A pavement pollutant sweeping system and deployment method

Abstract

The application provides a road surface pollutant cleaning system and a deployment method, which comprises: the system is composed of a remote sensing satellite, a road inspection sensing terminal, a distributed edge computing module, a cloud management and control platform and an intelligent cleaning unmanned vehicle. A "sky-ground-end" three-level linkage architecture is adopted, and through multi-dimensional interaction of intelligent edge nodes and autonomous mobile platforms, intelligent sensing and accurate cleaning of urban road surfaces are realized. The system adopts distributed computing, breaks through the performance bottleneck of traditional centralized control, can make up for the problems of insufficient field of view of the intelligent cleaning unmanned vehicle and insufficient response time of abnormal working condition processing, and significantly improves the efficiency from finding garbage to cleaning garbage.

Description

Technical Field

[0001] This invention relates to the field of intelligent waste cleaning technology, and in particular to a road pollutant cleaning system and deployment method. Background Technology

[0002] Road cleaning is typically done using intelligent unmanned vehicles. However, these vehicles face several challenges in urban environmental management: their limited field of vision prevents them from achieving full coverage of city roads; they are susceptible to various factors such as urban traffic and flow, making optimal path planning and scheduling difficult; and the intelligent system design, which involves roadside data collection and centralized computation, increases computational burden and results in the transmission of invalid images, leading to low efficiency in current intelligent cleaning systems.

[0003] Existing solutions include systems comprising roadside monitoring units, cloud-based monitoring and management platforms, and autonomous cleaning vehicles. The roadside monitoring units perform image monitoring of pre-defined service areas and generate image data. The cloud-based monitoring and management platform uses object detection algorithms to identify litter within the monitored area and sends task messages to the autonomous cleaning vehicles. The autonomous cleaning vehicles then perform image recognition, location tracking, and cleaning based on the tasks sent by the cloud management platform to complete the task. It can be seen that this solution only utilizes the near-end monitoring unit and the remote cloud platform for collaborative work, failing to address the collaborative scheduling of multiple areas, resulting in significant limitations in efficiency and scope of application. Summary of the Invention

[0004] This invention provides a road pollutant cleaning system and deployment method to address the shortcomings of existing technologies. It adopts a three-level linkage architecture of "sky-ground-end" and achieves intelligent perception and precise cleaning of urban roads through multi-dimensional interaction of intelligent edge nodes and autonomous mobile platforms.

[0005] In a first aspect, the present invention provides a road surface pollutant cleaning system, comprising:

[0006] The system consists of remote sensing satellites, roadside inspection and sensing terminals, distributed edge computing modules, cloud management and control platforms, and intelligent cleaning unmanned vehicles. The cloud management and control platform is the main control unit, the distributed edge computing modules are edge units, and the remote sensing satellites, inspection and sensing terminals, and intelligent cleaning unmanned vehicles are slave units. The main control unit, edge units, and slave units are connected via a wireless network.

[0007] The remote sensing satellite and the roadside inspection and sensing terminal work together to perform synchronous positioning and mapping of the target area;

[0008] The remote sensing satellite detects traffic conditions and traffic flow in the target area and uploads the detection results to the cloud-based management and control platform.

[0009] The distributed edge computing module identifies road pollutants and uploads the identification results to the cloud-based management and control platform.

[0010] The cloud-based management platform transmits information on road pollutants, traffic conditions in the target area, and traffic flow information to each intelligent unmanned sweeping vehicle.

[0011] The intelligent unmanned sweeping vehicles perform global planning for their respective road areas, reach the target garbage point based on the planned path, and clean up the pollutants on the road surface using the sweeping operation device.

[0012] According to the present invention, a road pollutant cleaning system is provided, wherein the remote sensing satellite includes an onboard computing platform, a LiDAR, an optical camera, a synthetic aperture radar, an inertial measurement unit, a star sensor, a GNSS module, a data transceiver module, and a power supply module;

[0013] The onboard computing platform receives information and control commands from the lidar, optical camera, synthetic aperture radar, inertial measurement unit, and GNSS module. The lidar acquires three-dimensional point cloud data, the optical camera extracts and matches features from urban road images, the synthetic aperture radar acquires all-weather radar images, the inertial measurement unit measures acceleration and angular velocity, the GNSS module provides absolute position information, and the communication module is used for wireless communication between the remote sensing satellite and the outside world.

[0014] The remote sensing satellite uses multi-source sensor fusion technology for global positioning and environmental map construction. It uses SLAM algorithm to process onboard sensor data and establishes high-precision three-dimensional point cloud maps or digital elevation models of the Earth's surface. During the map construction process, it combines satellite attitude data and orbital information to achieve accurate geographic positioning and terrain modeling on a global scale.

[0015] The surface images are acquired by optical cameras or synthetic aperture radar. Deep learning image classification and target detection algorithms deployed on cloud servers are used to automatically identify garbage accumulation areas in the surface images. The coordinate information of the identified garbage areas is combined with satellite positioning data and converted to the world coordinate system.

[0016] The remote sensing satellite uses satellite video or radar remote sensing data, combined with optical flow analysis algorithms or vehicle trajectory tracking algorithms, to achieve real-time monitoring of road traffic flow. By extracting vehicle motion vectors and calculating road cross-section traffic flow density, a dynamic traffic flow map is generated, and the dynamic traffic flow map is used for global path optimization.

[0017] According to the present invention, a road pollutant cleaning system is provided, wherein the roadside inspection and sensing terminal includes a first end-side computing unit, a first RGB-D vision module, a first solid-state LiDAR, a first GNSS module, a communication module, and a power module;

[0018] The first end-side computing unit receives information and control commands from the first RGB-D vision module, the first solid-state LiDAR and the first GNSS module. The first RGB-D vision module acquires road pollutant information, the first solid-state LiDAR acquires surrounding environmental information, and SLAM technology is used for positioning and mapping. The first GNSS module realizes the positioning of the roadside inspection and sensing terminal, and the communication module realizes wireless communication between the roadside inspection and sensing terminal and the outside world.

[0019] The positioning, mapping, and waste identification of the roadside inspection and sensing terminal include the roadside inspection and sensing terminal using SLAM technology to build a two-dimensional grid map of the target road segment, using a deep learning target detection algorithm to identify pollutants on the road surface, calculating the XYZ coordinates of the point cloud information within the identification box, converting the XYZ coordinates to the world coordinate system, and sending the converted coordinates to the intelligent cleaning unmanned vehicle through a cloud server.

[0020] According to the present invention, a road pollutant cleaning system is provided, wherein the distributed edge computing module includes an MEC computing chip and a communication private network slice;

[0021] The MEC computing chip receives remote sensing, video, and image data sent by remote sensing satellites and roadside inspection and sensing terminals, analyzes and identifies traffic communication conditions in real time, and identifies the name, category, and location information of road pollutants.

[0022] The dedicated communication network slice provides independent bandwidth to enable wireless communication between the distributed edge computing module and the outside world.

[0023] According to the present invention, a road pollutant cleaning system is provided, wherein the cloud management platform is responsible for assigning tasks to intelligent cleaning unmanned vehicles, monitoring the status information of intelligent cleaning unmanned vehicles, adjusting control commands, and realizing the interaction between intelligent cleaning unmanned vehicles and environmental information.

[0024] According to the present invention, a road pollutant cleaning system is provided, wherein the intelligent cleaning unmanned vehicle is equipped with a MEMS inertial measurement unit, a second GNSS module, a second end-side computing platform, a second RGB-D vision module, a second solid-state LiDAR, a cleaning operation device, a garbage storage device, a drive device, and a communication module.

[0025] The second solid-state LiDAR acquires information about the surrounding environment and uses SLAM technology for positioning and navigation. The second GNSS module provides positioning for the intelligent cleaning unmanned vehicle. The second RGB-D vision module acquires information about road pollutants. The MEMS inertial measurement unit and the second GNSS module use the UKF filtering algorithm to fuse time, position and velocity information. After feedback correction, the time, position and velocity are output.

[0026] The MEMS inertial measurement unit calculates the velocity acceleration, angular velocity, and yaw angle through the inertial system to obtain time, position, and velocity; the second GNSS module calculates time, position, and velocity through the navigation system.

[0027] The intelligent cleaning unmanned vehicle's positioning, mapping, and garbage identification include the cleaning device using SLAM technology for positioning and mapping, a second GNSS module providing positioning for the cleaning device, and a second RGB-D vision module identifying garbage information, including garbage name, category, and location information.

[0028] The coordinate transformation of the intelligent cleaning unmanned vehicle includes treating the motion of the intelligent cleaning unmanned vehicle in the plane as rigid body motion and transforming it to the world coordinate system through homogeneous coordinate transformation;

[0029] The positioning of the intelligent cleaning unmanned vehicle includes obtaining the position of the intelligent cleaning unmanned vehicle on the map, using a MEMS inertial measurement unit and a second GNSS module for global positioning, using the UKF algorithm for data fusion, estimating the state error of the inertial navigation and the second GNSS module, and correcting each navigation system through the estimation feedback of the state error to improve positioning accuracy.

[0030] The path planning of the intelligent cleaning unmanned vehicle includes dividing a two-dimensional grid map into regions by a cloud-based control platform, obtaining the world coordinate system of each intelligent cleaning unmanned vehicle, obtaining the garbage coordinate relationship of the area under the responsibility of the intelligent cleaning unmanned vehicle through coordinate transformation, and planning the path from near to far based on the garbage coordinate information of the area under its responsibility and the road traffic congestion. In the navigation, autonomous navigation and path planning algorithms are used to perform global path planning and calculate the optimal route from the intelligent cleaning unmanned vehicle to the target location.

[0031] The intelligent cleaning unmanned vehicle cleans up road pollutants by using a point cloud generated by a second RGB-D vision module after reaching a designated location, detecting road pollutants, calculating the coordinates of the point cloud within the recognition box, converting the acquired garbage coordinates to the intelligent cleaning unmanned vehicle coordinate system, and then starting the intelligent cleaning unmanned vehicle to move and start the cleaning operation device to clean up the garbage.

[0032] Secondly, the present invention also provides a method for deploying a road surface pollutant cleaning system, comprising:

[0033] Initialize the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle, and determine the parameters of the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle.

[0034] Establish a wireless network connection between the remote sensing satellite, the roadside inspection and sensing terminal, the cloud management and control platform, the distributed edge computing module, and the intelligent cleaning unmanned vehicle;

[0035] The remote sensing satellite and the roadside inspection and sensing terminal respectively collect data from the target area. The collected data is processed by the distributed edge computing module and then transmitted back to the cloud management and control platform.

[0036] The cloud-based management platform establishes a pollutant distribution network based on the returned data and issues task instructions to the intelligent unmanned cleaning vehicle.

[0037] The intelligent cleaning unmanned vehicle, based on the planned route obtained by autonomous navigation and path planning algorithms, arrives at the areas where each pollutant is located in turn, identifies the road pollutants, starts the cleaning operation device, and moves to the next task target area after completing the cleaning task.

[0038] After completing all cleaning tasks, the intelligent cleaning unmanned vehicle returns to the work standby area.

[0039] According to a method for deploying a road surface pollutant cleaning system provided by the present invention, the remote sensing satellite and the roadside inspection and sensing terminal respectively collect data on the target area, including:

[0040] Start SLAM mapping and create a 2D raster map;

[0041] Using a deep learning target detection algorithm, the remote sensing satellite identifies road pollutants, traffic conditions, and vehicle flow, while the roadside inspection and sensing terminal identifies road pollutants through a first RGB-D vision module.

[0042] The remote sensing satellite transmits road pollutants, traffic conditions, and vehicle flow to the distributed edge computing module, and the road inspection and sensing terminal transmits road pollutants to the distributed edge computing module.

[0043] The data collected by the distributed edge computing module is processed and then transmitted back to the cloud management platform, including:

[0044] Create a two-dimensional map and garbage coordinate information;

[0045] Transform the coordinates of the identified road pollutants to the world coordinate system;

[0046] Based on the identified traffic conditions and traffic flow, predict the duration of congestion and the duration of smooth traffic flow;

[0047] The prediction results are then sent back to the cloud-based management platform.

[0048] According to a deployment method of a road pollutant cleaning system provided by the present invention, the cloud-based management platform establishes a pollutant distribution network based on the returned data and issues task instructions to the intelligent cleaning unmanned vehicle, including:

[0049] The returned 2D map, garbage coordinate information, and traffic condition prediction information are processed.

[0050] The responsible road sections for intelligent cleaning unmanned vehicles are determined by dividing the area based on a two-dimensional map, the number of vehicles, and the target urban area.

[0051] Transform the coordinates of each intelligent unmanned cleaning vehicle to the world coordinate system;

[0052] The system iterates through the pollutant coordinates and traffic condition predictions of the areas covered by each intelligent cleaning drone to perform global path planning.

[0053] The coordinates of the garbage in the target sub-area are transmitted to the intelligent cleaning unmanned vehicle responsible for the target sub-area.

[0054] According to a method for deploying a road pollutant cleaning system provided by the present invention, the intelligent cleaning unmanned vehicle sequentially reaches each pollutant area based on a planned route obtained by an autonomous navigation and path planning algorithm, including:

[0055] Receive garbage coordinate information and traffic condition prediction information in the world coordinate system;

[0056] Based on the coordinates of the garbage and the traffic congestion situation, waypoints are set from near to far.

[0057] Global path planning is performed based on the grid map. The optimal path is determined by combining the traditional Dijkstra algorithm and the intelligent PSO algorithm. The combination of the traditional Dijkstra algorithm and the intelligent PSO algorithm includes a particle encoding method using key node sequences and attention marker vectors, a particle fitness evaluation mechanism based on local relaxation improvement, and a priority queue scheduling function that integrates global attention.

[0058] Based on the optimal route, the aircraft will travel along the waypoints to the areas where the pollutants are located.

[0059] Identify road pollutants and activate the sweeping operation equipment, including:

[0060] Based on the point cloud information within the target bounding box, the coordinate information of the garbage is obtained, and the relative coordinates between the garbage and the cleaning device are obtained through coordinate transformation.

[0061] Based on relative coordinates, the robot operating system motion framework is used to plan the travel route of the intelligent cleaning unmanned vehicle. The cleaning device reaches the garbage location according to the travel route and performs garbage cleaning.

[0062] The road pollutant cleaning system and deployment method provided by this invention adopts a three-level linkage architecture of "sky-ground-edge," achieving intelligent perception and precise cleaning of urban roads through multi-dimensional interaction between intelligent edge nodes and autonomous mobile platforms. This compensates for the insufficient field of view and inadequate response time in handling abnormal conditions of intelligent cleaning unmanned vehicles, significantly improving the efficiency from finding and cleaning up litter. Furthermore, by employing distributed computing and supported by an edge computing platform, it completes local target recognition tasks, solving the efficiency problem of centralized control in the management center, and further improving the work efficiency of urban road pollutant cleaning. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0064] Figure 1 This is a general framework diagram of the road pollutant cleaning system provided by the present invention;

[0065] Figure 2 This is a schematic diagram of distributed computing provided by the present invention;

[0066] Figure 3 This is a schematic diagram of the autonomous navigation and path planning method for the intelligent cleaning unmanned vehicle provided by the present invention;

[0067] Figure 4 This is a schematic diagram of the deployment method of the road pollutant cleaning system provided by the present invention;

[0068] Figure 5 This is a flowchart of the "sky-ground-end" three-level linkage architecture provided by the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0070] Figure 1 This is a general framework diagram of the road pollutant cleaning system provided in the embodiments of the present invention, as shown below. Figure 1 As shown, it includes:

[0071] Remote sensing satellites, roadside inspection and sensing terminals, distributed edge computing modules, cloud-based management and control platforms, and intelligent unmanned cleaning vehicles;

[0072] In this system, the cloud-based management platform serves as the master controller, while the distributed edge computing modules, remote sensing satellites, inspection and sensing terminals, and intelligent cleaning drones act as slave devices. The master controller, edge devices, and slave devices are connected via a wireless network. The master controller is the core of the system, responsible for assigning tasks to slave devices, monitoring their status, adjusting control commands, and facilitating interaction between slave devices and their environment. Edge devices are deployed in the edge computing environment, close to the data source or user, handling local tasks to reduce latency and improve response speed. Edge devices possess independent computing capabilities and collaborate with the master controller or other edge nodes. Slave devices primarily execute tasks assigned by the master controller. Figure 2 As shown, the slave machine is responsible for collecting data from the data source. The edge machine divides the data into multiple sub-tasks and distributes the sub-tasks to multiple edge computing nodes. The master machine integrates the calculation results, performs task allocation and status monitoring, and sends them to the slave machine. The intelligent cleaning unmanned vehicle then goes to the target point to perform the cleaning task.

[0073] It is understandable that the cloud-based management and control platform and the distributed edge computing module belong to the data computing and processing layer, the remote sensing satellite and the roadside inspection and sensing terminal belong to the data monitoring and detection layer, and the intelligent cleaning unmanned vehicle belongs to the control and execution layer.

[0074] Specifically, remote sensing satellites and roadside inspection and sensing terminals work together to perform synchronous positioning and mapping of the target area. The remote sensing satellites detect traffic conditions and traffic flow in the target area and upload the data to the cloud management platform. The distributed edge computing module identifies road pollutants and uploads the data to the cloud management platform. The cloud management platform transmits the road pollutants to each intelligent unmanned sweeping vehicle. The cloud management platform also transmits traffic conditions and traffic flow information of the target area to each intelligent unmanned sweeping vehicle. The intelligent unmanned sweeping vehicle performs global planning for the road area it is responsible for, reaches the garbage point, and cleans the road pollutants using its sweeping equipment.

[0075] The remote sensing satellite carries an onboard computing platform, LiDAR, optical camera, synthetic aperture radar (SAR), inertial measurement unit (IMU), star sensor, GNSS module, data transceiver module, and power supply module. The onboard computing platform receives data from the LiDAR, optical camera, SAR, IMU, GNSS module, and control commands. The LiDAR acquires 3D point cloud data, the optical camera acquires urban road images for feature extraction and matching, the SAR acquires all-weather radar images, the IMU measures acceleration and angular velocity, and the GNSS module provides absolute position information. The communication module is used for wireless communication between the remote sensing satellite and the outside world. The remote sensing satellite uses multi-source sensor fusion technology (such as optical camera, SAR, and LiDAR) for global positioning and environmental mapping. It utilizes SLAM algorithms to process onboard sensor data and establish high-precision 3D point cloud maps or digital elevation models (DEMs) of the Earth's surface. During map construction, precise global geolocation and terrain modeling are achieved by combining satellite attitude data (provided by star sensors and inertial measurement units, IMUs) and orbital information (acquired by GPS receivers). High-resolution optical cameras or synthetic aperture radars onboard remote sensing satellites capture surface images. Deep learning image classification and target detection algorithms deployed on cloud servers automatically identify garbage accumulation areas in the images. The coordinates of the identified garbage areas, combined with satellite positioning data, can be converted to the world coordinate system. Remote sensing satellites utilize satellite video or radar remote sensing data, combined with optical flow analysis or vehicle trajectory tracking algorithms from computer vision, to achieve real-time monitoring of road traffic flow. By extracting vehicle motion vectors and calculating road cross-section traffic density, dynamic traffic flow maps are generated. This data is further used for urban traffic planning, congestion warning, and global path optimization for intelligent driving systems.

[0076] The roadside inspection and sensing terminal is equipped with a first edge computing unit, a first RGB-D vision module, a first solid-state LiDAR, a first GNSS module, a communication module, and a power module. The first edge computing unit receives data from the first RGB-D vision module, the first solid-state LiDAR, and control commands. The first RGB-D vision module acquires information about road pollutants, including pollutant names, categories, and locations. The first solid-state LiDAR acquires information about the surrounding environment and uses SLAM technology for localization and mapping. The first GNSS module enables the inspection and sensing terminal to locate itself. The communication module enables wireless communication between the inspection and sensing terminal and the outside world. The localization, mapping, and waste identification of the roadside inspection and sensing terminal include: the terminal uses SLAM technology to locate and map the target road segment, creating a two-dimensional grid map. During mapping, a deep learning object detection algorithm is used to identify road pollutants. The XYZ coordinates of the point cloud information within the identification box are calculated, converted to the world coordinate system, and sent to the intelligent cleaning unmanned vehicle via a cloud server.

[0077] The distributed edge computing module is equipped with a mobile edge computing (MEC) chip and a dedicated communication network slice. The MEC chip receives remote sensing, video, and image data sent by remote sensing satellites and roadside inspection and perception terminals, and analyzes and identifies traffic communication conditions, road pollutant names, categories, and location information in real time. The dedicated communication network slice provides independent dedicated bandwidth to enable wireless communication between the distributed edge computing module and the outside world.

[0078] The cloud-based management platform is responsible for assigning tasks to the intelligent cleaning drones, monitoring their status, adjusting control commands, and enabling interaction between the intelligent cleaning drones and the environment.

[0079] The intelligent cleaning unmanned vehicle is equipped with a Micro-Electro-Mechanical System (MEMS) inertial measurement unit, a second GNSS module, a second end-side computing platform, a second RGB-D vision module, a second solid-state LiDAR, a cleaning device, a waste storage device, a drive unit, and a communication module. The second solid-state LiDAR acquires information about the surrounding environment and uses Simultaneous Localization and Mapping (SLAM) technology for positioning and navigation. The second GNSS module provides positioning for the intelligent cleaning unmanned vehicle. The second RGB-D vision module acquires information about road pollutants, including pollutant name, category, and location information. The MEMS inertial measurement unit and the second GNSS module use an unscented Kalman filter (UKF) algorithm to fuse time, position, and velocity information. The final output, after feedback correction, shows time, position, and velocity. Specifically, the MEMS inertial measurement unit calculates time, position, and velocity using velocity acceleration, angular velocity, and yaw angle through the inertial system, while the second GNSS module calculates time, position, and velocity through the navigation system. Sweeping equipment cleans up road pollutants, typically using a combination of sweeping and suction devices.

[0080] The localization, mapping, and waste identification of the intelligent cleaning drone include: the cleaning device uses SLAM technology for localization and mapping; a second GNSS module provides positioning for the cleaning device; and a second RGB-D vision module identifies waste information, including waste name, category, and location information. The coordinate transformation of the intelligent cleaning drone involves considering its motion in a plane as rigid body motion, which is then transformed to the world coordinate system through homogeneous coordinate transformation. The intelligent cleaning drone operates within the world coordinate system.

[0081] The positioning of the intelligent cleaning drone is mainly achieved by obtaining its location on a map. However, due to the obstruction caused by surrounding tall buildings, tunnels, and other factors while driving on the road, the positioning of the intelligent cleaning drone may deviate. A MEMS inertial measurement unit and a second GNSS module are used for global positioning. The UKF algorithm is used for data fusion to estimate the state errors of the inertial navigation and GNSS modules. The estimated state errors are then used to correct the various navigation systems through feedback, thereby improving the positioning accuracy.

[0082] The path planning for the intelligent cleaning drone involves a cloud-based management platform dividing a two-dimensional grid map into regions and obtaining the world coordinate system for each drone. Coordinate transformation is then used to determine the coordinate relationships of the garbage within the areas each drone is responsible for. Based on the received garbage coordinate information for their assigned area and considering traffic congestion, each drone plans its path from nearest to farthest. Navigation utilizes autonomous navigation and path planning algorithms for global path planning, calculating the optimal route from the drone to the target location. As a preferred approach, the autonomous navigation and path planning algorithm combines the traditional Dijkstra algorithm with intelligent Particle Swarm Optimization (PSO) algorithms.

[0083] Specifically, such as Figure 3 As shown, the first input data source is primarily map data, including a raster map, start-point coordinates, and end-point coordinates. The Dijkstra algorithm generates an initial path, including initializing a distance array, constructing a set of unprocessed nodes, iteratively selecting the node with the shortest distance from the unprocessed nodes, updating the distances to adjacent nodes and predecessor nodes, and generating the initial shortest path. Next, real-time map data is input, including a raster map and real-time traffic conditions. Finally, the PSO algorithm optimizes the path, including initializing a particle swarm (each example representing a path), defining a fitness function including path length, safety, and smoothness, iteratively updating particle velocity and position, adjusting the path based on individual and swarm optimization, and considering real-time traffic conditions such as congestion and closures, ultimately outputting the optimal path.

[0084] The autonomous navigation and path planning algorithm used in this invention combines the traditional Dijkstra algorithm with the intelligent PSO algorithm, which further improves the efficiency of urban road pollutant cleaning.

[0085] The intelligent unmanned sweeping vehicle cleans up road pollutants. This process primarily involves using a point cloud generated by a second RGB-D vision module upon arrival at the designated location to detect pollutants, calculating the coordinates of the point cloud within the recognition box, transforming the acquired garbage coordinates into the intelligent sweeping vehicle's coordinate system, and then the vehicle begins its movement, activating its sweeping equipment and starting to clean up the garbage. As a preferred solution, the YOLOv12 algorithm is used for object detection, offering a large receptive field, good model stability, and performance.

[0086] This invention employs the YOLOv12 algorithm for road pollutant identification, introducing an attention mechanism into a single-stage target detection framework for the first time. This reduces computational complexity while maintaining a large receptive field, thereby improving identification speed. Simultaneously, it introduces a residual high-efficiency layer aggregation network, employing block-level residual design with scaling techniques and a redesigned feature aggregation method to improve model stability and performance, ensuring both speed and reliability in identification.

[0087] In one embodiment, the present invention provides a method for deploying a road surface pollutant cleaning system, including:

[0088] Step 100: Initialize the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle, and determine the parameters of the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle;

[0089] Step 200: Establish a wireless network connection between the remote sensing satellite, the roadside inspection and sensing terminal, the cloud management and control platform, the distributed edge computing module, and the intelligent cleaning unmanned vehicle;

[0090] Step 300: The remote sensing satellite and the roadside inspection and sensing terminal respectively collect data from the target area. The collected data is processed by the distributed edge computing module, and the processed data is transmitted back to the cloud management and control platform.

[0091] Step 400: The cloud-based management platform establishes a pollutant distribution network based on the returned data and issues task instructions to the intelligent unmanned cleaning vehicle;

[0092] Step 500: The intelligent cleaning unmanned vehicle arrives at each pollutant area in sequence based on the planned route obtained from the task instructions, identifies the road pollutants, starts the cleaning operation device, and moves to the next task target area after completing the cleaning task.

[0093] Step 600: After completing all cleaning tasks, the intelligent cleaning unmanned vehicle returns to the work standby area.

[0094] like Figure 4 As shown, it includes:

[0095] First, initialize the status of the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle, and confirm the parameters of the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle.

[0096] Specifically, the system sequentially confirms that the remote sensing satellite is over the target urban area, confirms that the roadside inspection and sensing terminal is located in the target area, and confirms that the intelligent cleaning unmanned vehicle is located in the work standby area.

[0097] Then, wireless communication is established for remote sensing satellites, roadside inspection and sensing terminals, edge computing and cloud management and control platforms;

[0098] Remote sensing satellites collect data on the target area, including:

[0099] Remote sensing satellites perform SLAM mapping to create two-dimensional grid maps. While mapping, deep learning object detection algorithms are used to identify road pollutants, traffic conditions, and vehicle flow. This information is then transmitted to a distributed edge computing module.

[0100] The roadside inspection and sensing terminal collects data on the target area, including:

[0101] The roadside inspection and sensing terminal performs SLAM mapping to build a two-dimensional grid map. While building the map, a deep learning object detection algorithm is used to identify road pollutants through the first RGB-D vision module and transmit the information to the distributed edge computing module.

[0102] Furthermore, the data source is processed by the distributed edge computing module, including:

[0103] A two-dimensional map and garbage coordinate information are established. The coordinates of the identified road pollutants are converted to the world coordinate system. The traffic conditions and traffic flow are predicted to determine the duration of congestion and smooth traffic flow. The data is then transmitted back to the cloud management platform.

[0104] The cloud-based management platform establishes a pollutant distribution network by transmitting back data and assigns tasks to the intelligent unmanned cleaning vehicles, including:

[0105] The cloud-based management platform processes the transmitted 2D map, garbage coordinates, and traffic prediction information. Based on the map, number of vehicles, and target urban areas, it rationally divides the road sections to be handled by the intelligent cleaning drones. The cloud-based management platform converts the coordinates of each intelligent cleaning drone to the world coordinate system. It then traverses the pollutant coordinates and traffic prediction information of the areas handled by each intelligent cleaning drone, performs global path planning, and transmits the garbage coordinates of the target sub-area to the intelligent cleaning drone responsible for that area.

[0106] The intelligent unmanned cleaning vehicle proceeds along the planned route to each area containing pollutants, including:

[0107] The intelligent cleaning unmanned vehicle receives garbage coordinate information and traffic condition prediction information in the world coordinate system. Based on the garbage coordinate information and road traffic congestion, it sets waypoints from near to far. It performs global path planning based on the grid map and finds the optimal path through the autonomous navigation and path planning technology algorithm of the intelligent cleaning unmanned vehicle. Based on the optimal path and waypoints, the intelligent cleaning unmanned vehicle arrives near the target area and identifies road pollutants.

[0108] The intelligent cleaning unmanned vehicle starts its cleaning operation device, identifies the target, processes the point cloud information within the recognition frame, obtains the coordinate information of the garbage, calculates the relative coordinates between the garbage and the cleaning operation device through coordinate transformation, and plans the intelligent cleaning unmanned vehicle's travel route using the robot operating system motion framework (Moveit in this embodiment). The cleaning operation device reaches the garbage location and performs garbage cleaning.

[0109] Once completed, the intelligent cleaning vehicle moves to the next target area. After all work is finished, the intelligent cleaning vehicle returns to the work standby area.

[0110] In one embodiment, the autonomous navigation and path planning algorithm of this invention adopts a combination of the traditional Dijkstra algorithm and the intelligent PSO algorithm.

[0111] Specifically, the PSO iteration and the Dijkstra process are not separate, i.e., not PSO followed by Dijkstra or vice versa, nor are they completely parallel, but rather deeply intertwined and iteratively coordinated. The optimization and improvement schemes include a particle encoding method using key node sequences and attention marker vectors, a particle fitness evaluation mechanism based on local relaxation improvement amounts, and a priority queue scheduling function that integrates global attention.

[0112] The particle encoding method for key node sequences and attention marker vectors separates path structure information from regional value judgments, achieving separate storage of the solution space and guidance information. (Particle position) The expression is as follows:

[0113]

[0114] in, It is an ordered sequence of key nodes. For the number of nodes, , is a vector indicating the level of attention in the peripheral region. The number of vectors, This is the particle position index label.

[0115] The particle fitness evaluation mechanism based on local relaxation improvement avoids ineffective full-path computation, quantifies particle guidance value, and constructs a particle fitness function. as follows:

[0116]

[0117] in, Based on The initial path length estimate, Is The path improvement after performing local Dijkstra relaxation on the marked region. It is the penalty factor for the time consumed in the simulation calculation.

[0118] The priority queue scheduling function, which integrates global attention, breaks through the traditional distance-first principle, incorporates swarm intelligence prediction into the search order, and modifies Dijkstra's priority queue priority function. as follows:

[0119]

[0120] Where dist(v) is the minimum cost of the current node v. It is the edge The global attention level is aggregated by particle swarm marker frequencies. It is the attention adjustment coefficient, which usually ranges from 0 to 1.

[0121] In one embodiment, the present invention constructs a three-level linkage architecture of "space-ground-device".

[0122] Specifically, such as Figure 5 As shown, it consists of three parts: the base layer, the ground layer, and the terminal layer.

[0123] At the ground level, a low-orbit satellite constellation composed of remote sensing satellites monitors and acquires ground road image monitoring data, which is transmitted to the onboard computing platform to obtain urban road congestion and pollutant heat maps. The AI ​​decision-making module then converts these data into mission event commands, which are transmitted to the ground level via the communication downlink.

[0124] In the ground layer, the distributed edge computing module decomposes the task, the MEC computing chip performs local path planning, and the wireless communication private network slice is responsible for sending control commands to the cloud management platform to obtain accurate positioning data and environmental perception sharing information, which is then transmitted to the terminal layer.

[0125] At the terminal layer, the intelligent cleaning unmanned vehicle cluster provides operational status feedback based on precise positioning data. The roadside inspection and perception terminal constructs road surface scenes based on shared environmental perception information, and feeds the data back to the distributed edge computing module for data integration and updating. The data is then sent to the cloud management platform, which generates strategy optimization instructions and returns them to the low-orbit satellite constellation at the Tianji level for iterative optimization.

[0126] Therefore, the "sky-ground-end" three-level linkage architecture adopted in this invention realizes intelligent perception and precise cleaning of urban roads through multi-dimensional interaction between intelligent edge nodes and autonomous mobile platforms. This can make up for the problems of insufficient field of vision and insufficient response time for abnormal working conditions of intelligent sweeping unmanned vehicles, and significantly improve the efficiency from finding garbage to cleaning garbage.

[0127] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0128] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A road surface pollutant cleaning system, characterized in that, include: The system consists of remote sensing satellites, roadside inspection and sensing terminals, distributed edge computing modules, cloud management and control platforms, and intelligent cleaning unmanned vehicles. The cloud management and control platform is the main control unit, the distributed edge computing modules are edge units, and the remote sensing satellites, inspection and sensing terminals, and intelligent cleaning unmanned vehicles are slave units. The main control unit, edge units, and slave units are connected via a wireless network. The remote sensing satellite and the roadside inspection and sensing terminal work together to perform synchronous positioning and mapping of the target area; The remote sensing satellite transmits road pollutants and traffic conditions, including vehicle flow, to the distributed edge computing module, and the road inspection and sensing terminal transmits road pollutants to the distributed edge computing module. The distributed edge computing module processes the collected data and sends the processed data back to the cloud management platform, including: Establish a two-dimensional map and pollutant coordinate information; transform the identified road pollutant coordinates to the world coordinate system; predict congestion duration and traffic flow duration based on the identified traffic conditions including vehicle flow; and transmit the prediction information back to the cloud-based management platform. The cloud-based management platform transmits the coordinate information of road pollutants and traffic condition prediction information of the target area to each intelligent unmanned sweeping vehicle. The intelligent unmanned sweeping vehicles perform global planning for their respective road areas, reach the target pollutant points based on the planned paths, and clean the road surface pollutants using sweeping devices, including: Based on pollutant coordinates and road traffic congestion, waypoints are set from near to far. Global path planning is performed based on the grid map. The optimal path is determined by combining Dijkstra's algorithm and the PSO algorithm. A particle encoding method using key node sequences and attention marker vectors, a particle fitness evaluation mechanism based on local relaxation improvements, and a priority queue scheduling function incorporating global attention are employed. This includes constructing the particle fitness function as follows: ,in, Based on ordered key node sequence The initial path length estimate, Attention marker vectors in the edge region The path improvement after performing local Dijkstra relaxation on the marked region. This is a penalty factor for the time consumption of simulation calculations; the Dijkstra priority queue priority function is modified as follows: Where dist(v) is the minimum cost of the current node v. It is the edge The global attention level is aggregated by particle swarm marker frequencies. It is a attention adjustment coefficient; Based on the optimal route, the train travels along the waypoints to the areas where each pollutant is located.

2. The road pollutant cleaning system according to claim 1, characterized in that, The remote sensing satellite includes an onboard computing platform, LiDAR, optical camera, synthetic aperture radar, inertial measurement unit, star sensor, GNSS module, data transceiver module and power supply module; The onboard computing platform receives information and control commands from the lidar, optical camera, synthetic aperture radar, inertial measurement unit, and GNSS module. The lidar acquires three-dimensional point cloud data, the optical camera extracts and matches features from urban road images, the synthetic aperture radar acquires all-weather radar images, the inertial measurement unit measures acceleration and angular velocity, the GNSS module provides absolute position information, and the communication module is used for wireless communication between the remote sensing satellite and the outside world. The remote sensing satellite uses multi-source sensor fusion technology for global positioning and environmental map construction. It uses SLAM algorithm to process onboard sensor data and establishes high-precision three-dimensional point cloud maps or digital elevation models of the Earth's surface. During the map construction process, it combines satellite attitude data and orbital information to achieve accurate geographic positioning and terrain modeling on a global scale. The surface images are acquired by optical cameras or synthetic aperture radar. Deep learning image classification and target detection algorithms deployed on cloud servers are used to automatically identify pollutant accumulation areas in the surface images. The coordinate information of the identified pollutant areas is combined with satellite positioning data and converted to the world coordinate system. The remote sensing satellite uses satellite video or radar remote sensing data, combined with optical flow analysis algorithms or vehicle trajectory tracking algorithms, to achieve real-time monitoring of road traffic flow. By extracting vehicle motion vectors and calculating road cross-section traffic flow density, a dynamic traffic flow map is generated, and the dynamic traffic flow map is used for global path optimization.

3. The road pollutant cleaning system according to claim 1, characterized in that, The roadside inspection and sensing terminal includes a first end-side computing unit, a first RGB-D vision module, a first solid-state LiDAR, a first GNSS module, a communication module, and a power module; The first end-side computing unit receives information and control commands from the first RGB-D vision module, the first solid-state LiDAR and the first GNSS module. The first RGB-D vision module acquires road pollutant information, the first solid-state LiDAR acquires surrounding environmental information, and SLAM technology is used for positioning and mapping. The first GNSS module realizes the positioning of the roadside inspection and sensing terminal, and the communication module realizes wireless communication between the roadside inspection and sensing terminal and the outside world. The positioning, mapping, and pollutant identification of the roadside inspection and sensing terminal include the roadside inspection and sensing terminal using SLAM technology to build a two-dimensional grid map of the target road segment, using a deep learning target detection algorithm to identify pollutants on the road surface, calculating the XYZ coordinates of the point cloud information within the identification box, converting the XYZ coordinates to the world coordinate system, and sending the converted coordinates to the intelligent cleaning unmanned vehicle through a cloud server.

4. The road pollutant cleaning system according to claim 1, characterized in that, The distributed edge computing module includes an MEC computing chip and a private communication network slice; The MEC computing chip receives remote sensing, video, and image data sent by remote sensing satellites and roadside inspection and sensing terminals, analyzes and identifies traffic communication conditions in real time, and identifies the name, category, and location information of road pollutants. The dedicated communication network slice provides independent bandwidth to enable wireless communication between the distributed edge computing module and the outside world.

5. The road pollutant cleaning system according to claim 1, characterized in that, The cloud-based management platform is responsible for assigning tasks to the intelligent cleaning unmanned vehicle, monitoring the status information of the intelligent cleaning unmanned vehicle, adjusting control commands, and realizing the interaction between the intelligent cleaning unmanned vehicle and environmental information.

6. The road pollutant cleaning system according to claim 1, characterized in that, The intelligent cleaning unmanned vehicle is equipped with a MEMS inertial measurement unit, a second GNSS module, a second end-side computing platform, a second RGB-D vision module, a second solid-state LiDAR, a cleaning operation device, a pollutant storage device, a drive device, and a communication module. The second solid-state LiDAR acquires information about the surrounding environment and uses SLAM technology for positioning and navigation. The second GNSS module provides positioning for the intelligent cleaning unmanned vehicle. The second RGB-D vision module acquires information about road pollutants. The MEMS inertial measurement unit and the second GNSS module use the UKF filtering algorithm to fuse time, position and velocity information. After feedback correction, the time, position and velocity are output. The MEMS inertial measurement unit calculates the velocity acceleration, angular velocity, and yaw angle through the inertial system to obtain time, position, and velocity; the second GNSS module calculates time, position, and velocity through the navigation system. The intelligent cleaning unmanned vehicle's positioning, mapping, and pollutant identification include the cleaning device using SLAM technology for positioning and mapping, a second GNSS module providing positioning for the cleaning device, and a second RGB-D vision module identifying pollutant information, including pollutant name, category, and location information. The coordinate transformation of the intelligent cleaning unmanned vehicle includes treating the motion of the intelligent cleaning unmanned vehicle in the plane as rigid body motion and transforming it to the world coordinate system through homogeneous coordinate transformation; The positioning of the intelligent cleaning unmanned vehicle includes obtaining the position of the intelligent cleaning unmanned vehicle on the map, using a MEMS inertial measurement unit and a second GNSS module for global positioning, using the UKF algorithm for data fusion, estimating the state error of the inertial navigation and the second GNSS module, and correcting each navigation system through the estimation feedback of the state error to improve positioning accuracy. The path planning of the intelligent cleaning unmanned vehicle includes dividing the two-dimensional grid map into regions by the cloud management platform, obtaining the world coordinate system of each intelligent cleaning unmanned vehicle, obtaining the pollutant coordinate relationship of the area under the responsibility of the intelligent cleaning unmanned vehicle through coordinate transformation, and planning the path from near to far according to the pollutant coordinate information of the area under its responsibility and the road traffic congestion situation. In the navigation, autonomous navigation and path planning algorithms are used to perform global path planning and calculate the optimal route from the intelligent cleaning unmanned vehicle to the target location. The intelligent cleaning unmanned vehicle cleans up road pollutants by using a point cloud generated by a second RGB-D vision module after reaching a designated location, detecting the pollutants, calculating the coordinates of the point cloud within the recognition box, converting the pollutant coordinates to the intelligent cleaning unmanned vehicle coordinate system, and then starting the intelligent cleaning unmanned vehicle to move forward and start the cleaning operation device to clean up the pollutants.

7. A method for deploying a road surface contaminant cleaning system, based on the road surface contaminant cleaning system according to any one of claims 1 to 6, characterized in that, include: Initialize the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle, and determine the parameters of the remote sensing satellite, the roadside inspection and sensing terminal, and the intelligent cleaning unmanned vehicle. Establish a wireless network connection between the remote sensing satellite, the roadside inspection and sensing terminal, the cloud management and control platform, the distributed edge computing module, and the intelligent cleaning unmanned vehicle; The remote sensing satellite and the roadside inspection and sensing terminal respectively collect data from the target area. The collected data is processed by the distributed edge computing module and then transmitted back to the cloud management and control platform. The cloud-based management platform establishes a pollutant distribution network based on the returned data and issues task instructions to the intelligent unmanned cleaning vehicle. The intelligent cleaning unmanned vehicle, based on the planned route obtained by autonomous navigation and path planning algorithms, arrives at the areas where each pollutant is located in turn, identifies the road pollutants, starts the cleaning operation device, and moves to the next task target area after completing the cleaning task. After completing all cleaning tasks, the intelligent cleaning unmanned vehicle returns to the work standby area.

8. The deployment method of the road pollutant cleaning system according to claim 7, characterized in that, The remote sensing satellite and the roadside inspection and sensing terminal respectively collect data on the target area, including: Start SLAM mapping and create a 2D raster map; Using a deep learning target detection algorithm, the remote sensing satellite identifies road pollutants and traffic conditions including vehicle flow, while the roadside inspection and sensing terminal identifies road pollutants through a first RGB-D vision module.

9. The deployment method of the road pollutant sweeping system according to claim 7, characterized in that, The cloud-based management platform establishes a pollutant distribution network based on the returned data and issues task instructions to the intelligent unmanned cleaning vehicle, including: The returned two-dimensional map, pollutant coordinate information, and traffic condition prediction information are processed. The responsible road sections for intelligent cleaning unmanned vehicles are determined by dividing the area based on a two-dimensional map, the number of vehicles, and the target urban area. Transform the coordinates of each intelligent unmanned cleaning vehicle to the world coordinate system; The system iterates through the pollutant coordinates and traffic condition predictions of the areas covered by each intelligent cleaning drone to perform global path planning. The coordinates of pollutants in the target sub-region are transmitted to the intelligent cleaning unmanned vehicle responsible for the target sub-region.

10. The deployment method of the road pollutant cleaning system according to claim 7, characterized in that, The intelligent unmanned cleaning vehicle, based on a planned route obtained through autonomous navigation and path planning algorithms, sequentially reaches each area containing pollutants, including: Identify road pollutants and activate the sweeping operation equipment, including: Based on the point cloud information within the target bounding box, the coordinate information of the pollutants is obtained, and the relative coordinates between the pollutants and the cleaning device are obtained through coordinate transformation. Based on relative coordinates, the robot operating system motion framework is used to plan the travel route of the intelligent cleaning unmanned vehicle. The cleaning device reaches the location of the pollutants according to the travel route and cleans the pollutants.

Images