Intelligent decision-making system for emergency risk prevention of mine robot

Abstract

The invention discloses a mine robot emergency risk avoiding intelligent decision-making system, and the system comprises a disaster risk prediction module which is used for carrying out the time sequence prediction based on the data of a multi-source sensor, and outputting a disaster occurrence probability and a risk comprehensive index; the dynamic risk avoiding path planning module is used for constructing a dynamic risk map and performing path planning by adopting an improved A * algorithm; according to the system, advanced early warning, dynamic path planning, hierarchical response and multi-agent coordination in a mine catastrophe environment are realized by constructing an intelligent closed-loop architecture of perception, prediction, decision making, execution and coordination. Multi-source sensor data are fused, an LSTM time sequence prediction model is adopted to carry out disaster probability prediction, an A * algorithm is improved to realize risk sensitive path planning, a hierarchical response strategy is realized based on a finite-state machine, and robot group collaborative risk avoiding is realized through a publishing and subscribing mechanism. According to the method, the survivability, the decision-making efficiency and the collaborative operation level of the mine robot in the catastrophe environment are improved.

Description

Technical Field

[0001] This invention relates to the field of mining robot technology, specifically to an intelligent decision-making system for emergency avoidance in mining robots. Background Technology

[0002] The mine disaster environment is highly uncertain and dynamically evolving. Existing mine robot hazard avoidance systems face the following technical bottlenecks.

[0003] 1. Disaster prediction is severely lagging behind. Most systems can only achieve real-time hazard detection and cannot effectively predict the evolution trend of disasters.

[0004] 2. Path planning algorithms are simplistic, often only considering the geometric shortest path and ignoring dynamically changing environmental risk factors.

[0005] 3. The emergency response mechanism is rigid, and a uniform response strategy is used to deal with different types of disasters, which lacks specificity and adaptability.

[0006] 4. Insufficient inter-system collaboration capabilities: There is a lack of effective information sharing and collaborative decision-making mechanisms between robots and between robots and the ground control center, which increases the risk of secondary accidents.

[0007] These technical deficiencies severely restrict the survivability and operational efficiency of mining robots in disaster environments. To address this, an intelligent decision-making system for emergency hazard avoidance in mining robots is proposed. Summary of the Invention

[0008] The purpose of this invention is to provide an intelligent decision-making system for emergency avoidance of mine robots, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an intelligent decision-making system for emergency avoidance of mine robots, comprising:

[0010] The disaster risk prediction module is used to perform time-series predictions based on multi-source sensor data and output the probability of disaster occurrence and a comprehensive risk index.

[0011] The dynamic risk avoidance path planning module is used to construct a dynamic risk map and perform path planning using an improved A* algorithm.

[0012] The tiered emergency response module is used to implement differentiated response strategies based on the warning level and disaster type.

[0013] The collaborative early warning unit is used to enable information sharing and collaborative decision-making among robots and with the ground control center.

[0014] Preferably, the disaster risk prediction module adopts a long short-term memory network model, receives multi-source time-series data from gas sensors, vibration sensors and infrared thermal imagers, outputs the conditional probability of disasters occurring in the future, and calculates a comprehensive disaster risk index, which integrates current risk indicators and risk change gradients.

[0015] Preferably, the formula for calculating the comprehensive disaster risk index is:

[0016] ;

[0017] in, It is an indicator of the probability of disasters occurring. As an immediate risk indicator based on current sensor readings, Indicates the risk change gradient, For learnable weight coefficients, satisfying .

[0018] Preferably, the dynamic risk avoidance path planning module discretizes the environment into a grid map, with each grid having a dynamic risk value at each moment. The path planning uses an improved A* algorithm, whose cost function integrates path length and cumulative risk, and achieves a trade-off between safety and efficiency through risk sensitivity factors.

[0019] Preferably, the improved A* algorithm is based on the traditional A* algorithm, but introduces a risk-sensitive cost function as follows:

[0020] ;

[0021] in, From the starting point to the node The actual cost of movement, For heuristic cost estimation to reach the target point, As a risk-sensitive factor, Accumulated risks for the path;

[0022] The system dynamically adjusts based on the warning level. Value, Level I warning Level II warning Level III warning This ensures that safety is prioritized in high-risk situations.

[0023] Preferably, the dynamic risk avoidance path planning module has a dynamic replanning mechanism, which immediately triggers path replanning when the risk value of any grid in the path exceeds a set threshold.

[0024] Preferably, the graded emergency response module adopts a hierarchical decision-making architecture based on finite state machines. The state transition is controlled by a rule-based reasoning system, and the decision logic is based on multi-dimensional feature vectors and Q-learning algorithm, with a predefined set of response strategies for different disaster types.

[0025] Preferably, the collaborative early warning unit adopts a distributed publish-subscribe communication architecture, the early warning message format is standardized, the communication protocol adopts the MQTT standard, the group collaborative decision-making is based on a distributed consensus algorithm, and the utility of itself and the utility of neighboring robots are integrated.

[0026] Preferably, the decision utility function of the distributed consensus algorithm is:

[0027] ;

[0028] in, For robots Decision effectiveness To assess its own utility, For the benefit of the neighbor robot, To influence the weight, This represents the autonomy coefficient.

[0029] Preferred: The system executes in the following five stages in sequence: environmental monitoring and risk prediction, intelligent decision-making and path planning, hierarchical response and execution control, collaborative early warning and group risk avoidance, effect evaluation and system recovery.

[0030] Compared with existing technologies, the beneficial effects of this invention are as follows: This system, by constructing an intelligent closed-loop architecture encompassing perception, prediction, decision-making, execution, and collaboration, achieves advanced early warning, dynamic path planning, hierarchical response, and multi-agent collaboration in mine disaster environments. It integrates multi-source sensor data, employs an LSTM time-series prediction model for disaster probability prediction, improves the A* algorithm for risk-sensitive path planning, implements a hierarchical response strategy based on a finite state machine, and achieves collaborative hazard avoidance among robot swarms through a publish-subscribe mechanism. This invention enhances the survivability, decision-making efficiency, and collaborative operation level of mine robots in disaster environments. Attached Figure Description

[0031] Figure 1 This is a block diagram of the overall system architecture of the present invention;

[0032] Figure 2 This is a flowchart of the disaster risk prediction module of the present invention;

[0033] Figure 3 This is the graded emergency response state transition diagram of the present invention;

[0034] Figure 4 This is a schematic diagram of information interaction in the collaborative early warning unit of the present invention. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0036] Please see Figure 1-4 The technical solution of this invention achieves the above objectives by constructing a complete intelligent closed-loop architecture of "perception-prediction-decision-execution-coordination". This system comprises four core modules, which are integrated into an organic whole through deep learning algorithms.

[0037] The disaster risk prediction module employs a time-series prediction model based on Long Short-Term Memory (LSTM) networks. This module receives continuous time-series data from multiple sensors (including gas sensors, vibration sensors, and infrared thermal imagers) and constructs the input sequence:

[0038] ;

[0039] in, This represents the length of the time window. LSTM networks learn long-term dependencies in sequences through their unique gating mechanism, outputting future... Conditional probability of various disasters occurring within a given time period:

[0040] ;

[0041] in, For LSTM hidden states, and For model parameters, This is the activation function. To further improve the accuracy of risk assessment, the system introduces a comprehensive disaster risk index:

[0042] ;

[0043] in, As an immediate risk indicator based on current sensor readings, Indicates the risk change gradient, For learnable weight coefficients, satisfying .

[0044] The dynamic risk avoidance path planning module introduces a dynamic risk assessment mechanism on top of traditional path planning. This module first constructs a grid-based environmental map, with each grid cell... At any moment It has a dynamic risk value:

[0045] ;

[0046] in, For the first Quantification function of risk class, The corresponding weights are used. Path planning employs an improved A* algorithm, whose cost function is redefined as:

[0047] ;

[0048] After discretization, it is represented as:

[0049] ;

[0050] in, From the starting point to the node The actual cost, For heuristic functions, As a risk-sensitive factor, This is the path segment length. The algorithm adjusts... Achieve security ( Larger) and efficiency ( (A smaller) trade-off.

[0051] The tiered emergency response module adopts a hierarchical decision-making architecture based on a finite state machine (FSM). The system state is defined as follows:

[0052] ;

[0053] State transitions are controlled by a rule-based reasoning system, whose decision-making logic is based on multi-dimensional feature vectors:

[0054] ;

[0055] in, For immediate return, Let be the state transition probability. For state value functions, This is a discount factor. The system predefines a set of response strategies for different disaster types:

[0056] ;

[0057] Each strategy It includes an ordered sequence of actions to ensure the accuracy and timeliness of the response.

[0058] The collaborative early warning unit implements swarm intelligence based on a distributed publish-subscribe architecture. The standardized format of the early warning message is as follows:

[0059] ;

[0060] The communication protocol adopts the lightweight MQTT standard, implementing message filtering and routing through a topic subscription mechanism. Group cooperative obstacle avoidance is achieved through a distributed consensus algorithm.

[0061] ;

[0062] in, For robots Decision effectiveness To assess its own utility, For the benefit of the neighbor robot, To influence the weight, This represents the autonomy coefficient.

[0063] This system covers the entire process from data collection, risk prediction, path planning, emergency response to collaborative risk avoidance, and its specific implementation process is as follows.

[0064] Module 1: Implementation of the Disaster Risk Prediction Module. The core of this module is to achieve early warning of disasters through multi-sensor data fusion and time-series prediction models. It mainly includes the following implementation process:

[0065] Step 1.1: Data Acquisition and Preprocessing. The system acquires environmental data in real time through a multi-source sensor array deployed on the robot body. Gas sensors monitor CH4 and CO concentrations at a frequency of 1 Hz. Vibration sensors monitor the vibration signal of the top plate at a sampling rate of 100 Hz. An infrared thermal imager captures the temperature field distribution at a frequency of 10 Hz. All sensor data are aligned with timestamps, filtered, and normalized to form standardized time-series data.

[0066] ;

[0067] in, The time window length is the historical data time window used for prediction.

[0068] Step 1.2: LSTM Prediction Model Implementation. The preprocessed data is input into a pre-trained Long Short-Term Memory (LSTM) network. This network contains three hidden layers, each with 128 neurons, and learns long-term dependencies in sequences through a gating mechanism. The network outputs the conditional probabilities of various disasters occurring within the next 30 seconds:

[0069] ;

[0070] in, For LSTM hidden states, and For model parameters, This is the Sigmoid activation function.

[0071] Step 1.3: Comprehensive Risk Assessment. To provide a more comprehensive risk assessment, the system calculates a comprehensive disaster risk index:

[0072] ;

[0073] in, It serves as an immediate risk indicator based on current sensor readings, reflecting the degree of danger in a timely manner; It represents the risk change gradient, reflecting the disaster development trend; For learnable weight coefficients, satisfying It is continuously optimized based on historical data.

[0074] Module 2: Implementation of the Dynamic Risk Avoidance Path Planning Module. This module plans a safe and efficient evacuation path for the robot based on real-time risk prediction results.

[0075] Step 2.1: Dynamic Risk Map Construction. The environment is discretized into 0.1m × 0.1m grids, each grid... At any moment It has a dynamic risk value:

[0076] ;

[0077] in, For the first Quantification function of risk class, These are the weighting coefficients for the corresponding risks. The risk map is updated every 0.1 seconds to ensure it reflects the latest environmental conditions.

[0078] Step 2.2: Improve the A* path planning algorithm. Based on the traditional A* algorithm, a risk-sensitive cost function is introduced:

[0079] ;

[0080] in, From the starting point to the node The actual cost of movement, For heuristic cost estimation to reach the target point, As a risk-sensitive factor, Accumulated risks along the path. The system dynamically adjusts based on the warning level. Value, Level I warning Level II warning Level III warning This ensures that safety is prioritized in high-risk situations.

[0081] Step 2.3: Dynamic Replanning Mechanism. The system performs a risk assessment of the current path every second. If the risk value of any grid cell in the path exceeds the threshold of 0.8, path replanning is triggered immediately. The replanning process is completed within 1 second, ensuring that the robot always follows the optimal path in the current environment.

[0082] Module 3: Implementation of the Tiered Emergency Response Module. This module implements precise emergency response strategies based on the warning level and disaster type.

[0083] Step 3.1: Finite State Machine Architecture. The system uses a hierarchical finite state machine to manage the robot's behavioral states:

[0084] ;

[0085] State transitions are controlled by a rule-based reasoning decision system, with decision logic based on multi-dimensional feature vectors and the Q-learning algorithm.

[0086] ;

[0087] in, For immediate return, Let be the state transition probability. For state value functions, This is the discount factor.

[0088] Module 4: Implementation of the Collaborative Early Warning Unit. This module enables efficient collaboration between robots and between robots and the ground control center.

[0089] Step 4.1: Publish-subscribe communication mechanism. A standardized alert message format is used:

[0090] ;

[0091] When a robot triggers a Level II or higher alert, it publishes a structured alert message to a pre-defined topic via a wireless mesh network within one second. Other robots and the ground control center receive these messages in real time as subscribers.

[0092] Step 4.2: Distributed Collaborative Decision-Making. Upon receiving the warning information, each robot adjusts its behavior based on a distributed consensus algorithm:

[0093] ;

[0094] in, For robots Decision effectiveness To assess its own utility, For the benefit of the neighbor robot, To influence the weight, This represents the autonomy coefficient.

[0095] Furthermore, the aforementioned emergency avoidance intelligent decision-making system executes in the following five stages in sequence, forming a complete closed-loop control process.

[0096] Phase 1: Environmental Monitoring and Risk Prediction. The system continuously monitors the mine environment using a multi-source sensor array, collecting key parameters such as gas concentration, vibration signals, and temperature distribution in real time. After preprocessing and feature extraction, the collected data is input into an LSTM-based time-series prediction model. This model analyzes historical data sequences to identify disaster evolution patterns and generate advanced risk predictions. When a parameter change trend indicates a significant increase in disaster risk, the system automatically triggers an early warning mechanism and determines the warning level based on the risk probability.

[0097] Phase Two: Intelligent Decision-Making and Path Planning. Upon receiving an early warning signal, the system immediately initiates a dynamic path planning process. First, a grid map containing real-time risk information is constructed, quantifying the risk level of each area. Then, an improved A* algorithm is used for path search. This algorithm seeks an optimal balance between path length and safety risk by optimizing the cost function. The system generates multiple candidate paths and performs a comprehensive evaluation, selecting the path with the lowest cumulative risk as the optimal avoidance route. Simultaneously, a dynamic replanning mechanism is established to ensure timely adjustments to the path strategy when environmental conditions change.

[0098] Phase Three: Tiered Response and Execution Control. Based on the warning level and disaster type, the system activates the corresponding emergency response protocol. Differentiated response strategies are implemented for different disaster characteristics: rapid evacuation procedures are initiated for widespread disasters, while on-site protective measures are taken for localized disasters. During the response, the system coordinates and controls the robot's power system, actuators, and protective devices to carry out evacuation operations in an orderly manner according to a preset sequence of actions. Simultaneously, the execution status is monitored in real time to ensure the accurate implementation of all response measures.

[0099] Phase Four: Collaborative Early Warning and Group Risk Avoidance. Simultaneously with initiating autonomous risk avoidance, the system broadcasts standardized early warning information to the ground control center and other operational robots via wireless communication. Robots receiving the warning adjust their behavior based on a distributed collaborative algorithm, achieving group obstacle avoidance and resource coordination. The ground control center simultaneously updates the environmental risk map, implements regional control and personnel scheduling, and constructs a comprehensive safety protection system.

[0100] Phase 5: Effectiveness Assessment and System Recovery. After completing the evacuation operation, the system conducts a comprehensive assessment of the entire process, including the effectiveness of the evacuation path, the appropriateness of the response measures, and the reliability of the coordination mechanism. Collected environmental data and execution records are used to optimize the predictive model and decision-making algorithm. Once the environmental risk is confirmed to have been eliminated, the system gradually returns to normal monitoring status and remains ready for response.

[0101] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent decision-making system for emergency hazard avoidance in mining robots, characterized in that, include: The disaster risk prediction module is used to perform time-series predictions based on multi-source sensor data and output the probability of disaster occurrence and a comprehensive risk index. The dynamic risk avoidance path planning module is used to construct a dynamic risk map and perform path planning using an improved A* algorithm. The tiered emergency response module is used to implement differentiated response strategies based on the warning level and disaster type. The collaborative early warning unit is used to enable information sharing and collaborative decision-making among robots and with the ground control center.

2. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 1, characterized in that: The disaster risk prediction module adopts a long short-term memory network model, receives multi-source time-series data from gas sensors, vibration sensors and infrared thermal imagers, outputs the conditional probability of disasters occurring in the future, and calculates a comprehensive disaster risk index. The index integrates current risk indicators and risk change gradients.

3. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 2, characterized in that: The formula for calculating the comprehensive disaster risk index is as follows: ； in, It is an indicator of the probability of disasters occurring. As an immediate risk indicator based on current sensor readings, Indicates the risk change gradient, For learnable weight coefficients, satisfying .

4. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 1, characterized in that: The dynamic risk avoidance path planning module discretizes the environment into a grid map, with each grid having a dynamic risk value at each moment. The path planning uses an improved A* algorithm, whose cost function integrates path length and cumulative risk, and achieves a trade-off between safety and efficiency through risk sensitivity factors.

5. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 4, characterized in that: The improved A* algorithm is based on the traditional A* algorithm, but introduces a risk-sensitive cost function: ； in, From the starting point to the node The actual cost of movement, For heuristic cost estimation to reach the target point, As a risk-sensitive factor, Accumulated risks for the path; The system dynamically adjusts based on the warning level. Value, Level I warning Level II warning Level III warning This ensures that safety is prioritized in high-risk situations.

6. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 4, characterized in that: The dynamic risk avoidance path planning module has a dynamic replanning mechanism. When the risk value of any grid in the path exceeds the set threshold, path replanning is triggered immediately.

7. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 1, characterized in that: The graded emergency response module adopts a hierarchical decision-making architecture based on finite state machines. The state transition is controlled by a rule-based reasoning system, and the decision logic is based on multi-dimensional feature vectors and Q-learning algorithm, with a predefined set of response strategies for different disaster types.

8. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 1, characterized in that: The collaborative early warning unit adopts a distributed publish-subscribe communication architecture, with standardized early warning message format, MQTT standard communication protocol, and group collaborative decision-making based on distributed consensus algorithm, integrating its own utility with that of neighboring robots.

9. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 8, characterized in that: The decision utility function of the distributed consensus algorithm is: ； in, For robots Decision effectiveness To assess its own utility, For the benefit of the neighbor robot, To influence the weight, This represents the autonomy coefficient.

10. The intelligent decision-making system for emergency avoidance of a mining robot according to claim 1, characterized in that: The system executes in the following five stages in sequence: environmental monitoring and risk prediction, intelligent decision-making and path planning, hierarchical response and execution control, collaborative early warning and group risk avoidance, and effect evaluation and system recovery.

Images