Unmanned aerial vehicle assisted railway sensor data collection and compute offload method and system

Abstract

The application discloses a kind of unmanned aerial vehicle assisted railway sensing data collection and computing unloading method and system, based on DRL intelligent algorithm decision unmanned aerial vehicle UAV flight trajectory and computing task unloading, to wireless sensor network WSN sensing data is collected, and task unloading calculation, while meeting certain real-time demand of railway environment monitoring data makes WSN transmission energy consumption and UAV flight and computing energy consumption minimization.The present application can be applied to the monitoring of railway operating environment, through the deployment flexibility of UAV, ensure certain real-time data collection of wireless sensor network, and minimize the transmission energy consumption of wireless sensor network as far as possible, while using the computing resources of base station and high-speed train, reduce the computing load of UAV, prolong the life cycle of entire monitoring system.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and in particular relates to a method and system for collecting and calculating railway sensor data assisted by an unmanned aerial vehicle (UAV). Background Art

[0002] With the rapid development of China's railways, effective monitoring of railway infrastructure and the environment has become increasingly important. In environments with challenging geological conditions, where laying and maintaining wired backhaul networks is difficult, and where real-time information transmission must be guaranteed, wireless sensor networks (WSNs) have become an effective solution for monitoring the railway operating environment. For example, video surveillance cameras, foreign object intrusion detection sensors, and disaster monitoring sensors deployed along the railway line transmit data to ground control centers via wireless backhaul. Large-scale deployment of WSNs for monitoring railway operating environments captures and analyzes WSN data to track changes in the railway operating environment. However, the massive amount of monitoring data generated in this process poses significant challenges to the transmission bearer network and requires significant computing power for data processing and calculation. The quality of service in this process can be measured using the AoI (AoI) metric, which measures the freshness of the calculated results. In railway WSN scenarios, the AoI of data represents the time difference between data acquisition, collection, offloading, calculation, and the return of the final result to the data fusion center.

[0003] With advances in unmanned aerial vehicle (UAV) technology, UAVs are increasingly being used in civil and commercial applications, including surveillance, aerial imaging, precision agriculture, smart logistics, and disaster response. UAVs offer advantages such as rapid deployment, ease of programming, and high scalability, providing reliable and cost-effective wireless communication solutions for a variety of real-world scenarios. Furthermore, UAVs can fly to locations with favorable wireless channel conditions, enabling devices to transmit data with low energy consumption. Therefore, in railway wireless sensor networks, using UAVs for sensor data collection can effectively reduce the transmission energy consumption of the sensor network and extend the sensor network's lifecycle. UAV trajectory control can also achieve better data transmission channel conditions, ensuring data reliability. Furthermore, UAV-mounted servers can directly process the collected data, further improving data processing latency. However, given the limited battery life and computing power of UAVs, UAVs can offload some data processing to the base station (BS). Summary of the Invention

[0004] To address the short battery lifecycle in railway wireless sensor networks, the railway's real-time demand for environmental data, and the computational processing of massive amounts of monitoring data, the present invention provides a drone-assisted railway sensor data collection and computation offloading method and system.

[0005] The present invention provides a UAV-assisted railway sensor data collection and computation offloading method. Based on the Deep Reinforcement Learning (DRL) algorithm, the UAV flight trajectory and computation task offloading are determined. The method collects wireless sensor network (WSN) sensor data and offloads the computation tasks. This method minimizes the WSN transmission energy consumption and the UAV flight and computation energy consumption while meeting certain real-time requirements for railway environmental monitoring data. The method specifically includes the following steps:

[0006] Data collection phase of WSN: SN (Sensor Node) collects railway environment data and stores the data in the buffer area; SN sends a data status indication frame to BS (Base Station), which includes the SN's location coordinates, SN buffer queue status, data timestamp, and data priority.

[0007] UAV data collection phase: The base station (BS) executes the DRL algorithm to make decisions based on the WSN data status and UAV computing capabilities, including UAV data collection flight trajectory and computing task offloading. The BS sends a trajectory indication frame to the UAV, which contains the locations that the UAV needs to reach first and then later. The UAV flies to the corresponding SN location based on the trajectory and sends a wake-up indication frame to the SN. The SN then uploads the sensor data in the buffer area.

[0008] Computational task offloading phase: After completing data collection, the UAV sends a collection completion indication frame to the BS. The BS then sends a task offloading indication frame to the UAV, which includes the offloaded data ratio and offloaded data type. The offload ratio ranges from 0 to 1. When the BS computing resources are insufficient, the BS offloads part of the task to the train for calculation, and the final calculation results are returned to the BS.

[0009] Algorithm training and execution: The DRL algorithm is executed in the base station (BS). The algorithm's environmental status information includes the location coordinates of the SN, the SN buffer queue status, data timestamps, data priority, and the computing resources of the UAV. The algorithm's actions include the UAV trajectory and computing task offloading decisions. The algorithm's reward indicators include the freshness of the transmitted data (AoI), transmission energy consumption, and UAV energy consumption. The algorithm trains the model using historical data or experimental data. When the training reaches the preset rounds or meets the preset feedback conditions, the BS uses the trained model to make decisions on the UAV trajectory, data collection, and computing task offloading.

[0010] Furthermore, the data status indication frame is used to inform the BS of the sensor data information in the SN cache area, including SN location, cache queue status, data timestamp, and data priority; based on this information, the BS executes the DRL algorithm to decide the UAV data collection flight trajectory and computing task offloading.

[0011] Furthermore, the trajectory indication frame is used to control the flight trajectory of the UAV data collection, including the location information of the SNs that the UAV needs to reach first and then. Based on the flight trajectory, the UAV adjusts the speed and direction to fly to the corresponding SN positions first and then sends a wake-up indication frame to the corresponding SN, and the SN uploads the sensor data.

[0012] Furthermore, the task offloading indication frame is used to inform the UAV to offload the computing task, including the offloading ratio and the type of offloaded data. Based on this indication, the UAV sends the computing task to the BS. When the BS computing resources are insufficient, the BS offloads part of the task to the train for calculation. After the corresponding device completes the calculation, the calculation results are returned to the BS.

[0013] The present invention provides a drone-assisted railway sensor data collection and calculation offloading system, which is suitable for the above-mentioned drone-assisted railway sensor data collection and calculation offloading method, including but not limited to railway WSN constructed by Wi-Fi, Star Flash, ZigBee and Bluetooth communication technologies and gateway equipment for network protocol conversion, as well as UAV, ground BS and high-speed trains. The cellular communication standards used between key equipment include but not limited to 4G and 5G communication technologies.

[0014] Among them, WSN perceives the railway operating environment and sends data status indication frames to the base station. The BS executes the DRL decision algorithm based on this information; the BS sends a trajectory indication frame to the UAV. The UAV flies based on the trajectory and sends a wake-up indication frame when it reaches the corresponding SN position. The SN sends the sensor data to the UAV; after the collection is completed, the UAV flies back to the BS according to the trajectory and informs the BS. The BS sends a task offloading indication frame to the UAV. The UAV offloads the computing task to the BS. When the BS computing resources are insufficient, the BS offloads part of the task to the train for calculation. Finally, all the results are returned to the BS.

[0015] The beneficial technical effects of the present invention are:

[0016] This invention is designed for data collection and computation in wireless sensor networks used for railway operation and maintenance environmental monitoring. The base station offloads data collection and computation tasks by determining the flight trajectory of the UAV based on the WSN's data status and the UAV's computing resources. This approach not only meets the need for fresh railway environmental monitoring data, but also reduces the WSN's energy consumption for data transmission and the UAV's flight and computing energy consumption, extending the WSN network lifecycle. This invention provides strong support for future intelligent railway operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the principle architecture of the drone-assisted railway sensor data collection and computation offloading method of the present invention.

[0018] Figure 2 for Figure 1Communication frame timing diagram of UAV data collection and computing task offloading in the scenario.

[0019] Figure 3 for Figure 1 Signaling transmission process between different devices in the scenario. DETAILED DESCRIPTION

[0020] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The present invention provides a method for unmanned aerial vehicle (UAV)-assisted railway sensor data collection and computation offloading. This method utilizes a DRL algorithm to determine UAV trajectory planning, data collection, and computation offloading. The algorithm is executed in the base station (BS). Communication between the UAV, BS, and WSN is based on 5G communication technology. The UAV and BS serve as the air and ground communication platforms for the railway environment, respectively, providing wireless access, data collection, and computation offloading for the railway WSN. High-speed trains connect ground users to the BS. When the BS's computational resources are insufficient, some computation tasks can be offloaded to the train. The WSN senses the railway environment, stores the collected data in a buffer, and sends a data status indication frame to the BS. This data-rich frame only informs the BS of the current sensor data cache status. The BS executes the DRL algorithm based on the WSN's data status and sends the trajectory decision result to the UAV via a trajectory indication frame. The UAV then flies along the indicated trajectory and collects WSN data. After the UAV completes data collection, the BS sends a task offloading indication frame to the UAV based on the decision result. The UAV sends some of the task to the BS for computation, and all computation results are returned to the BS. BS acts as the data fusion center to integrate the final calculation results.

[0022] Figure 1 This is a network for collecting and computing offloading tasks of railway sensors assisted by drones of the present invention. The communication technology of this network is based on the 5G cellular network. UAV and BS serve as air and ground base stations, respectively, for access to railway WSN and high-speed trains. In this network, UAV, BS and high-speed trains are equipped with servers with different computing capabilities. The DRL algorithm is executed in the BS, and model training is performed based on historical data or experimental phase data. When the training reaches a preset number of rounds or meets the preset feedback conditions, the model parameters are used to decide the UAV flight trajectory and computing task offloading. It should be noted that the deep reinforcement learning model is not limited in this embodiment. The deep reinforcement learning model can be: deep Q learning, deep deterministic policy gradient, proximal policy optimization and other models.

[0023] The railway environmental monitoring WSN samples the environment at different sampling frequencies and transmits data status indication frames to the ground-based base station (BS). Based on the received data status indication frames and the UAV's computing resources, the BS uses a DRL model to determine the UAV's flight trajectory for data collection and offload computational tasks. The BS sends a trajectory indication frame to the UAV, which includes the locations of the SNs the UAV must visit first and then the SNs. The UAV then flies along this trajectory and wakes up the designated SNs to collect uplink sensor data. After data collection is complete, it sends a collection completion indication frame to the BS and returns to the BS. The BS also sends a computational task offload indication frame to the UAV, indicating the offloaded computational task ratio and the type of data being offloaded. The UAV then sends the collected data to the BS as instructed. If the BS's computing resources are insufficient, the BS can offload some computational tasks to the train. After the corresponding devices complete the computation, the results are returned to the BS, which acts as the data fusion center for data analysis and integration.

[0024] Figure 2 The wireless communication frame timing diagram shows a UAV completing a data collection and computation task offloading process. First, the WSN sends a data status indication frame to the ground BS. The frame includes the SN location, cache queue status, data timestamp, and data priority information. The data timestamp is the sampling time of the data, the cache queue status indicates the amount of data in the SN cache queue, and the data priority indicates the urgency of data collection and computation. The data timestamp is used to calculate the data's Area of ​​Interest (AoI). The AoI increases linearly with time until the data calculation results are returned to the BS, at which point it decreases to the time difference between data sampling, transmission, offloading, and final return to the BS. The BS then executes the DRL algorithm based on the received sensor data status indication frame and the UAV's computational resources to determine the UAV's flight trajectory and computation task offloading. The BS sends a trajectory indication frame to the UAV, which includes the locations of the SNs the UAV must visit. Based on the flight trajectory, the UAV adjusts its speed and direction to fly to the corresponding SNs and sends a wake-up indication frame to the corresponding SNs, requesting uplink data from the SNs. After data collection is complete, the UAV returns to the BS according to the trajectory indication and notifies the BS of the completion of data collection. Finally, the BS sends a data computation offload indication frame to the UAV, including the offloaded computation ratio and the type of data being offloaded. Based on this indication, the UAV then sends the computation task to the BS. If the BS's computing resources are insufficient, the BS can offload some of the computation to the train. After the corresponding devices complete the computation, the results are returned to the BS, which acts as the data fusion center for data analysis and integration.

[0025] Figure 3The corresponding signaling flow chart shows that the SN sends a data status indication frame to the base station, and the UAV informs the BS of its computing resources. The BS then executes the DRL decision algorithm based on this information. The BS then sends a trajectory indication frame to the UAV. The UAV flies along this trajectory and sends a wake-up indication frame when it reaches the corresponding SN location. The SN then sends the sensor data to the UAV. Finally, after the collection is complete, the UAV flies back to the BS according to the trajectory and informs the BS. The BS then sends a task offload indication frame to the UAV, and the UAV offloads the computing task to the BS. If the BS's computing resources are insufficient, the BS can offload some of the tasks to the train for calculation. Finally, all results are returned to the BS, and the BS calculates the algorithm reward based on the entire process.

[0026] The above description is only part of the embodiments of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the core ideas and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A method for collecting and computing railway sensor data assisted by a drone, characterized in that: Based on the deep reinforcement learning algorithm DRL, the UAV flight trajectory and computing task offloading are determined. The wireless sensor network (WSN) sensor data is collected and the task is offloaded to the calculation. While meeting the real-time requirements of railway environmental monitoring data, the WSN transmission energy consumption and the UAV flight and computing energy consumption are minimized. The specific steps include: Data collection phase of WSN: SN collects railway environment data and stores the data in the buffer area; SN sends a data status indication frame to BS, which includes the SN's location coordinates, SN buffer queue status, data timestamp, and data priority; UAV data collection phase: The base station (BS) executes the DRL algorithm to make decisions based on the WSN data status and the UAV computing capability, including the UAV data collection flight trajectory and computing task offloading. The BS sends a trajectory indication frame to the UAV, which contains the locations the UAV needs to reach in sequence. The UAV flies to the corresponding SN location based on the trajectory and sends a wake-up indication frame to the SN. The SN then uploads the sensor data in the buffer. Computational task offloading phase: After completing data collection, the UAV sends a collection completion indication frame to the BS. The BS then sends a task offloading indication frame to the UAV, which includes the offloaded data ratio and offloaded data type. The offload ratio ranges from 0 to 1. If the BS computing resources are insufficient, the BS offloads part of the task to the train for calculation, and the final calculation results are returned to the BS. Algorithm training and execution: The DRL algorithm is executed in the base station (BS). The algorithm's environmental status information includes the location coordinates of the SN, the SN buffer queue status, data timestamps, data priority, and the UAV's computing resources. The algorithm's actions include the UAV trajectory and computing task offloading decisions. The algorithm's reward indicators include the AoI of the transmitted data, transmission energy consumption, and UAV energy consumption. The algorithm trains the model using historical data or experimental data. When the training reaches the preset rounds or meets the preset feedback conditions, the BS uses the trained model to make decisions on the UAV trajectory, data collection, and computing task offloading.

2. The method for collecting and computing railway sensor data assisted by a drone according to claim 1, characterized in that: The data status indication frame is used to inform the BS of the sensor data information of the SN cache area, including SN location, cache queue status, data timestamp, and data priority; Based on this information, the BS executes the DRL algorithm to decide the flight trajectory of the UAV data collection and offload the computational tasks.

3. The method for collecting and computing railway sensor data assisted by a drone according to claim 1, characterized in that: The trajectory indication frame is used to control the flight trajectory of the UAV data collection, including the location information of the SNs that the UAV needs to reach first and then. The UAV adjusts the speed and direction based on the flight trajectory and flies to the corresponding SN positions first and then sends a wake-up indication frame to the corresponding SN, and the SN uploads the sensor data.

4. The method for collecting and computing railway sensor data assisted by a drone according to claim 1, characterized in that: The task offloading indication frame is used to inform the UAV to offload the computing task, including the offloading ratio and the offloading data type of the computing task. The UAV sends the computing task to the BS based on the indication; When the BS computing resources are insufficient, the BS offloads some tasks to the train for computation; After the corresponding device completes the calculation, the calculation results are returned to the BS.

5. A UAV-assisted railway sensor data collection and calculation unloading system, characterized in that: The method for collecting and offloading railway sensor data assisted by a drone as described in any one of claims 1 to 4 includes but is not limited to railway WSN constructed with Wi-Fi, Starflash, ZigBee and Bluetooth communication technologies and gateway equipment for network protocol conversion, as well as UAVs, ground BSs and high-speed trains, and cellular communication standards including but not limited to 4G and 5G communication technologies are used between key equipment; Among them, WSN perceives the railway operating environment and sends data status indication frames to the base station. The BS executes the DRL decision algorithm based on this information; the BS sends a trajectory indication frame to the UAV. The UAV flies based on the trajectory and sends a wake-up indication frame when it reaches the corresponding SN position. The SN sends the sensor data to the UAV; after the collection is completed, the UAV flies back to the BS according to the trajectory and informs the BS. The BS sends a task offloading indication frame to the UAV. The UAV offloads the computing task to the BS. When the BS computing resources are insufficient, the BS offloads part of the task to the train for calculation. Finally, all the results are returned to the BS.

Images