are aggregated to obtain the collision risk feature aggregation interval. Based on the collision risk feature aggregation interval, the vehicle operating status and driver control behavior features are correlated, and the collision feature correlation window is constructed by adaptively adjusting the window length to match the risk change rate under different operating conditions.

[0028] In this embodiment, a remote control platform for mining trucks in an underground coal mine is used as an example. This mine uses multiple electric mining trucks in a confined working environment, including tunnels, turns, and intersections where multiple vehicles converge. The system enables remote driving of the trucks through a remote control center, with a focus on preventing collision risks (such as collisions with walls or between vehicles). Driver status data is obtained from a head-mounted device, while mining truck data is obtained from an onboard sensing system.

[0029] The system hardware includes an onboard computing unit, a remote server, and a driver monitoring system. The software is developed using Python, integrating TensorFlow for risk modeling and OpenCV for image parsing. The digital twin uses a Unity simulator to construct a virtual mine environment, combining real-time data to verify the early warning logic.

[0030] The specific technical solutions include: Acquire mining truck and driver status data, construct collision feature association window, map risk time-series state vector, and obtain mining truck control parameter set; The mining truck status data includes vehicle speed, braking status, and the relative position and distance between the vehicle and obstacles, which are used to characterize the dynamic operating characteristics of the mining truck during remote control. The driver state data includes driver's line of sight, operation response time, continuous operation interval, and fatigue characteristic parameters. These serve as the basic inputs for driver risk assessment, operation stability analysis, and interaction intensity adjustment in the graded collision warning model, and are used to assess the risk of driver state and operation behavior. Collision feature association window construction: timestamp marking is performed according to the sampling period, time alignment is performed using Kalman filtering fusion, the preset time step Δt=0.2s is used to construct a sliding time window, aggregate multidimensional risk features (e.g., velocity variation coefficient, turning volatility), and adaptively adjust the window.

[0031] Human-computer interaction feature analysis and action semantic recognition: Control command interaction: parsing input signals (steering: angle °, braking: pressure kPa), status feedback interaction: visual feedback (dashboard displays risk).

[0032] The risk time-series state vector is mapped as follows: [speed risk 0.25, steering risk 0.35, fatigue risk 0.15]. This yields the following set of control parameters for the mining truck: {speed change rate: 0.03 m / s², steering angle fluctuation: 3°, response time: 0.25 s}.

[0033] Establish a graded collision early warning model, establish the relationship between fatigue and stability through the risk energy level fluctuation algorithm, and generate a remote control collision early warning scheme; Establish a graded collision early warning model: extract characteristic parameters of collision factors (rate of change of speed, fluctuation of steering angle, response time), set multi-level early warning trigger thresholds, quantify the contribution, and generate a risk index sequence (index = weighted sum).

[0034] Risk level fluctuation algorithm: Extract fatigue feature vector, manipulate stability vector, couple model: level E = vector inner product, fluctuation amplitude ΔE>0.15 to assess risk, and construct risk fluctuation relationship curve.

[0035] Energy level renormalization is performed on the directional sensitivity of the warning level to generate a remote-controlled collision warning scheme: evaluation function f = exponential Sensitivity, matching threshold, solution: {Prompt: light + voice, frequency: 3s / time, intervention: low level}.

[0036] Based on the remote control collision warning scheme, execute remote control actions, dynamically adjust the interaction intensity, and intervene in threshold control; Execute remote control actions: Receive vehicle control commands, reconstruct signals, match and verify deviations, and adjust risk constraints: amplitude = initial (1 - risk 0.15), with priority sorted for high risk.

[0037] Collision warning interaction logic network: Risk information interaction layer (encoding risk, mapping prompts: low - green light, medium - yellow voice, high - red vibration), control feedback interaction layer (verifying response delay, adjusting level -0.1, frequency +1s / time, priority +1).

[0038] Interaction intensity adjustment: Quantization modeling initial vector, based on fatigue threshold mapping function f(x)=x The adjustment coefficient is obtained by exp, and the potential collision is used as the threshold for the main driving intervention risk.

[0039] Construct an early warning feedback optimization mechanism, iteratively correct weights, and control response priority; Collect feedback data (instruction status, early warning records, response time), redistribute contribution based on deviation, verify the change in collision projection, and increase priority, adjust execution frequency and intervention level when approaching the threshold.

[0040] Risk-guided control framework: Dynamic guidance relationship (evolutionary analysis vector, calculation of gain coefficient), transformed into guidance quantity, and adjustment of amplitude / priority.

[0041] In this embodiment, the human-computer interaction includes control command interaction and status feedback interaction: The control command interaction is used to receive control commands input by the driver and to parse and process the control commands to obtain vehicle input signals. The vehicle input signals include vehicle steering control signals, acceleration control signals, and braking control signals; The status feedback interaction is used to feed back the operating status information, environmental perception information and collision warning information of the mining truck to the collision warning interaction logic network in a visual or signaling manner, and to evaluate the driver's control response time and operational stability.

[0042] In this embodiment, the process of establishing the graded collision warning model is as follows: Extract collision factor feature parameters within the collision feature association window; The collision factor characteristic parameters include vehicle speed change rate, steering angle fluctuation, and driver control response time. Set multi-level warning trigger thresholds, quantify the contribution of different risk characteristics, and generate a risk index sequence that characterizes the overall collision risk level; The risk index sequence is mapped to different levels of collision warning status, and combined with the risk guidance control framework in a remote control environment, a hierarchical collision warning model with adaptive update capability is established.

[0043] In this embodiment, the process by which the risk level fluctuation algorithm establishes the quantitative relationship between driver fatigue state and handling stability is as follows: Extract driver state features, perform statistical analysis on driver stability and operation response time, and construct a set of fatigue feature vectors characterizing driver fatigue level; Stability analysis was performed on the handling behavior data of mining trucks to extract braking response deviation and handling trajectory offset, and a set of handling stability feature vectors characterizing the smoothness of vehicle handling was constructed. The risk energy level fluctuation algorithm is used to couple and model the driver fatigue feature vector set and the handling stability feature vector set, and to evaluate the collision risk factors based on the fluctuation amplitude of energy level changes within a continuous time window, thereby generating a risk energy level sequence. A risk volatility relationship curve is constructed based on the risk energy level sequence; The risk fluctuation relationship curve is used to reflect the degree of coupling between changes in driver state and vehicle handling behavior.

[0044] In this embodiment, the method for generating the remote-controlled collision warning scheme is as follows: A collision risk assessment function is constructed based on the risk time-series state vector to calculate the collision risk index of the vehicle at the current moment. Based on the directional sensitivity of different warning levels, the risk contribution of the potential collision direction of the vehicle is reorganized, and multi-level warning trigger thresholds are set. It also combines the collision risk index with preset multi-level warning trigger thresholds to match different risk levels with corresponding warning trigger strategies, generating a remote-controlled collision warning scheme. The remote-controlled collision warning scheme includes warning prompt methods, frequency of interactive information output, and level of remote control intervention.

[0045] In this embodiment, based on the acquired mining truck operating status data, driver status data, and environmental perception data, the collision risk of the mining truck during remote control is assessed in real time, and risk parameters for graded collision warning are generated.

[0046] First, a collision risk assessment function is constructed. Let the characteristic vector of the mining truck's operating state at time t be: , Where v(t) represents vehicle speed, a(t) represents vehicle acceleration, θ(t) represents vehicle steering angle, d(t) represents distance between vehicle and nearest obstacle, Δv(t) represents relative speed between vehicle and obstacle, and τ(t) represents driver control response delay.

[0047] After normalizing the above features, a collision risk assessment function is constructed to calculate the collision risk index of the vehicle at the current moment. The specific calculation formula is as follows: , Where R(t) represents the collision risk index at time t, and wi is the weight coefficient of the i-th risk feature.

[0048] After obtaining the overall risk index, in order to achieve directionally sensitive early warning, it is necessary to calculate the risk gain coefficient in different motion directions. The specific calculation formula is as follows: , Among them, G i (t) represents the risk gradient in each direction, and K i (t) represents the risk gain coefficient in the i-th direction, and m is the number of directions.

[0049] In this embodiment, the execution process of the remote control action is as follows: Receive vehicle control commands sent by a remote control terminal, separate and reconstruct the control fields of the vehicle control commands, and obtain the vehicle remote control signal; The vehicle remote control signals include steering control signals, speed control signals, and braking control signals; The remote control signal of the vehicle is matched and verified with the real-time operating status data of the mining truck to obtain the deviation of the control command execution. Based on the deviation of the control command execution and the current collision risk level, the risk constraint adjustment is performed on the remote control action; The risk constraint adjustment includes the execution range of remote control actions and the execution priority of control commands.

[0050] In this embodiment, as Figure 2 The collision warning interaction logic network shown includes a risk information interaction layer and a control feedback interaction layer: The risk information interaction, etc.: The warning information corresponding to different collision risk levels is structured and encoded, and the risk warning signal is mapped into a multi-level warning warning instruction according to the preset information hierarchy rules. The warning information with differentiated warning intensity is output to the driver through the remote control interface. The control feedback interaction layer receives the driver's control response signals during remote control, performs timing verification on the execution results of control commands and the driver's response delay, generates control feedback scheduling parameters, and dynamically adjusts the warning trigger level, interaction information refresh frequency, and execution priority of remote control commands based on the control feedback scheduling parameters.

[0051] In this embodiment, the process of adjusting the interaction intensity is as follows: Quantitatively model the interactive behavior in the remote control system to generate an initial vector of interaction intensity; A fatigue threshold mapping function is introduced, and the initial interaction intensity vector is used as an input parameter to dynamically correct the initial interaction intensity vector, thereby obtaining the interaction intensity adjustment coefficient. The interaction intensity adjustment coefficient is used to control the interaction intensity at different risk stages.

[0052] In this embodiment, the execution process of the early warning feedback optimization mechanism is as follows: Collect feedback data generated during remote control; The feedback data includes the execution status of remote control commands, collision warning trigger records, and driver response time. Based on the deviation between the early warning triggering results and the actual risk evolution process, the contribution of risk impact factors is redistributed, and the collision projection change corresponding to the short-term running trajectory of mining trucks is verified. When the change in the collision projection approaches the warning level trigger threshold, the priority of the warning signal response in the high-risk direction is increased, and the execution frequency and intervention level of the remote control command are dynamically adjusted.

[0053] In this embodiment, the graded collision warning model also includes a risk guidance control framework: The risk guidance control framework is used to establish a dynamic guidance relationship between collision risk assessment results and remote control decisions; The dynamic guidance relationship is formed by continuously evolving the risk time-series state vector and calculating the risk gain coefficient in different motion directions to create a risk guidance vector with directional constraint characteristics. The risk guidance vector is converted into an executable control guidance quantity, and the execution amplitude and response priority of the remote control command are adjusted in real time according to the collision risk level and the driver fatigue threshold, so as to complete the control intervention in advance when the collision risk increases. The risk guidance vector includes a steering correction coefficient, a speed inhibition coefficient, and a braking intervention coefficient.

[0054] This invention also provides a graded collision warning system for remotely controlled mining trucks, specifically including: Multi-source data perception and analysis module: acquires mining truck and driver status data, constructs a collision feature association window, maps the mining truck and driver status data into a risk time-series state vector, and performs feature analysis and action semantic recognition on human-machine interaction during operation to obtain a mining truck control parameter set; Graded collision warning modeling module: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. Remote control interaction control module: Executes remote control actions according to the remote control collision warning scheme, and dynamically adjusts the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the control of the warning level trigger threshold; Early warning feedback optimization control module: Constructs an early warning feedback optimization mechanism, iteratively corrects the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and controls the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A graded collision warning method for remote control of mining trucks, characterized in that, include: S1: Obtain the status data of the mining truck and the driver, construct a collision feature association window, map the status data of the mining truck and the driver into a risk time-series state vector, and perform feature analysis and action semantic recognition on the human-machine interaction during operation to obtain the mining truck control parameter set. S2: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. S3: Execute remote control actions according to the remote control collision warning scheme, and dynamically adjust the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the regulation of the warning level trigger threshold; S4: Construct an early warning feedback optimization mechanism, iteratively correct the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and control the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

2. The method according to claim 1, characterized in that, The method for constructing the collision feature association window is as follows: Based on the sampling period of different data sources, the acquired mining truck and driver status data are timestamped. The timestamps are time-aligned with the risk features, a sliding time window is constructed with a preset time step, and the multi-dimensional risk features obtained by continuous sampling are aggregated to obtain the collision risk feature aggregation interval. Based on the collision risk feature aggregation interval, the vehicle operating status and driver control behavior features are correlated, and the collision feature correlation window is constructed by adaptively adjusting the window length to match the risk change rate under different operating conditions.

3. The method according to claim 1, characterized in that, The human-computer interaction includes control command interaction and status feedback interaction: The control command interaction is used to receive control commands input by the driver and to parse and process the control commands to obtain vehicle input signals. The vehicle input signals include vehicle steering control signals, acceleration control signals, and braking control signals; The status feedback interaction is used to feed back the operating status information, environmental perception information and collision warning information of mining trucks to the collision warning interaction logic network in a visual or signal-based manner, and to evaluate the driver's control response time and operational stability.

4. The method according to claim 1, characterized in that, The process of establishing the graded collision early warning model is as follows: Extract collision factor feature parameters within the collision feature association window; The collision factor characteristic parameters include vehicle speed change rate, steering angle fluctuation, and driver control response time. Set multi-level warning trigger thresholds, quantify the contribution of different risk characteristics, and generate a risk index sequence that characterizes the overall collision risk level; The risk index sequence is mapped to different levels of collision warning status, and combined with the risk guidance control framework in a remote control environment, a hierarchical collision warning model with adaptive update capability is established.

5. The method according to claim 1, characterized in that, The process by which the risk level fluctuation algorithm establishes the quantitative relationship between driver fatigue and handling stability is as follows: Extract driver state features, perform statistical analysis on driver stability and operation response time, and construct a set of fatigue feature vectors characterizing driver fatigue level; Stability analysis was performed on the handling behavior data of mining trucks to extract braking response deviation and handling trajectory offset, and a set of handling stability feature vectors characterizing the smoothness of vehicle handling was constructed. The risk energy level fluctuation algorithm is used to couple and model the driver fatigue feature vector set and the handling stability feature vector set, and the collision risk factors are evaluated based on the fluctuation amplitude of energy level changes within a continuous time window to generate a risk energy level sequence. A risk volatility relationship curve is constructed based on the risk energy level sequence; The risk fluctuation relationship curve is used to reflect the degree of coupling between changes in driver state and vehicle handling behavior.

6. The method according to claim 1, characterized in that, The method for generating the remote-controlled collision warning scheme is as follows: A collision risk assessment function is constructed based on the risk time-series state vector to calculate the collision risk index of the vehicle at the current moment. Based on the directional sensitivity of different warning levels, the risk contribution of the potential collision direction of the vehicle is reorganized, and multi-level warning trigger thresholds are set. It also combines the collision risk index with preset multi-level warning trigger thresholds to match different risk levels with corresponding warning trigger strategies, generating a remote-controlled collision warning scheme. The remote-controlled collision warning scheme includes warning prompt methods, frequency of interactive information output, and level of remote control intervention.

7. The method according to claim 1, characterized in that, The execution process of the remote control action is as follows: Receive vehicle control commands sent by a remote control terminal, separate and reconstruct the control fields of the vehicle control commands, and obtain the vehicle remote control signal; The vehicle remote control signals include steering control signals, speed control signals, and braking control signals; The remote control signal of the vehicle is matched and verified with the real-time operating status data of the mining truck to obtain the deviation of the control command execution. Based on the deviation of the control command execution and the current collision risk level, the risk constraint adjustment is performed on the remote control action; The risk constraint adjustment includes the execution range of remote control actions and the execution priority of control commands.

8. The method according to claim 3, characterized in that, The collision warning interaction logic network includes a risk information interaction layer and a control feedback interaction layer: The risk information interaction, etc.: The warning information corresponding to different collision risk levels is structured and encoded, and the risk warning signal is mapped into a multi-level warning warning instruction according to the preset information hierarchy rules. The warning information with differentiated warning intensity is output to the driver through the remote control interface. The control feedback interaction layer receives the driver's control response signals during remote control, performs timing verification on the execution results of control commands and the driver's response delay, generates control feedback scheduling parameters, and dynamically adjusts the warning trigger level, interaction information refresh frequency, and execution priority of remote control commands based on the control feedback scheduling parameters.

9. The method according to claim 1, characterized in that, The process of adjusting the interaction intensity is as follows: Quantitatively model the interactive behavior in the remote control system to generate an initial vector of interaction intensity; A fatigue threshold mapping function is introduced, and the initial interaction intensity vector is used as an input parameter to dynamically correct the initial interaction intensity vector, thereby obtaining the interaction intensity adjustment coefficient. The interaction intensity adjustment coefficient is used to control the interaction intensity at different risk stages.

10. The method according to claim 1, characterized in that, The execution process of the early warning feedback optimization mechanism is as follows: Collect feedback data generated during remote control; The feedback data includes the execution status of remote control commands, collision warning trigger records, and driver response time. Based on the deviation between the early warning triggering results and the actual risk evolution process, the contribution of risk impact factors is redistributed, and the collision projection change corresponding to the short-term running trajectory of mining trucks is verified. When the change in the collision projection approaches the warning level trigger threshold, the priority of the warning signal response in the high-risk direction is increased, and the execution frequency and intervention level of the remote control command are dynamically adjusted.

11. The method according to claim 4, characterized in that, The graded collision warning model also includes a risk-guided control framework: The risk guidance control framework is used to establish a dynamic guidance relationship between collision risk assessment results and remote control decisions; The dynamic guidance relationship is formed by continuously evolving the risk time-series state vector and calculating the risk gain coefficient in different motion directions to create a risk guidance vector with directional constraint characteristics. The risk guidance vector is converted into an executable control guidance quantity, and the execution amplitude and response priority of the remote control command are adjusted in real time according to the collision risk level and the driver fatigue threshold, so as to complete the control intervention in advance when the collision risk increases. The risk guidance vector includes a steering correction coefficient, a speed inhibition coefficient, and a braking intervention coefficient.

12. A graded collision warning system for remote-controlled mining trucks, used to perform the method as described in any one of claims 1-11, characterized in that, include: Multi-source data perception and analysis module: acquires mining truck and driver status data, constructs a collision feature association window, maps the mining truck and driver status data into a risk time-series state vector, and performs feature analysis and action semantic recognition on human-machine interaction during operation to obtain a mining truck control parameter set; Graded collision warning modeling module: Establish a graded collision warning model, take the mining truck control parameter set as a constraint, establish the quantitative relationship between driver fatigue state and control stability through risk energy level fluctuation algorithm, and perform energy level reshaping on the directional sensitivity of the warning level according to the risk time sequence state vector to generate a remote control collision warning scheme. Remote control interaction control module: Executes remote control actions according to the remote control collision warning scheme, and dynamically adjusts the interaction intensity in combination with the driver fatigue threshold, taking potential collision interaction as the main driving variable of graded collision warning, and intervening in the control of the warning level trigger threshold; Early warning feedback optimization control module: Constructs an early warning feedback optimization mechanism, iteratively corrects the weight coefficients in the collision early warning coupling parameter set based on the collected feedback data, and controls the response priority of the graded early warning signal when the change in collision projection approaches the early warning level trigger threshold.

Images