Air quality underway monitoring device based on unmanned driving

Abstract

The invention relates to an unmanned-driving-based air quality underway monitoring device which comprises an unmanned aerial vehicle main body, an air quality monitoring sensor module is arranged at the bottom of the unmanned aerial vehicle main body, and a data transmission module is arranged on the air quality monitoring sensor module. A positioning module, a control module and an energy supply module are arranged at the top of the unmanned aerial vehicle body, the control module controls the flight path, height, speed and take-off and landing points of the unmanned aerial vehicle body, a Gaussian diffusion model is constructed based on real-time monitoring data, the flight height and course are dynamically adjusted through deep reinforcement learning, and the flight speed is adjusted. An area with the maximum pollutant concentration gradient is preferentially explored. The air quality underway monitoring device is high in flexibility, high in real-time performance, low in cost, capable of monitoring multiple parameters, capable of monitoring multiple air quality monitoring parameters at the same time, capable of providing comprehensive air quality information and capable of providing a new generation of high-temporal-spatial-resolution monitoring solution for intelligent environmental protection.

Description

Technical Field

[0001] The present invention relates to an air quality monitoring device, in particular to an unmanned air quality cruise monitoring device. Background Art

[0002] With the acceleration of industrialization and urbanization, air quality is receiving increasing attention. Air pollution is becoming increasingly serious, particularly with rising concentrations of pollutants such as PM2.5, sulfur dioxide, and nitrogen oxides, posing a serious threat to human health. Existing air quality monitoring methods rely on fixed-site monitoring equipment. These devices are typically located in fixed locations, are costly, and lack real-time, comprehensive coverage of air quality conditions across the entire region, resulting in insufficient data timeliness and accuracy.

[0003] Therefore, how to develop a convenient, efficient and low-cost air quality monitoring system has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of the present invention is to provide an unmanned air quality cruise monitoring device to achieve real-time and comprehensive coverage of the air quality status of the entire monitoring target area.

[0005] To achieve the above objectives, the present invention provides an unmanned air quality cruise monitoring device, comprising a drone body, an air quality monitoring sensor module disposed at the bottom of the drone body, a data transmission module disposed on the air quality monitoring sensor module, and a positioning module, a control module, and an energy supply module disposed on the top of the drone body.

[0006] The air quality monitoring sensor module includes a PM2.5 laser scattering sensor, a VOCs photoionization detector, a SO2 electrochemical gas sensor, a NOx electrochemical gas sensor, and an O3 electrochemical gas sensor integrated in a 3D stacked package. The air quality monitoring sensor module also has a built-in micro standard gas chamber and automatically triggers sensor calibration every 30 minutes.

[0007] The positioning module integrates GPS, UWB indoor positioning chip, and visual SLAM for multimodal positioning, wherein the GPS provides a global reference, the UWB indoor positioning chip compensates for indoor positioning, and the visual SLAM constructs a local map. The absolute position of the GPS, the relative distance measurement value of the UWB, and the pose estimation of the visual SLAM are input into the EKF algorithm, and the multi-source data are dynamically integrated through the state equation and the observation equation to output the optimal position estimate;

[0008] The control module controls the flight trajectory, altitude, speed, take-off and landing points of the drone body, and builds a Gaussian diffusion model based on real-time monitoring data. It dynamically adjusts the flight altitude and heading through deep reinforcement learning, prioritizing exploration of areas with the highest pollutant concentration gradient.

[0009] The energy supply module includes a perovskite solar film and a battery set on the wing surface of the drone body, and dynamically adjusts the flight speed and sensor sampling frequency based on the energy consumption prediction model of the deep Q network.

[0010] Preferably, the air quality monitoring sensor module further includes a micro quantum dot spectrometer with a wavelength range of 250 to 1100 nm mounted on the head of the drone body.

[0011] Preferably, multiple layers of Permalloy magnetic shielding layers are provided around the motor of the drone body and around the communication module.

[0012] Preferably, after the ground station sends the coordinates of the target area, the air quality cruise monitoring device automatically calls OpenStreetMap data to generate a three-dimensional flight restricted area map, and flies around risk points.

[0013] Preferably, the positioning module dynamically adjusts the weight of each sensor according to signal quality, and the signal quality includes GPS signal-to-noise ratio and UWB link stability.

[0014] Preferably, the visual SLAM corrects the accumulated error through loop detection, while the anchor point information of the UWB indoor positioning chip is used as a global constraint.

[0015] Preferably, in terms of energy management strategy, when the lighting conditions are good, the energy consumption prediction model will automatically increase the operating speed and data collection frequency; when the energy supply is limited, it will take deceleration measures and selectively shut down some non-core sensor functions.

[0016] Preferably, the scheduling strategy of the energy supply module based on task priority includes: once the freshness of the sensor data approaches a preset threshold, the drone will be assigned to the area with priority and the sampling frequency will be increased.

[0017] Preferably, the energy supply module continuously updates network parameters through online learning technology.

[0018] Based on the above technical solution, the advantages of the present invention are:

[0019] High flexibility: It can be quickly deployed to areas that need to be monitored without being restricted by terrain, and can monitor areas that traditional monitoring stations cannot cover.

[0020] Strong real-time performance: capable of collecting and transmitting air quality data in real time.

[0021] Low cost: Compared with building and maintaining ground fixed monitoring stations, the present invention has lower costs.

[0022] Multi-parameter monitoring: It can monitor multiple air quality monitoring parameters simultaneously and provide comprehensive air quality information.

[0023] Decision support: It has achieved a leapfrog upgrade from "passive monitoring" to "active tracing", from "single-machine operation" to "swarm intelligence", and from "data collection" to "decision-making", providing a new generation of high-temporal and spatial resolution monitoring solutions for smart environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0025] Figure 1 This is a structural diagram of the air quality cruise monitoring device. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments.

[0027] The present invention provides an air quality monitoring device based on unmanned navigation, such as Figure 1 As shown, a preferred embodiment of the present invention is shown.

[0028] Specifically, if Figure 1 As shown, the air quality cruise monitoring device includes a drone body 1, an air quality monitoring sensor module 2 is provided at the bottom of the drone body 1, a data transmission module 3 is provided on the air quality monitoring sensor module 2, and a positioning module 4, a control module 5 and an energy supply module 6 are provided on the top of the drone body 1.

[0029] The drone body 1 has an overall cross-shaped structure, including: a fuselage frame, arms, a power system (propellers, brushless motors), a flight control system, a monitoring equipment installation area (equipment cabin, protective devices), and a communication system (data transmission module, antenna). It can be modified using existing commercially available drones.

[0030] The air quality monitoring sensor module 2 includes a PM2.5 laser scattering sensor, a VOCs photoionization detector, a SO2 electrochemical gas sensor, a NOx electrochemical gas sensor and an O3 electrochemical gas sensor integrated in a 3D stacked package. The air quality monitoring sensor module 2 is also equipped with a built-in micro standard gas chamber and automatically triggers sensor calibration every 30 minutes.

[0031] The present invention's 3D stacked packaging utilizes TSV technology to vertically stack and interconnect multiple chips. In 3D stacked packaging, the chips are placed very close together, resulting in lower latency. Furthermore, the shortened interconnect length reduces parasitic effects, enabling devices to operate at higher frequencies, resulting in improved performance and significantly lower costs.

[0032] The PM2.5 laser scattering sensor (particle size resolution 0.3μm), VOCs photoionization detector (PID) and electrochemical gas sensor (SO2 / NOx / O3) are integrated in the equipment cabin using 3D stacking packaging technology, reducing the volume by 60% and reducing power consumption by 40%; a miniature quantum dot spectrometer (wavelength range 250~1100nm) is installed on the head of the drone to improve the accuracy of VOCs fingerprint spectral recognition through the quantum confinement effect, and can detect benzene series compounds with concentrations as low as 0.1ppb.

[0033] The air quality monitoring sensor module 2 has a built-in micro standard gas chamber (including NIST-certified zero gas / standard gas), which automatically triggers sensor calibration every 30 minutes to eliminate the impact of temperature and humidity drift.

[0034] The positioning module 4 integrates GPS, UWB indoor positioning chip, and visual SLAM for multimodal positioning, inputs the absolute position of GPS, the relative distance measurement value of UWB, and the pose estimation of visual SLAM into the EKF algorithm, dynamically fuses multi-source data through state equations and observation equations, and outputs the optimal position estimate.

[0035] Furthermore, the positioning module 4 utilizes a three-dimensional stacked hardware layout (e.g., three layers of stacked circuit boards), integrating a GPS module (e.g., the UBlox NEO-M8N), a UWB chip, and the visual SLAM camera / IMU components into a single platform. This physical isolation reduces electromagnetic interference. The module is encapsulated using lightweight, high-strength materials such as magnesium alloy, ensuring a total weight of less than 500 grams, meeting the drone's payload limits.

[0036] GPS provides a global baseline: In open outdoor environments, GPS outputs latitude, longitude, and altitude information as the foundational positioning data for the global coordinate system. UWB compensates for indoor positioning: In indoor scenarios where GPS signals are blocked, UWB uses ultra-wideband ranging technologies (such as ToF or TDoA) to achieve centimeter-level relative positioning and correct for position drift. Visual SLAM builds local maps: A camera captures environmental feature points in real time, combined with motion data from an IMU (gyroscope and accelerometer) to construct a high-precision local map and infer position and pose. This compensates for the shortcomings of other positioning sources, especially in dynamic or markerless environments.

[0037] Environmentally adaptive weight allocation: Dynamically adjusts the weight of each sensor based on signal quality (such as GPS signal-to-noise ratio and UWB link stability). For example, GPS is primarily used in outdoor environments, while UWB and visual SLAM are used indoors, with inertial navigation (IMU) used to compensate for short signal interruptions.

[0038] Loop closure detection and relocalization: Visual SLAM corrects accumulated errors through loop closure detection. At the same time, UWB anchor information can be used as a global constraint to further improve system robustness. The positioning accuracy remains at ±0.3m in areas blocked by satellite signals.

[0039] Furthermore, a digital compass anti-magnetic interference design is adopted, and multiple layers of permalloy magnetic shielding layers are provided around the motor of the drone body 1 and around the communication module 3, so that the heading angle measurement error is less than 0.5°.

[0040] The control module 5 controls the flight trajectory, altitude, speed, take-off and landing points of the drone body 1, and constructs a Gaussian diffusion model based on real-time monitoring data. It dynamically adjusts the flight altitude and heading through deep reinforcement learning, and prioritizes exploring areas with the largest pollutant concentration gradient.

[0041] Specifically, the multimodal action space approach breaks through traditional reinforcement learning strategies. It utilizes a Gaussian diffusion model through a reverse denoising process to model multimodal distributions. The Diffusion-QL algorithm employs this approach: In offline reinforcement learning, the diffusion model directly generates action sequences and optimizes the policy using a loss function. Experiments show that it outperforms traditional methods (such as BC-CVAE) on multimodal datasets. The Diffusion Policy treats action generation as a stepwise denoising process, enabling flexible control of diverse driving styles (aggressive / conservative). Data augmentation improves sample efficiency and generalization. The diffusion model generates high-quality synthetic data, addressing the data scarcity problem in reinforcement learning. The S4RL algorithm blends data generated by the diffusion model with real datasets to enhance the generalization capability of offline reinforcement learning. The Cloud World Model utilizes the diffusion model to construct a virtual-real world simulation environment, accelerating training for long-tail scenarios through closed-loop reinforcement learning. Trajectory Planning: With its long-term and dynamic adaptability, the diffusion model combines efficiency and flexibility in trajectory generation. Fast Trajectory Generation: High-quality trajectories can be generated in just 2-3 steps, achieving a 5x improvement in efficiency over traditional methods and supporting dynamic adjustments. Diffuser algorithm: Models the joint distribution of action-state sequences, enables long-term planning, and optimizes future trajectories through guided sampling.

[0042] Furthermore, the data transmission module (3) includes:

[0043] (1) Onboard edge AI processor: Jetson Orin Nano running the YOLOv8s model - integrated with a micro mass spectrometer for pollutant component analysis, real-time pollution source identification and in-depth detection;

[0044] (2) Using knowledge distillation technology, the LSTM pollution prediction model trained on the cloud is compressed to less than 10MB, realizing the blockchain-edge computing integration of airborne pollution trend deduction for the next 30 minutes;

[0045] (3) Deploy a Hyperledger Fabric light node on the airborne end to monitor data and generate SHA-256 hash chain evidence in real time. FPGA accelerates ZK-SNARKs (Groth16 protocol) AES-256 encrypted data offline storage (ZK-SNARKs) technology to achieve zero-knowledge proof data sharing.

[0046] (4) Using software-defined radio (SDR) technology, dynamic frequency hopping (50Hz switching rate) is achieved in the 902-928MHz frequency band, and the anti-interference capability is improved by 10 times compared with traditional digital transmission;

[0047] (5) Use the developed LoRa+5G dual-mode communication module, give priority to the 5G slicing network in urban environments (end-to-end latency <20ms), and switch to the LoRa long-distance mode (transmission distance up to 15km) in suburban areas;

[0048] The energy supply module (6) includes a perovskite solar film and a storage battery arranged on the wing surface of the drone body (1), and dynamically adjusts the flight speed and sensor sampling frequency based on an energy consumption prediction model of a deep Q network.

[0049] Specifically, the energy consumption prediction model framework: the state space is composed of multiple key parameters, including but not limited to energy data (such as solar power generation and remaining battery power), flight control parameters (speed and altitude, etc.), mission-related requirements (freshness of sensor information, priority of data processing) and external environmental factors (light intensity, potential obstacle location).

[0050] The range of operations covers adjusting the flight speed (divided into three levels: high, medium and low), changing the frequency of sensor data collection based on mission criticality, and improving the flight trajectory by reducing energy consumption during turns.

[0051] The reward mechanism aims to achieve a balance between energy efficiency and task execution quality. It achieves this goal by incentivizing low energy consumption and reducing data latency, while imposing corresponding penalties for insufficient power or failure to successfully complete the task.

[0052] Dynamic adjustment logic:

[0053] In terms of energy management strategy, when lighting conditions are good, the system will automatically increase the operating speed and data collection frequency to speed up task execution efficiency; when energy supply is relatively limited, it will take deceleration measures and selectively shut down some non-core sensor functions to achieve the purpose of extending working time.

[0054] Scheduling strategy based on task priority: Once the sensor data freshness (AoI) approaches a preset threshold, drones are assigned to that area first and their sampling frequency is increased. This approach aims to ensure task execution reliability by temporarily increasing local energy consumption.

[0055] Real-time environmental response mechanism: Continuously updates network parameters through online learning technology to flexibly respond to sudden obstacles or changing weather conditions. For example, when a detour is required, the system can automatically adjust the path planning strategy, effectively reducing energy consumption fluctuations.

[0056] Technical Advantages: End-to-end Optimization: DQN autonomously learns optimal strategies in complex environments, overcoming the limitations of traditional heuristic rule design and achieving an 8% to 30% reduction in energy consumption in dynamic scenarios. A multi-objective collaborative strategy aims to simultaneously improve flight operations, perception, and path planning decisions across a wide range of state spaces. This approach not only improves the system's sustained operational capabilities but also enhances mission execution efficiency.

[0057] Furthermore, the integration of a perovskite solar film (with a 28% conversion efficiency) on the wing surface, combined with a bionic circuit design inspired by dragonfly wing veins, extends the average daily flight time by three hours. An energy consumption prediction model based on a Deep Q-Network (DQN) dynamically adjusts flight speed and sensor sampling frequency to ensure stable communication with the ground control station during the drone's mission.

[0058] The air quality cruise monitoring device of the present invention can monitor air quality by referring to the following steps:

[0059] (1) Autonomous task decomposition: After the ground station sends the target area coordinates, the drone automatically uses OpenStreetMap data to generate a three-dimensional flight restricted area map to avoid risk points such as high-voltage lines and high-rise buildings;

[0060] (2) When an emergency pollution incident is detected, it automatically switches to "swarm emergency mode" and drone clusters within a 10km radius autonomously assemble to form a dynamic monitoring network; the Trusted Execution Environment (TEE) is embedded in the ARM TrustZone security zone in the flight control chip to protect the core algorithms and communication keys from malicious attacks.

[0061] The air quality cruise monitoring device of this invention supports real-time 3D pollution digital twins. Using the LiDAR (Velodyne VLP-16) and multispectral camera onboard a drone, a centimeter-level 3D model of the monitored area is constructed. The Unity3D engine is used in the cloud to dynamically render the pollutant diffusion process, allowing users to view the "virtual pollution cloud" overlaid with the real scene through AR glasses. The specific process is as follows:

[0062] (1) Data collection and transmission: The drone is equipped with a Velodyne VLP-16 LiDAR (centimeter-level point cloud) and a multispectral camera to simultaneously collect spectral and spatial data of pollution sources. It is combined with a catalytic oxidation sensor to monitor more than a thousand gas parameters in real time and upload the data to the cloud via the 5G network.

[0063] (2) Three-dimensional modeling and diffusion simulation: DJI Terra is used in the cloud to fuse LiDAR point clouds and multispectral images to build a centimeter-level real-scene three-dimensional model. Combined with meteorological data to drive the numerical model, the diffusion path and concentration distribution of the pollution cloud are dynamically simulated.

[0064] (3) Real-time rendering and AR interaction: The Unity3D engine maps pollution data into a dynamic particle system and pushes it to the terminal through WebGL or streaming media; AR glasses overlay the virtual pollution cloud on the real scene based on SLAM technology, and support gesture interaction to view concentration and diffusion trends.

[0065] (4) Multi-terminal collaborative decision-making: The web platform provides three-dimensional heat maps, pollution source tracing and historical backtracking functions, and the mobile terminal generates law enforcement reports; the system integrates meteorological forecasts and historical data, optimizes the layout of monitoring sites and issues early warnings.

[0066] The air quality monitoring device of the present invention can also implement learning-assisted decision-making: a cross-regional federated learning platform is established, each drone locally trains a pollution feature extraction model, and uses homomorphic encryption to exchange model parameters rather than raw data. A generative adversarial network (GAN) simulates the effects of different treatment plans and outputs a report recommending the optimal emission reduction path, as follows:

[0067] (1) Build a cross-regional federated learning system to promote data sharing and model training among regions, thereby enhancing the learning efficiency and prediction accuracy of the algorithm while ensuring the data security of the participants.

[0068] (2) Localized training of pollution feature recognition models for each UAV: ​​By independently executing the training process of the pollution feature recognition model on each UAV, the aim is to enhance the model's adaptability to different environmental conditions and its actual application efficiency.

[0069] (3) Using homomorphic encryption technology to achieve secure exchange of model parameters: With the help of homomorphic encryption methods, the transmission and update process of model parameters can be completed without directly sharing the original data, thereby significantly reducing the risk of information leakage.

[0070] (4) Using generative adversarial networks (GANs) to simulate the effects of various governance measures: Through GAN technology, different governance strategies can be effectively simulated, and the effectiveness of these strategies can be evaluated based on the generated results, thereby providing more accurate support for decision-making.

[0071] (5) Provide a recommendation report on the optimal emission reduction path: Based on the above analysis process, we can extract the most effective emission reduction strategy recommendations, thereby providing decision makers with more specific and accurate reference basis.

[0072] The air quality cruise monitoring device of the present invention is highly flexible and can be quickly deployed to areas that need to be monitored. It is not restricted by terrain and can monitor areas that traditional monitoring stations cannot cover. It has strong real-time performance and can collect and transmit air quality data in real time. It is low-cost and is lower than the cost of building and maintaining fixed ground monitoring stations. It has multi-parameter monitoring and can monitor multiple air quality monitoring parameters at the same time to provide comprehensive air quality information. It provides decision support and has achieved a leap-forward upgrade from "passive monitoring" to "active tracing", from "single-machine operation" to "swarm intelligence", and from "data collection" to "decision-making", providing a new generation of high-temporal and spatial resolution monitoring solutions for smart environmental protection.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention. They should all be included in the scope of the technical solution for protection of the present invention.

Claims

1. An unmanned air quality cruise monitoring device, comprising an unmanned aerial vehicle (UAV) body (1), characterized in that: An air quality monitoring sensor module (2) is provided at the bottom of the drone body (1), a data transmission module (3) is provided on the air quality monitoring sensor module (2), and a positioning module (4), a control module (5), and an energy supply module (6) are provided on the top of the drone body (1); The air quality monitoring sensor module (2) includes a PM2.5 laser scattering sensor, a VOCs photoionization detector, a SO2 electrochemical gas sensor, a NOx electrochemical gas sensor, and an O3 electrochemical gas sensor integrated in a 3D stacked package. The air quality monitoring sensor module (2) is also equipped with a built-in micro standard gas chamber and automatically triggers sensor calibration every 30 minutes. The positioning module (4) integrates GPS, UWB indoor positioning chip, and visual SLAM to perform multimodal positioning, wherein the GPS provides a global reference, the UWB indoor positioning chip compensates for indoor positioning, and the visual SLAM constructs a local map. The absolute position of the GPS, the relative distance measurement value of the UWB, and the pose estimation of the visual SLAM are input into the EKF algorithm, and the multi-source data are dynamically integrated through the state equation and the observation equation to output the optimal position estimate; The control module (5) controls the flight trajectory, altitude, speed, take-off and landing points of the UAV body (1), and constructs a Gaussian diffusion model based on real-time monitoring data, dynamically adjusts the flight altitude and heading through deep reinforcement learning, and prioritizes exploring areas with the maximum pollutant concentration gradient; The energy supply module (6) includes a perovskite solar film and a storage battery arranged on the wing surface of the drone body (1), and dynamically adjusts the flight speed and sensor sampling frequency based on an energy consumption prediction model of a deep Q network.

2. The air quality cruise monitoring device according to claim 1, characterized in that: The air quality monitoring sensor module (2) further comprises a micro quantum dot spectrometer with a wavelength range of 250-1100 nm, which is mounted on the head of the drone body (1).

3. The air quality cruise monitoring device according to claim 1, characterized in that: Multiple layers of Permalloy magnetic shielding layers are provided around the motor of the drone body (1) and around the communication module (3).

4. The air quality cruise monitoring device according to claim 1, characterized in that: After the ground station sends the coordinates of the target area, the air quality cruise monitoring device automatically calls OpenStreetMap data to generate a three-dimensional flight restricted area map, and flies around risk points.

5. The air quality cruise monitoring device according to claim 1, characterized in that: The positioning module (4) dynamically adjusts the weight of each sensor according to signal quality, and the signal quality includes GPS signal-to-noise ratio and UWB link stability.

6. The air quality cruise monitoring device according to claim 5, characterized in that: The visual SLAM corrects the accumulated error through loop closure detection, while the anchor point information of the UWB indoor positioning chip is used as a global constraint.

7. The air quality cruise monitoring device according to claim 1, characterized in that: In terms of energy management strategy, the energy supply module (6) automatically increases the operating speed and data collection frequency when the lighting conditions are good; and when the energy supply is limited, it takes deceleration measures and selectively shuts down some non-core sensor functions.

8. The air quality cruise monitoring device according to claim 1, characterized in that: The scheduling strategy of the energy supply module (6) based on task priority includes: once the freshness of the sensor data approaches a preset threshold, the drone will be assigned to the area first and the sampling frequency will be increased.

9. The air quality cruise monitoring device according to claim 1, characterized in that: The energy supply module (6) continuously updates network parameters through online learning technology.

Images