Intelligent metasurface assisted coal mine deep well hoist communication system and method

Abstract

A coal mine deep well hoist communication system and method based on intelligent metasurface assistance, the system comprises an Internet of Things device, an edge computing node, a cloud server and an intelligent metasurface, the Internet of Things device comprises a tension sensor, a weighing sensor and an acceleration sensor; the tension sensor, the weighing sensor and the acceleration sensor are connected with the edge computing node through the intelligent metasurface, and the edge computing node is connected with the cloud server through a wired transmission medium. The method comprises the following steps: arranging the intelligent metasurface and constructing an intelligent communication system; modeling the energy loss and the calculation time delay in the calculation offloading process of the intelligent communication system; constructing a comprehensive benefit function integrating the energy efficiency and the calculation time delay; taking the maximization of the comprehensive benefit function as the optimization target of the calculation offloading strategy; constructing a security reinforcement learning model; and realizing the optimal calculation offloading benefit. The system and method can ensure the stability of the wireless data link inside the hoist and improve the transmission efficiency of data.

Description

Technical Field

[0001] This invention belongs to the fields of coal mine automation, intelligent communication and industrial Internet of Things technology, specifically a communication system and method for deep well hoists in coal mines based on intelligent metasurface assistance. Background Technology

[0002] As a core component of the mine production system, the safety and efficiency of deep-shaft hoisting systems in coal mines directly impact the stable operation of production processes and the safety of personnel. To address the potential risks of overload and rope breakage caused by factors such as terminal load overload, slack rope jamming, operational errors, and maintenance negligence during hoisting operations, academia and industry are currently exploring innovative solutions aimed at significantly improving the reliability and safety of hoisting systems. Low-power wide-area network (LPWAN) technologies, especially LoRa, have demonstrated significant application potential in mine remote monitoring systems due to their advantages of low power consumption and long-distance communication. LoRa technology effectively utilizes unlicensed radio frequency bands to achieve efficient wireless data transmission, significantly reducing system energy consumption and communication latency. Furthermore, combining it with mobile edge computing (MEC) technology to bring data processing capabilities to the edge of the device not only accelerates real-time data processing and analysis but also significantly improves system response speed, laying a solid foundation for the intelligent development of the Internet of Things (IoT) in mines. However, in the complex metallic environment of underground coal mines, the attenuation of LoRa signals becomes a key factor limiting its application effectiveness.

[0003] To address this technological bottleneck, Intelligent Metasurface (RIS) technology has been introduced into mine communication systems as a cutting-edge wireless communication solution. RIS achieves intelligent adaptive adjustment of beamforming by precisely controlling the phase shift parameters of the reflector units and integrating advanced reinforcement learning algorithms, effectively enhancing signal strength and communication stability, and providing an innovative and efficient solution for wireless communication in coal mines. However, the complex and ever-changing mine environment, with electromagnetic interference, coal dust, water mist, and large mobile devices posing severe challenges to wireless communication, makes accurately acquiring real-time communication models and channel state information a pressing problem. Given the limitations of traditional optimization methods in dealing with the dynamic changes and uncertainties in the mine environment, introducing more advanced optimization algorithms and artificial intelligence technologies has become an inevitable choice. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a communication system and method for deep coal mine hoists based on intelligent metasurface assistance. This system, utilizing LoRa wireless communication technology and combining the signal enhancement characteristics and adjustable reflection properties of intelligent metasurfaces, ensures the stability of the wireless data link within the hoist, thereby improving data transmission efficiency and ensuring the reliability and security of data transmission. This facilitates real-time data fusion processing by edge computing nodes and cloud servers, enabling comprehensive and reliable monitoring of the hoist system status. The method is highly intelligent and simple to implement. Addressing the signal attenuation and communication instability caused by the dense metal components in underground coal mines, it is the first to apply intelligent metasurface technology to the unique environment of hoist roadways. Furthermore, through continuous interactive learning with the environment, deep reinforcement learning algorithms are used to continuously optimize the configuration strategy of the intelligent metasurface, precisely controlling the propagation path, amplitude, and phase of electromagnetic waves. This significantly improves the strength and transmission stability of the wireless signal, effectively adapting to dynamic changes such as the rapid movement of the hoist, and ensuring the long-term stability and sustainable optimization capability of the communication quality within the hoist.

[0005] To achieve the above objectives, the present invention provides a communication system for a deep coal mine hoist based on a smart metasurface, comprising an Internet of Things device, an edge computing node, a cloud server, and a smart metasurface.

[0006] The IoT device includes a tension sensor, a weighing sensor, and an acceleration sensor. The tension sensor has wireless communication capabilities and is installed on the wire rope to collect the tension signal of the wire rope in real time, and transmits the tension signal to an external device via LoRa wireless communication technology. The weighing sensor also has wireless communication capabilities and is installed at the connection point between the wire rope and the cage or skip, to collect the weight signal of the cage or skip in real time, and transmits the weight signal to an external device via LoRa wireless communication technology. The acceleration sensor also has wireless communication capabilities and is installed on the cage or skip to collect the acceleration signal of the cage or skip in real time, and transmits the acceleration signal to an external device via LoRa wireless communication technology.

[0007] The edge computing node has wireless communication capabilities. It is installed on the mine wall and located near the mine entrance. It is used to receive wireless signals through LoRa wireless communication technology, complete edge computing tasks based on the received wireless signals, and then transmit the obtained data to the cloud server through a wired transmission medium.

[0008] The cloud server is located remotely and is connected to the edge computing nodes. It is used to receive data uploaded by the edge computing nodes and to complete core computing tasks.

[0009] The intelligent metasurface is installed on the mine wall to receive and enhance the strength of wireless signals emitted by IoT devices, and transmits the received wireless signals to edge computing nodes by reflection. At the same time, it optimizes the communication quality between IoT devices and edge computing nodes by adjusting its own reflection phase shift coefficient, ensuring reliable communication between IoT devices and edge computing nodes.

[0010] To facilitate real-time monitoring of the hoist's operation and to collect environmental information about the hoist, a monitoring camera is also included. The monitoring camera has wireless communication capabilities and is installed on the mine shaft wall to collect image data of the monitoring area in real time. The intelligent metasurface is used as a relay and enhancement medium for data transmission to ensure that the image data can be transmitted quickly and stably to the edge computing node.

[0011] Furthermore, to ensure reliable communication, the intelligent metasurface is located in the middle section of the deep coal mine shaft in the vertical direction.

[0012] As a preferred embodiment, the edge computing node consists of a wireless access point and a mobile edge computing server.

[0013] As a preferred embodiment, the wired transmission medium is an optical fiber or a cable.

[0014] In this invention, tension sensors are installed on the wire rope, load cells are installed at the connection points between the wire rope and the cage or skip, and acceleration sensors are installed on the cage or skip. This facilitates real-time acquisition of tension signals from the wire rope, weight signals from the cage or skip, and acceleration signals from the cage or skip during hoist operation, thus enabling comprehensive acquisition of hoist operating data. The tension, load, and acceleration sensors all utilize LoRa wireless communication technology to transmit the acquired signals to external devices, achieving long-distance signal transmission with low power consumption. Edge computing nodes receive wireless signals via LoRa wireless communication technology, enabling long-distance signal reception with low power consumption, and further extending communication distance through the LoRa wireless network. Deploying smart metasurfaces in deep coal mine shafts and integrating them into the LoRa wireless network system significantly enhances the strength and coverage of the wireless signal, effectively improving the stability of the wireless sensor network and further extending the wireless signal coverage. Utilizing edge computing nodes to perform edge computing tasks can effectively reduce the computational burden on cloud servers. Simultaneously, autonomous computing can further reduce dependence on cloud servers. Connecting cloud servers and edge computing nodes via wired transmission media ensures reliable long-distance data transmission. Because intelligent metasurfaces can autonomously adjust their reflective properties to optimize signal propagation and direction, wireless signals can bypass existing physical obstacles and signal blind spots. This allows communication links to dynamically adjust according to actual needs, overcoming the limitations of traditional methods in handling dynamic changes and significantly improving the adaptive capabilities and intelligence level of the communication system, effectively guaranteeing the reliability and security of data transmission.

[0015] This system utilizes LoRa wireless communication technology, combined with the signal enhancement characteristics of intelligent metasurfaces and the adjustable reflectivity of the surface itself, to ensure the stability of the wireless data link within the hoist. This improves data transmission efficiency and ensures the reliability and security of data transmission. It also enables edge computing nodes and cloud servers to perform real-time data fusion processing on the data monitored by the load cells, accelerometers, and tension sensors, thereby achieving comprehensive and reliable monitoring of the hoist system status. This effectively solves the challenges of severe signal attenuation of wireless sensors in the metal environment and high requirements for data stability and latency in deep-well hoists in coal mines.

[0016] This invention also provides a communication method for deep mine hoists based on intelligent metasurface assistance, employing a communication system for deep mine hoists based on intelligent metasurface assistance, comprising the following steps:

[0017] Step 1: Deploy intelligent metasurfaces and construct intelligent communication systems;

[0018] Deploy smart metasurfaces on the mine shaft; utilize the programmable reflection characteristics of smart metasurfaces to actively adjust the direction and phase of reflected beams, constructing virtual and enhanced wireless communication links in complex mine environments, and using wireless communication links to establish communication connections between IoT devices and edge computing nodes; use wired transmission media to establish communication connections between edge computing nodes and remote cloud servers.

[0019] Step 2: Model the energy loss and computation delay issues during the computation offloading process in the intelligent communication system;

[0020] S21: Establish the channel model of the intelligent metasurface according to formula (1) and obtain the channel gain of the intelligent metasurface. ;

[0021] (1);

[0022] In the formula, Indicates the direct path gain; This represents the gain of the nth reflection path; Indicating the first intelligent metasurface n The complex reflection coefficient of each reflecting element;

[0023] S22: Establish a model for receiving signals on a smart metasurface based on formula (2);

[0024] (2);

[0025] In the formula, Indicates the first k Signals received by an IoT device; Indicates the transmission power; Represents the edge computing node and the first k Channel model between IoT devices This represents the channel estimation error. Indicates the transmission of a signal; This represents Gaussian white noise with variance . ;

[0026] S23: Establish a model for the transport rate of the intelligent metasurface based on formula (3);

[0027] (3);

[0028] In the formula, Indicates the first k The transmission rate of an IoT device; Indicates bandwidth;

[0029] S24: Use formula (4) to characterize the energy consumption at IoT devices ;

[0030] (4);

[0031] In the formula, T Indicates the duration of processing a subframe; Indicates the energy efficiency factor of IoT devices; Indicates the first k CPU cycle frequency of an IoT device;

[0032] S25: Use formula (5) to characterize the energy consumption at the edge computing node. ;

[0033] (5);

[0034] In the formula, Indicates the transmit power of IoT devices; Indicates to the first k The proportion of transmission duration for each IoT device; The energy efficiency factor representing the edge computing node; This indicates the computing power of the mobile edge computing server;

[0035] S26: Use formula (6) to characterize the energy consumption at the cloud server. ;

[0036] (6);

[0037] In the formula, This indicates the transmit power of the edge computing node; Indicates the edge computing node to the first k The proportion of transmission duration for each IoT device; The energy efficiency factor represents the energy consumption of a cloud server. This indicates the CPU cycle frequency of the cloud server;

[0038] S27: Average energy consumption and average equivalent delay are defined as indicators for measuring the performance of intelligent communication systems; the average energy consumption during the calculation unloading process is characterized by formula (7). Formula (8) is used to characterize the average equivalent delay during the unloading process. ;

[0039] (7);

[0040] (8);

[0041] In the formula, ; This indicates the number of unloading tasks assigned to IoT devices. This indicates the number of offloaded tasks allocated to the edge computing nodes. This indicates the number of unloading tasks allocated to the cloud server. This indicates the total number of uninstallation tasks; This indicates the number of computation cycles required to process a single bit of computation. This indicates the compute offloading mode. When set to 1, it triggers a task migration mechanism for different compute resource levels, activating the task offloading mode for direct processing capabilities of IoT devices, the local optimized processing capabilities of edge computing nodes, and the powerful centralized processing capabilities of cloud servers, respectively. Conversely, the corresponding mode is turned off.

[0042] Step 3: Considering the characteristics of coal mine hoist fault detection based on multi-source data fusion, a computational unloading strategy that balances real-time performance and resource constraints is constructed; a comprehensive benefit function that integrates energy efficiency and computational latency is constructed according to formula (9). ;

[0043] (9);

[0044] In the formula, Weighting parameters representing energy efficiency, Weight parameters representing the total computation delay;

[0045] Step 4: Maximizing the comprehensive benefit function is established as the optimization objective of the computation offloading strategy, and an optimization framework for the computation offloading strategy is constructed based on formulas (a), (b), (c), (d), (e), and (f). Constraint (a) ensures a balance between security and energy efficiency in the transmit power of IoT devices and edge computing nodes; constraint (b) ensures that IoT devices, edge computing nodes, and cloud servers operate at their respective maximum computing power limits based on CPU frequency; each level measures its ability to process communication data using CPU frequency to ensure system efficiency and stability, achieving efficient data flow and intelligent processing from source to cloud; constraint (c) specifies the fairness of power allocation in the intelligent communication system network; constraint (d) considers the diversity of task modes; constraint (e) defines the range of values ​​for the phase shift of the intelligent metasurface reflection coefficient matrix; and constraint (f) guarantees the minimum transmission rate requirement when channel estimation errors exist.

[0046] ;

[0047] In the formula, Indicates the first k Computation offloading mode for individual IoT devices;

[0048] Step 5: Construct a security reinforcement learning model;

[0049] The dynamic computational unloading process and RIS phase shift adjustment process of the communication system for deep coal mine hoists assisted by intelligent metasurfaces are modeled as Markov decision processes, and the wireless sensor network assisted by intelligent metasurfaces and the intelligent metasurfaces are respectively used as the environment and learning agents. At the same time, a chance-constrained reinforcement learning method based on Lagrange is proposed. Finally, a safety reinforcement learning model for the communication system for deep coal mine hoists assisted by intelligent metasurfaces is constructed.

[0050] Step Six: Achieve optimal computational unloading efficiency;

[0051] S61: First, use the Lagrange multiplier method to transform the constrained Markov decision process problem into an unconstrained dual problem according to formula (10); then, use the integral separation method to formulate the proportional integral of the Lagrange multiplier to prevent the policy from being too conservative.

[0052] (10);

[0053] In the formula, Rewards for reinforcement learning are used to measure the performance of the algorithm. This is the Lagrange penalty factor, used to regulate the degree of relaxation of the constraints; Indicates the total transmission rate of the IoT devices used. Less than the rate threshold The probability of; A threshold representing the probability of interruption;

[0054] S62: Introduce a mining IoT computation offloading mechanism based on deep deterministic policy gradient, and use a deep deterministic policy gradient algorithm to solve the optimal task offloading strategy.

[0055] A1: Initialization based on the deep deterministic policy gradient algorithm; initialization of relevant learning parameters based on the deep deterministic policy gradient algorithm; initialization of Actor network parameters in the deep deterministic policy gradient algorithm. and Critic network parameters By and Network parameters assigned to the Target Actor network and the Target Critic network. , Complete the initialization of the Target network parameters and randomly generate the initial environmental state of the communication system for a deep coal mine hoist assisted by a smart metasurface. ;in, , and These represent the number of tasks assigned to different devices; , and All of these represent parameters for task unloading;

[0056] A2: Single-step exploration and policy sampling; performing one-step exploration within a round, the Actor network initializes the system state s with a deterministic policy based on the current input. Then, the mining IoT devices are based on Noise obtained by sampling the OU process Select the calculation unloading strategy according to formula (11) ;

[0057] During the execution of the computational offloading strategy, reflection phase shift is configured. Number of tasks Calculate the model factor and resource allocation factors This method optimizes system performance and evaluates the utility value of unloading, i.e., the reward. U Subsequently, based on real-time observation of the computational offloading process, the system state is updated to provide a new starting point for subsequent decision-making. During this process, the dynamic changes in computational workload and IoT device location information follow a Markov chain model, ensuring accurate re-estimation of environmental channel state information. Finally, the computational offloading experience containing the aforementioned key information is applied... These cases will be compiled and stored in the experience pool for future learning and optimization.

[0058] (11);

[0059] A3: Network parameter update; determine if the number of calculated unloading experiences in the experience pool has reached Z; if not, proceed directly to step A4 to continue exploring and accumulating calculated unloading experiences until the accumulated calculated unloading experiences reach Z sets; if so, based on experience replay technology, the mining IoT device randomly samples Z sets of calculated unloading experiences from the experience pool, i.e. And using these computational unloading experiences, the parameters of four neural networks are iteratively updated. The four neural network parameters include the Actor network parameters. Critic network parameters Target Actor network parameters and TargetCritic network parameters ;in, Indicates the current state; Indicates the action at the current moment; This represents the reward at the current moment; Indicates the state at the next moment;

[0060] A31: The Adam optimizer is used as the gradient descent algorithm, and the Critic network parameters are updated using formulas (12) and (13). ;

[0061] (12);

[0062] (13);

[0063] In the formula, Represents the loss function; This represents the conversion factor in reinforcement learning; Represents the target Critic network; Represents an Actor network; This indicates the first element taken from the experience replay array. n A loss function; Represents the Critic network; Indicates the current state; This indicates the first element taken from the experience replay array. n One action;

[0064] A32: Update the Actor network parameters according to formula (14) using the deterministic gradient ascent method. ;

[0065] (14);

[0066] In the formula, This represents the gradient of the deterministic policy in the Actor network; Represents the mathematical expectation; This represents the policy gradient of the Critic network; The policy gradient represents the parameters of the Critic network. Represents the policy function;

[0067] A33: Use soft update of learning rate Update the Target Actor network parameters according to formula (15). Update the Target Critic network parameters according to formula (16). This slows down the tracking speed of the Actor network parameters and Critic network parameters.

[0068] (15);

[0069] (16);

[0070] A4: End of Round Decision; Check if the number of time slots has reached the maximum number of training steps within a round. If not, change the state for the next round. For the new initial system state Return to A2; if reached, end the round of training.

[0071] As a preferred embodiment, in step A1 of step six, the relevant learning parameters include the learning rate. Discount Factor Memory pool size, OU noise Soft update learning rate , Batch size Z, number of rounds, and exploration step size within a round.

[0072] In this invention, a smart reflective surface is first deployed in a deep coal mine shaft. The wireless communication link constructed using this surface is then used to establish a communication connection between IoT devices and edge computing nodes. Leveraging the advantages of the smart metasurface's signal enhancement and adjustable reflective properties, combined with the low-power, long-distance transmission characteristics of LoRa wireless communication technology, the coverage of the internal wireless communication network of the hoist can be further expanded. Simultaneously, the reliability and stability of the wireless communication link can be further improved, ensuring that monitoring data can be transmitted to the edge computing nodes in a timely and reliable manner. For coal mine hoist fault detection based on multi-source data fusion, which has high requirements for data stability and latency, this invention, while designing the communication offloading benefit function for the deep coal mine hoist, also considers improving the minimum energy efficiency limit and reducing computational latency. A model of energy loss and computational latency during the computational offloading process of mine IoT devices in a network system assisted by the smart metasurface is constructed. Furthermore, the optimization objective of maximizing the computational offloading benefit function can be achieved by jointly designing the reflection phase shift and computational offloading rate of the smart metasurface. Because it is difficult to accurately obtain real-time wireless communication models and channel state information underground in coal mines, and because of the non-convex constraints of the intelligent metasurface reflection phase shift coefficient and minimum energy loss, this invention effectively addresses the challenges of obtaining real-time information and overcoming non-convex constraints. The dynamic computational unloading process and the intelligent metasurface phase shift adjustment process of the intelligent metasurface-assisted deep-well hoist communication in coal mines are modeled as Markov decision processes. Furthermore, this invention introduces a Lagrange chance-constrained reinforcement learning scheme to ensure real-time data processing and timely emergency response. Deep reinforcement learning, as an emerging technology in the field of artificial intelligence, demonstrates outstanding capabilities in handling optimization problems in complex dynamic environments. By constructing a model based on a finite-state Markov decision process, the deep reinforcement learning algorithm can adaptively select the optimal computational unloading strategy through continuous interactive learning with the environment, even without complete knowledge of the system model and specific channel state information. This ensures efficient real-time data processing and rapid emergency response capabilities for mine IoT devices. To ensure compliance with computational latency thresholds and guarantee real-time data processing and emergency response, a Lagrange-based chance-constrained reinforcement learning method is proposed. This method ultimately constructs a secure reinforcement learning model for communication in deep coal mine hoists assisted by intelligent metasurfaces. The Deep Deterministic Policy Gradient (DDPG) algorithm is then used to optimize the secure reinforcement model, achieving optimal computational offloading efficiency in complex environments and ensuring the high efficiency and safety of deep coal mine hoist communication. By cleverly integrating the secure learning mechanism into the DRL framework, the safety of the exploration process is ensured. This mechanism assesses the risk level of state-action pairs by setting safety indicators and automatically avoids high-risk behaviors, effectively preventing excessive computational latency caused by random exploration and ensuring the stable operation of the mine IoT system.Driven by reinforcement learning algorithms, the lift communication system can continuously monitor and analyze changes in the communication environment, and autonomously adjust the phase shift parameters of the reflector unit to meet communication needs under different environmental conditions with the optimal beam shape.

[0073] This method is highly intelligent and simple to implement. It addresses the signal attenuation and communication instability caused by the dense metal devices in underground coal mines. For the first time, it applies intelligent metasurface technology to the special environment of hoist roadways. At the same time, through continuous interactive learning with the environment, it uses deep reinforcement learning algorithms to continuously optimize the configuration strategy of intelligent metasurfaces. This enables precise control of the propagation path, amplitude, and phase of electromagnetic waves, significantly improving the strength and transmission stability of wireless signals. It effectively adapts to dynamic changes such as the rapid movement of the hoist, ensuring the long-term stability and sustainable optimization capability of the communication quality inside the hoist. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the structure of the hoist communication system in this invention;

[0075] Figure 2 This is a schematic diagram of the circuit part in this invention;

[0076] Figure 3 This is a flowchart of the hoist communication method part of the present invention;

[0077] Figure 4 This is a flowchart of the improved DDPG algorithm in this invention.

[0078] In the diagram: 1. Edge computing node, 2. Cloud server, 3. Smart metasurface, 4. Tension sensor, 5. Weighing sensor, 6. Accelerometer, 7. Wire rope, 8. Cage or basket. Detailed Implementation

[0079] The invention will now be further described with reference to the accompanying drawings.

[0080] As a key piece of equipment in mine production, the coal mine deep shaft hoist is responsible for the vertical transportation of materials and personnel. Its operational safety and efficiency directly affect the overall production efficiency and personnel safety of the mine. Therefore, it is necessary to monitor its operating status and environmental parameters in real time. To this end, sensor nodes based on LoRa technology are used to collect data in real time and transmit it through a low-power wide area network. At the same time, edge computing nodes are deployed nearby to perform preliminary processing of sensor data, reducing data transmission volume and latency. However, the unique environmental characteristics of deep coal mines, such as great depth, narrow tunnels, rough rock walls, and high concentrations of coal dust and ore dust, greatly hinder the transmission of wireless communication signals, creating serious signal blind spots. To address this, Reconfigurable Intelligence Surface (RIS) technology is introduced into the deep-shaft hoisting scenario in coal mines. The RIS is deployed at appropriate locations on the mine wall or at suitable heights around the hoist. Utilizing its programmable reflection characteristics, it actively adjusts the direction and phase of the reflected beam, thereby dynamically adjusting electromagnetic characteristics to optimize the signal propagation path. This constructs a virtual, enhanced wireless communication link in the complex mine environment, ensuring signal stability and coverage. Finally, all processed data is aggregated on cloud server 2 for advanced analysis and report generation, providing real-time, comprehensive decision support for coal mine managers and enabling remote monitoring and control of the deep-shaft hoist.

[0081] Specifically, such as Figure 1 and Figure 2 As shown, the present invention provides a communication system for a deep coal mine hoist based on a smart metasurface, including an Internet of Things device, an edge computing node 1, a cloud server 2, and a smart metasurface 3;

[0082] The IoT device includes a tension sensor 4, a weighing sensor 5, and an acceleration sensor 6. The tension sensor 4 has wireless communication capabilities and is installed on the wire rope 7 to collect the tension signal of the wire rope 7 in real time and transmit the tension signal to an external device via LoRa wireless communication technology. The weighing sensor 5 has wireless communication capabilities and is installed at the connection point between the wire rope 7 and the cage or skip 8 to collect the weight signal of the cage or skip 8 in real time and transmit the weight signal to an external device via LoRa wireless communication technology. The acceleration sensor 6 has wireless communication capabilities and is installed on the cage or skip 8 to collect the acceleration signal of the cage or skip 8 in real time and transmit the acceleration signal to an external device via LoRa wireless communication technology.

[0083] The edge computing node 1 has wireless communication capabilities. It is installed on the mine wall and located near the mine entrance. It is used to receive wireless signals through LoRa wireless communication technology, complete edge computing tasks based on the received wireless signals, and then transmit the obtained data to the cloud server 2 through a wired transmission medium.

[0084] The cloud server 2 is located at a remote location and is connected to the edge computing node 1. It is used to receive data uploaded by the edge computing node 1 and to complete core computing tasks.

[0085] The intelligent metasurface 3 is installed on the mine wall to receive and enhance the strength of wireless signals emitted by IoT devices, and transmits the received wireless signals to edge computing node 1 by reflection. At the same time, it is used to optimize the communication quality between IoT devices and edge computing node 1 by adjusting its own reflection phase shift coefficient, so as to ensure reliable communication between IoT devices and edge computing node 1.

[0086] To facilitate real-time monitoring of the hoist's operation, a monitoring camera is also included. The monitoring camera has wireless communication capabilities and is installed on the mine shaft wall to collect image data of the monitoring area in real time. The intelligent metasurface 3 serves as a relay and enhancement medium for data transmission, ensuring that the image data can be transmitted quickly and stably to the edge computing node 1.

[0087] To ensure reliable communication, the intelligent metasurface 3 is located in the middle section of the deep coal mine shaft in the height direction.

[0088] As a preferred embodiment, the edge computing node 1 consists of a wireless access point and a mobile edge computing (MEC) server. The wireless access point (AP) receives wireless signals and transmits them to the mobile edge computing server, which then processes computing tasks from Internet of Things (IoT) devices.

[0089] As a preferred embodiment, the wired transmission medium is an optical fiber or a cable.

[0090] Deploying Smart Metasurfaces (RIS) in deep coal mine shafts and integrating them into a LoRa wireless network system can significantly enhance signal coverage and provide stable wireless sensor network signals. This effectively solves the problem of severe signal attenuation caused by dense metal components inside the hoist, thus significantly extending the wireless signal coverage. Specifically, through the two reflection paths of the RIS, combined with LoRa wireless communication technology, it ensures that IoT devices (such as various sensors, monitoring cameras, control terminals, etc.) on the coal mine hoist can more effectively transmit measurement signals, images, and data to edge computing nodes, servers, or control centers. During this process, the smart metasurface adjusts its reflection characteristics by obtaining channel state information in a timely manner, thereby optimizing signal propagation and direction, enhancing signal coverage and data link stability. This virtual link not only bypasses the original physical obstacles and signal blind spots but also significantly improves the reliability and coverage of signal transmission. Simultaneously, the flexible configuration capability of the RIS allows the communication link to be dynamically optimized according to actual needs, further improving data transmission efficiency and security.

[0091] In this invention, tension sensors are installed on the wire rope, load cells are installed at the connection points between the wire rope and the cage or skip, and acceleration sensors are installed on the cage or skip. This facilitates real-time acquisition of tension signals from the wire rope, weight signals from the cage or skip, and acceleration signals from the cage or skip during hoist operation, thus enabling comprehensive acquisition of hoist operating data. The tension, load, and acceleration sensors all utilize LoRa wireless communication technology to transmit the acquired signals to external devices, achieving long-distance signal transmission with low power consumption. Edge computing nodes receive wireless signals via LoRa wireless communication technology, enabling long-distance signal reception with low power consumption, and further extending communication distance through the LoRa wireless network. Deploying smart metasurfaces in deep coal mine shafts and integrating them into the LoRa wireless network system significantly enhances the strength and coverage of the wireless signal, effectively improving the stability of the wireless sensor network and further extending the wireless signal coverage. Utilizing edge computing nodes to perform edge computing tasks can effectively reduce the computational burden on cloud servers. Simultaneously, autonomous computing can further reduce dependence on cloud servers. Connecting cloud servers and edge computing nodes via wired transmission media ensures reliable long-distance data transmission. Because intelligent metasurfaces can autonomously adjust their reflective properties to optimize signal propagation and direction, wireless signals can bypass existing physical obstacles and signal blind spots. This allows communication links to dynamically adjust according to actual needs, overcoming the limitations of traditional methods in handling dynamic changes and significantly improving the adaptive capabilities and intelligence level of the communication system, effectively guaranteeing the reliability and security of data transmission.

[0092] This system utilizes LoRa wireless communication technology, combined with the signal enhancement characteristics of intelligent metasurfaces and the adjustable reflectivity of the surface itself, to ensure the stability of the wireless data link within the hoist. This improves data transmission efficiency and ensures the reliability and security of data transmission. It also enables edge computing nodes and cloud servers to perform real-time data fusion processing on the data monitored by the load cells, accelerometers, and tension sensors, thereby achieving comprehensive and reliable monitoring of the hoist system status. This effectively solves the challenges of severe signal attenuation of wireless sensors in the metal environment and high requirements for data stability and latency in deep-well hoists in coal mines.

[0093] like Figure 3 and Figure 4 As shown, the present invention also provides a communication method for deep mine hoists based on intelligent metasurface assistance, employing a communication system for deep mine hoists based on intelligent metasurface assistance, comprising the following steps:

[0094] Step 1: Deploy the intelligent metasurface 3 and construct an intelligent communication system;

[0095] The intelligent metasurface 3 is deployed on the mine wall; the programmable reflection characteristics of the intelligent metasurface 3 are used to actively adjust the direction and phase of the reflected beam to build a virtual and enhanced wireless communication link in the complex mine environment, and the wireless communication link is used to establish a communication connection between the Internet of Things device and the edge computing node 1; the wired transmission medium is used to establish a communication connection between the edge computing node 1 and the remote cloud server 2.

[0096] Step 2: Model the energy loss and computation delay issues during the computation offloading process in the intelligent communication system;

[0097] S21: In a smart metasurface 3-assisted LoRa communication system in a coal mine, the construction of the channel model is crucial. The smart metasurface 3 is... N It consists of several independently controllable reflective elements, which can flexibly adjust the phase of the reflected signal, thereby optimizing the performance of the communication link. In the communication link between Internet of Things (IoT) devices and wireless access points (APs), the signal will experience two reflection paths through the smart metasurface 3, as well as a possible direct path. Channel gain h Taking into account the direct path gain Reflection path gain and the RIS reflection coefficient matrix The influence of this is used to comprehensively characterize the quality and characteristics of the communication link. Specifically, the channel model of the intelligent metasurface 3 is established according to formula (1), and the channel gain of the intelligent metasurface 3 is obtained. ;

[0098] (1);

[0099] In the formula, Indicates the direct path gain; This represents the gain of the nth reflection path; The third intelligent metasurface n The complex reflection coefficients of each reflecting element (including amplitude and phase information) are calculated, and the amplitude is set to 1, with only the phase adjusted.

[0100] S22: Internet of Things (IoT) devices are responsible for transmitting computing task data to wireless access points (APs) via wireless channels. Given the unique physical conditions of mines, direct communication links often face severe interference and coverage blind spots, which greatly challenges the reliability and efficiency of signal transmission. To solve this problem, the deployment of intelligent metasurfaces (RIS)3 is particularly important. As a key means of signal enhancement, it has immeasurable value in expanding signal coverage and improving transmission quality. For the signal reception process, a model of intelligent metasurface 3 receiving signals can be established according to formula (2).

[0101] (2);

[0102] In the formula, Indicates the first k Signals received by an IoT device; Indicates the transmission power; Represents edge computing node 1 and the... k Channel model between IoT devices This represents the channel estimation error. Indicates the transmission of a signal; This represents Gaussian white noise with variance . ;

[0103] S23: Under the overall architecture of the deep coal mine communication system, the transmission rate is an important standard for measuring the system's transmission capacity. A model of the transmission rate of the intelligent metasurface 3 can be established according to formula (3).

[0104] (3);

[0105] In the formula, Indicates the first k The transmission rate of an IoT device; Indicates bandwidth;

[0106] S24: The energy consumption of sensor nodes is a key indicator for measuring their sustainable operation capability. It not only relates to the cost of a single task execution, but also directly affects the long-term stable operation and energy efficiency optimization of the entire system. The energy consumption at IoT devices can be characterized by formula (4). ;

[0107] (4);

[0108] In the formula, T Indicates the duration of processing a subframe; This represents the energy efficiency factor of IoT devices; Indicates the first k CPU cycle frequency of an IoT device;

[0109] S25: Use formula (5) to characterize the energy consumption at edge computing node 1. ;

[0110] (5);

[0111] In the formula, Indicates the transmit power of IoT devices; Indicates to the first k The proportion of transmission duration for each user; This represents the energy efficiency factor of edge computing node 1; This indicates the computing power of the mobile edge computing server;

[0112] S26: Assume that the delay in transmitting the result information after the edge computing task is negligible, because the calculated information is usually very small. For cloud computing, all computing data is first transmitted to the MEC server, and then forwarded to the remote cloud server 2 via fiber optic transmission. The average data computing power and energy consumption of cloud server 2 can be characterized by formula (6);

[0113] (6);

[0114] In the formula, This represents the transmit power of edge computing node 1; This indicates that edge computing node 1 is assigned to the first... k The proportion of transmission duration for each user; This represents the energy efficiency factor of cloud server 2; This indicates the CPU cycle frequency of cloud server 2;

[0115] S27: In order to comprehensively evaluate the performance of the entire offloading strategy, average energy consumption and average equivalent delay are defined as indicators to measure the performance of intelligent communication systems; the average energy consumption during the calculation of the offloading process is characterized by formula (7). Formula (8) is used to characterize the average equivalent delay during the unloading process. ;

[0116] (7);

[0117] (8);

[0118] In the formula, ; This indicates the number of unloading tasks assigned to IoT devices. This indicates the number of unloaded tasks allocated to the edge computing node (1). This indicates the number of unloading tasks allocated to cloud server (2). This indicates the total number of uninstallation tasks; This indicates the number of computation cycles required to process a single bit of computation. This indicates the compute offloading mode. When set to 1, it triggers the task migration mechanism for different compute resource levels, respectively activating the task offloading modes for the direct processing capabilities of IoT devices, the local optimized processing capabilities of edge computing node 1, and the powerful centralized processing capabilities of cloud server 2. Conversely, the corresponding modes are turned off.

[0119] Step 3: For fault detection of coal mine hoists based on multi-source data fusion, considering the characteristics of data stability and high latency requirements, while designing the offloading benefit function of deep coal mine hoist communication, the issues of improving the minimum energy efficiency limit and reducing computational latency are fully considered. The energy loss and computational latency issues of mining IoT devices in the network system assisted by the intelligent metasurface (RIS) 3 are modeled. The optimization goal of maximizing the computational offloading benefit function is achieved by jointly designing the intelligent metasurface (RIS) 3 reflection phase shift and computational offloading rate.

[0120] Specifically, a computational offloading strategy that balances real-time performance and resource constraints is constructed. This strategy must ensure the immediate processing of computational tasks to maintain the safety and efficiency of mining operations. To this end, a comprehensive benefit function is constructed, which integrates energy efficiency and computational latency. This function aims to comprehensively evaluate and optimize offloading decisions. The comprehensive benefit function integrating energy efficiency and computational latency can be constructed according to formula (9). ;

[0121] (9);

[0122] In the formula, Weighting parameters representing energy efficiency, The weight parameter represents the total computational delay and is used to balance the impact of the utility function.

[0123] Step 4: Establish the maximization of the comprehensive benefit function as the optimization objective of the computation unloading strategy. That is, the strategy is adjusted to maximize the function, thereby taking into account both the real-time requirements of the computation task and the efficiency of resource utilization.

[0124] Specifically, an optimization framework for the computational offloading strategy is constructed based on formulas (a), (b), (c), (d), (e), and (f). In the process of constructing the optimization framework for the computational offloading strategy, multidimensional constraints are finely set to reflect the complexity of system operation and actual needs. Among them, constraint (a) ensures that the transmit power of IoT devices and edge computing node 1 achieves a balance between security and energy efficiency; constraint (b) ensures that IoT devices, edge computing node 1, and cloud server 2 operate at the maximum computing power limit of their respective CPU frequencies; each level measures the ability to process communication data through CPU frequency to ensure the system is efficient and stable, and realizes efficient data flow and intelligent processing from the source to the cloud; constraint (c) specifies the fairness of power allocation in the intelligent communication system network; constraint (d) considers the diversity of task modes; constraint (e) defines the range of values ​​for the phase shift of the reflection coefficient matrix of intelligent metasurface 3; constraint (f) guarantees the minimum requirement of transmission rate when channel estimation error exists. This series of constraints constitutes a high-dimensional, continuous, and non-convex optimization problem, and its solution requires the use of advanced optimization theories and algorithms to achieve the optimization of system performance.

[0125] ;

[0126] In the formula, Indicates the first k Computation offloading mode for individual IoT devices;

[0127] Step 5: Construct a security reinforcement learning model;

[0128] Because it is difficult to accurately obtain real-time wireless communication models and channel state information in underground coal mines, and due to the non-convex constraints of RIS reflection phase shift coefficient and minimum energy loss, this paper models the dynamic computational unloading process and RIS phase shift adjustment process of the coal mine deep shaft hoist communication system assisted by intelligent metasurface 3 as Markov decision processes. The wireless sensor network assisted by intelligent metasurface 3 and the intelligent metasurface 3 are respectively used as the environment and learning agent. Simultaneously, to ensure that the computational delay threshold requirements are met and to guarantee real-time data processing and emergency response, a Lagrange-based chance-constrained reinforcement learning method is proposed. Finally, a safety reinforcement learning model for the coal mine deep shaft hoist communication system assisted by intelligent metasurface 3 is constructed.

[0129] Step Six: Achieve optimal computational unloading efficiency;

[0130] S61: In order to avoid the difficulty of solving the problem caused by non-convex constraints, the Lagrange multiplier method is first used to transform the constrained Markov decision process problem into an unconstrained dual problem according to formula (10); then the proportional integral of the Lagrange is formulated by the integral separation method to prevent the policy from being too conservative.

[0131] (10);

[0132] In the formula, Rewards for reinforcement learning are used to measure the performance of the algorithm. This is the Lagrange penalty factor, used to regulate the degree of relaxation of the constraints; Indicates the total transmission rate of the IoT devices used. Less than the rate threshold The probability of; A threshold representing the probability of interruption;

[0133] S62: To address the complex high-dimensional continuous policy optimization challenge in the Smart Metasurface (RIS) 3-assisted LoRa-MEC system for underground coal mines, this application introduces a computational offloading mechanism for mining IoT based on Deep Deterministic Policy Gradient (DDPG). This mechanism relies on the secure reinforcement learning model constructed in step five and uses the DDPG algorithm to solve for the optimal task offloading policy, aiming to achieve unconstrained optimization. The specific implementation path includes precisely defining the state space, action space, and reward function; approximating the policy and value function through a deep neural network; and employing techniques such as experience replay and a dual-network structure to ensure the stability and convergence of the learning process, ultimately achieving efficient computational task offloading decisions.

[0134] To dynamically explore computation offloading strategies in complex and ever-changing hoist communication scenarios, it is necessary to design reasonable RIS reflection phase shifts and computation offloading rates to achieve the goal of optimal computation offloading efficiency. This involves the following steps:

[0135] A1: Initialization based on the deep deterministic policy gradient algorithm; Initialize the relevant learning parameters based on the deep deterministic policy gradient algorithm, including the learning rate. Discount Factor Memory pool size, OU noise Soft update learning rate , Batch size Z, number of rounds, and exploration step size within each round; initialize the Actor network parameters in the deep deterministic policy gradient algorithm. and Critic network parameters By and Network parameters assigned to the Target Actor network and the Target Critic network. , Complete the initialization of the Target network parameters and randomly generate the initial environmental state of the communication system for a deep coal mine hoist based on a smart metasurface 3-assisted system. ;in, , and These represent the number of tasks assigned to different devices (IoT devices, edge computing nodes, and cloud servers); , and All of these represent parameters for task unloading;

[0136] A2: Single-step exploration and policy sampling; performing one-step exploration within a round, the Actor network initializes the system state s with a deterministic policy based on the current input. Then, the mining IoT devices are based on Noise obtained by sampling the OU process Select the calculation unloading strategy according to formula (11) ;

[0137] During the execution of the computational offloading strategy, the reflection phase shift is carefully configured. Number of tasks Calculate the model factor and resource allocation factors This method optimizes system performance and evaluates the utility value of offloading (which quantifies the effectiveness of the offloading strategy), i.e., rewards. U Subsequently, based on real-time observation of the computational offloading process, the system state is updated to provide a new starting point for subsequent decision-making. During this process, the dynamic changes in computational workload and IoT device location information follow a Markov chain model, ensuring accurate re-estimation of environmental channel state information. Finally, the computational offloading experience containing the aforementioned key information is applied... These cases will be compiled and stored in the experience pool for future learning and optimization.

[0138] (11);

[0139] A3: Network parameter update; determine if the number of calculated unloading experiences in the experience pool has reached Z; if not, proceed directly to step A4 to continue exploring and accumulating calculated unloading experiences until the accumulated calculated unloading experiences reach Z sets; if so, based on experience replay technology, the mining IoT device randomly samples Z sets of calculated unloading experiences from the experience pool, i.e. And using these computational unloading experiences, the parameters of four neural networks are iteratively updated. The four neural network parameters include the Actor network parameters. Critic network parameters Target Actor network parameters and TargetCritic network parameters ;in, Indicates the current state; Indicates the action at the current moment; This represents the reward at the current moment; Indicates the state at the next moment;

[0140] A31: The Adam optimizer is used as the gradient descent algorithm, and the Critic network parameters are updated using formulas (12) and (13). ;

[0141] (12);

[0142] (13);

[0143] In the formula, Represents the loss function; This represents the conversion factor in reinforcement learning; Represents the target Critic network; Represents an Actor network; This indicates the first element taken from the experience replay array. n A loss function; Represents the Critic network; Indicates the current state; This indicates the first element taken from the experience replay array. n One action;

[0144] A32: Update the Actor network parameters according to formula (14) using the deterministic gradient ascent method. ;

[0145] (14);

[0146] In the formula, This represents the gradient of the deterministic policy in the Actor network; Represents the mathematical expectation; This represents the policy gradient of the Critic network; The policy gradient represents the parameters of the Critic network. Represents the policy function;

[0147] A33: Use soft update of learning rate Update the Target Actor network parameters according to formula (15). Update the Target Critic network parameters according to formula (16). This slows down the tracking speed of the Actor network parameters and Critic network parameters.

[0148] (15);

[0149] (16);

[0150] A4: End of Round Decision; Check if the number of time slots has reached the maximum number of training steps within a round. If not, change the state for the next round. For the new initial system state Return to A2; if reached, end the round of training.

[0151] In this invention, a smart reflective surface is first deployed in a deep coal mine shaft. The wireless communication link constructed using this surface is then used to establish a communication connection between IoT devices and edge computing nodes. Leveraging the advantages of the smart metasurface's signal enhancement and adjustable reflective properties, combined with the low-power, long-distance transmission characteristics of LoRa wireless communication technology, the coverage of the internal wireless communication network of the hoist can be further expanded. Simultaneously, the reliability and stability of the wireless communication link can be further improved, ensuring that monitoring data can be transmitted to the edge computing nodes in a timely and reliable manner. For coal mine hoist fault detection based on multi-source data fusion, which has high requirements for data stability and latency, this invention, while designing the communication offloading benefit function for the deep coal mine hoist, also considers improving the minimum energy efficiency limit and reducing computational latency. A model of energy loss and computational latency during the computational offloading process of mine IoT devices in a network system assisted by the smart metasurface is constructed. Furthermore, the optimization objective of maximizing the computational offloading benefit function can be achieved by jointly designing the reflection phase shift and computational offloading rate of the smart metasurface. Because it is difficult to accurately obtain real-time wireless communication models and channel state information underground in coal mines, and because of the non-convex constraints of the intelligent metasurface reflection phase shift coefficient and minimum energy loss, this invention effectively addresses the challenges of obtaining real-time information and overcoming non-convex constraints. The dynamic computational unloading process and the intelligent metasurface phase shift adjustment process of the intelligent metasurface-assisted deep-well hoist communication in coal mines are modeled as Markov decision processes. Furthermore, this invention introduces a Lagrange chance-constrained reinforcement learning scheme to ensure real-time data processing and timely emergency response. Deep reinforcement learning, as an emerging technology in the field of artificial intelligence, demonstrates outstanding capabilities in handling optimization problems in complex dynamic environments. By constructing a model based on a finite-state Markov decision process, the deep reinforcement learning algorithm can adaptively select the optimal computational unloading strategy through continuous interactive learning with the environment, even without complete knowledge of the system model and specific channel state information. This ensures efficient real-time data processing and rapid emergency response capabilities for mine IoT devices. To ensure compliance with computational latency thresholds and guarantee real-time data processing and emergency response, a Lagrange-based chance-constrained reinforcement learning method is proposed. This method ultimately constructs a secure reinforcement learning model for communication in deep coal mine hoists assisted by intelligent metasurfaces. The Deep Deterministic Policy Gradient (DDPG) algorithm is then used to optimize the secure reinforcement model, achieving optimal computational offloading efficiency in complex environments and ensuring the high efficiency and safety of deep coal mine hoist communication. By cleverly integrating the secure learning mechanism into the DRL framework, the safety of the exploration process is ensured. This mechanism assesses the risk level of state-action pairs by setting safety indicators and automatically avoids high-risk behaviors, effectively preventing excessive computational latency caused by random exploration and ensuring the stable operation of the mine IoT system.Driven by reinforcement learning algorithms, the lift communication system can continuously monitor and analyze changes in the communication environment, and autonomously adjust the phase shift parameters of the reflector unit to meet communication needs under different environmental conditions with the optimal beam shape.

[0152] This method is highly intelligent and simple to implement. It addresses the signal attenuation and communication instability caused by the dense metal devices in underground coal mines. For the first time, it applies intelligent metasurface technology to the special environment of hoist roadways. At the same time, through continuous interactive learning with the environment, it uses deep reinforcement learning algorithms to continuously optimize the configuration strategy of intelligent metasurfaces. This enables precise control of the propagation path, amplitude, and phase of electromagnetic waves, significantly improving the strength and transmission stability of wireless signals. It effectively adapts to dynamic changes such as the rapid movement of the hoist, ensuring the long-term stability and sustainable optimization capability of the communication quality inside the hoist.

Claims

1. A communication method for a deep coal mine hoist based on intelligent metasurface assistance, comprising an intelligent metasurface-assisted communication system for a deep coal mine hoist, the communication system comprising an Internet of Things (IoT) device, an edge computing node (1), a cloud server (2), and an intelligent metasurface (3); the IoT device comprising a tension sensor (4), a weighing sensor (5), and an acceleration sensor (6); the tension sensor (4) having wireless communication function, being installed on a wire rope (7) for real-time acquisition of the tension signal of the wire rope (7), and communicating via LoRa wireless communication technology. The tension signal is transmitted to an external device; the weighing sensor (5) has wireless communication function and is installed at the connection point between the wire rope (7) and the cage or skip (8) to collect the weight signal of the cage or skip (8) in real time and transmit the weight signal to an external device through LoRa wireless communication technology; the acceleration sensor (6) has wireless communication function and is installed on the cage or skip (8) to collect the acceleration signal of the cage or skip (8) in real time and transmit the acceleration signal to an external device through LoRa wireless communication technology; the edge computing node (1) has wireless communication function. Yes, it is installed on the mine wall and located near the mine entrance. It is used to receive wireless signals through LoRa wireless communication technology, and to complete edge computing tasks based on the received wireless signals. Then, it transmits the obtained data to the cloud server (2) through a wired transmission medium. The cloud server (2) is located at a remote end and is connected to the edge computing node (1). It is used to receive the data uploaded by the edge computing node (1) and to complete the core computing tasks. The smart metasurface (3) is installed on the mine wall to receive and enhance the strength of the wireless signals emitted by the Internet of Things devices, and to reflect the received wireless signals. The data is transmitted to the edge computing node (1). At the same time, it is used to optimize the communication quality between the IoT device and the edge computing node (1) by adjusting its own reflection phase shift coefficient, so as to ensure reliable communication between the IoT device and the edge computing node (1). It also includes a monitoring camera, which has wireless communication function. It is installed on the mine wall to collect image data of the monitoring area in real time, and uses the smart metasurface (3) as a relay and enhancement medium for data transmission to ensure that the image data can be transmitted to the edge computing node (1) quickly and stably. The smart metasurface (3) is located in the middle section of the deep shaft of the coal mine in the height direction. Its features are, Includes the following steps: Step 1: Arrange the intelligent metasurface (3) and construct the intelligent communication system; The intelligent metasurface (3) is deployed on the mine wall; the programmable reflection characteristics of the intelligent metasurface (3) are used to actively adjust the direction and phase of the reflected beam to build a virtual and enhanced wireless communication link in the complex mine environment, and the wireless communication link is used to establish a communication connection between the Internet of Things device and the edge computing node (1); the wired transmission medium is used to establish a communication connection between the edge computing node (1) and the remote cloud server (2); Step 2: Model the energy loss and computation delay issues during the computation offloading process in the intelligent communication system; S21: Establish the channel model of the intelligent metasurface (3) according to formula (1) and obtain the channel gain of the intelligent metasurface (3). ; （1）； In the formula, The channel gain of the smart metasurface (3) is represented; Indicates the direct path gain; Indicates the first Gain of each reflection path; Indicating the intelligent metasurface (3) The complex reflection coefficient of each reflecting element; S22: Establish a model for the intelligent metasurface (3) to receive signals based on formula (2); （2）； In the formula, Indicates the first Signals received by an IoT device; Indicates the transmission power; Represents the edge computing node (1) and the first Channel model between IoT devices; This refers to the channel estimation error. Indicates the transmission of a signal; This represents Gaussian white noise with variance . ; S23: Establish a model of the transmission rate of the smart metasurface (3) based on formula (3); （3）； In the formula, Indicates the first The transmission rate of an IoT device; Indicates bandwidth; S24: Use formula (4) to characterize the energy consumption at IoT devices ; （4）； In the formula, Indicates the duration of processing a subframe; Indicates the energy efficiency factor of IoT devices; Indicates the first CPU cycle frequency of an IoT device; S25: Use formula (5) to characterize the energy consumption at the edge computing node (1). ; （5）； In the formula, Indicates the transmit power of IoT devices; Indicates to the first The proportion of transmission duration for each IoT device; The energy efficiency factor represents the energy consumption of the edge computing node (1); This indicates the CPU cycle frequency of the mobile edge computing server; S26: Use formula (6) to characterize the energy consumption at cloud server (2). ; （6）； In the formula, This represents the transmit power of the edge computing node (1); The edge computing node (1) is given to the first The proportion of transmission duration for each IoT device; The energy efficiency factor of the cloud server (2) is represented; This indicates the CPU cycle frequency of the cloud server (2); S27: Average energy consumption and average equivalent delay are defined as indicators for measuring the performance of intelligent communication systems; the average energy consumption during the calculation unloading process is characterized by formula (7). Formula (8) is used to characterize the average equivalent delay during the unloading process. ; （7）； （8）； In the formula, ; This indicates the number of unloading tasks assigned to IoT devices. This indicates the number of unloaded tasks allocated to the edge computing node (1). This indicates the number of unloading tasks allocated to cloud server (2). This indicates the total number of uninstallation tasks; This indicates the number of computation cycles required to process a single bit of computation. This indicates the computing offload mode. When set to 1, it triggers the task migration mechanism for different computing resource levels, respectively activating the task offload mode for the direct processing capability of IoT devices, the local optimized processing capability of edge computing nodes (1), and the powerful centralized processing capability of cloud servers (2). Conversely, it also closes the corresponding mode. Step 3: Considering the characteristics of coal mine hoist fault detection based on multi-source data fusion, a computational unloading strategy that balances real-time performance and resource constraints is constructed; a comprehensive benefit function that integrates energy efficiency and computational latency is constructed according to formula (9). This function integrates two dimensions: energy efficiency and computational latency, aiming to comprehensively evaluate and optimize offloading decisions. （9）； In the formula, Weighting parameters representing energy efficiency, Weight parameters representing the total computation delay; Step 4: Maximize the overall benefit function The optimization objective of the computation offloading strategy is established, and the optimization framework of the computation offloading strategy is constructed according to formulas (a), (b), (c), (d), (e), and (f). Constraint (a) ensures that the transmission power of IoT devices and edge computing nodes (1) is balanced between safety and energy efficiency. Constraint (b) ensures that IoT devices, edge computing nodes (1), and cloud servers (2) operate at the maximum computing power limit of their respective CPU frequencies. Each level measures the ability to process communication data through CPU frequency to ensure the system is efficient and stable, and realizes efficient flow and intelligent processing of data from the source to the cloud. Constraint (c) specifies the fairness of power allocation in the intelligent communication system network. Constraint (d) considers the diversity of task modes. Constraint (e) defines the range of values ​​for the phase shift of the reflection coefficient matrix of the intelligent metasurface (3). Constraint (f) guarantees the minimum requirement of transmission rate when channel estimation error exists. This series of constraints constitutes a high-dimensional, continuous, and non-convex optimization problem, and its solution requires the use of advanced optimization theories and algorithms to achieve the optimization of system performance. （a）； （b）； （c）； （d）； （e）； （f）； In the formula, Indicates the first Computation offloading mode for individual IoT devices; Step 5: Construct a security reinforcement learning model; The dynamic computational unloading process and RIS phase shift adjustment process of the coal mine deep shaft hoist communication system assisted by the intelligent metasurface (3) are modeled as Markov decision processes, and the wireless sensor network assisted by the intelligent metasurface (3) and the intelligent metasurface (3) are respectively used as the environment and learning agent; at the same time, a chance-constrained reinforcement learning method based on Lagrange is proposed; finally, a safety reinforcement learning model of the coal mine deep shaft hoist communication system assisted by the intelligent metasurface (3) is constructed. Step Six: Achieve optimal computational unloading efficiency; S61: First, use the Lagrange multiplier method to transform the constrained Markov decision process problem into an unconstrained dual problem according to formula (10); then, use the integral separation method to formulate the proportional integral of the Lagrange multiplier to prevent the policy from being too conservative. （10）； In the formula, Rewards for reinforcement learning are used to measure the performance of the algorithm. This is the Lagrange penalty factor, used to regulate the degree of relaxation of the constraints; Indicates the total transmission rate of the IoT devices used. Less than the rate threshold The probability of; A threshold representing the probability of interruption; S62: Introduce a mining IoT computation offloading mechanism based on deep deterministic policy gradient, and use a deep deterministic policy gradient algorithm to solve the optimal task offloading strategy. A1: Initialization based on the deep deterministic policy gradient algorithm; initialization of relevant learning parameters based on the deep deterministic policy gradient algorithm; initialization of Actor network parameters in the deep deterministic policy gradient algorithm. and Critic network parameters By and Network parameters assigned to the Target Actor network and the Target Critic network. , Complete the initialization of the Target network parameters and randomly provide the initial environmental state of the coal mine deep shaft hoist communication system based on the intelligent metasurface (3) assisted. ;in, , and These represent the number of tasks assigned to different devices; ; and All of these represent parameters for task unloading; A2: Single-step exploration and policy sampling; performing one-step exploration within a round, the Actor network initializes the system state based on the current input. Deterministic strategy Then, the mining IoT devices are based on Noise obtained by sampling the OU process Select the calculation unloading strategy according to formula (11) ; During the execution of the computational offloading strategy, reflection phase shift is configured. Number of tasks Calculate the model factor and resource allocation factors This method optimizes system performance and evaluates the utility value of unloading, i.e., the reward. U Subsequently, based on real-time observation of the computational offloading process, the system state is updated to provide a new starting point for subsequent decision-making. During this process, the dynamic changes in computational workload and IoT device location information follow a Markov chain model, ensuring accurate re-estimation of environmental channel state information. Finally, the computational offloading experience containing the aforementioned key information is applied... These cases will be compiled and stored in the experience pool for future learning and optimization. （11）； A3: Network parameter update; determine if the number of calculated unloading experiences in the experience pool has reached Z; if not, proceed directly to step A4 to continue exploring and accumulating calculated unloading experiences until the accumulated calculated unloading experiences reach Z sets; if so, based on experience replay technology, the mining IoT device randomly samples Z sets of calculated unloading experiences from the experience pool, i.e. And using these computational unloading experiences, the parameters of four neural networks are iteratively updated. The four neural network parameters include the Actor network parameters. Critic network parameters Target Actor network parameters and TargetCritic network parameters ;in, Indicates the current state; Indicates the action at the current moment; This represents the reward at the current moment; Indicates the state at the next moment; A31: The Adam optimizer is used as the gradient descent algorithm, and the Critic network parameters are updated using formulas (12) and (13). ; （12）； （13）； In the formula, Represents the loss function; This represents the conversion factor in reinforcement learning; Represents the target Critic network; Represents an Actor network; This indicates the first element taken from the experience replay array. A loss function; Represents the Critic network; Indicates the current state; This indicates the first element taken from the experience replay array. One action; A32: Update the Actor network parameters according to formula (14) using the deterministic gradient ascent method. ; （14）； In the formula, This represents the gradient of the deterministic policy of the Actor network; Represents the mathematical expectation; This represents the policy gradient of the Critic network; The policy gradient represents the parameters of the Critic network. Represents the strategy function; A33: Use soft update of learning rate Update the Target Actor network parameters according to formula (15). Update the Target Critic network parameters according to formula (16). This slows down the tracking speed of the Actor network parameters and Critic network parameters. （15）； （16）； In the formula, This refers to hyperparameters that are manually adjusted; A4: End of Round Decision; Check if the number of time slots has reached the maximum training step size within a round. If not, change the state for the next round. For the new initial system state Return to A2; if the target is reached, end the training round.

2. The communication method for deep mine hoists based on intelligent metasurface assistance according to claim 1, characterized in that, The edge computing node (1) consists of a wireless access point and a mobile edge computing server.

3. The communication method for deep mine hoists based on intelligent metasurface assistance according to claim 1, characterized in that, The wired transmission medium is optical fiber or electrical cable.

4. The communication method for deep mine hoists based on intelligent metasurface assistance according to claim 1, characterized in that, In step A1 of step six, the relevant learning parameters include the learning rate. Discount Factor Memory pool size, OU noise Soft update learning rate , Batch size Z, number of rounds, and exploration step size within a round.

Images