A road surface operation automatic identification control system for a sanitation vehicle

Abstract

This invention provides an automatic identification and control system for sanitation vehicles operating on roads, relating to the fields of autonomous driving and control strategies. The system includes modules for data acquisition, processing, feature extraction, model building, stain identification, decision control, and execution. The core of the system lies in the decision control module, which constructs a conflict coordination mechanism by introducing a dynamic objective weight matrix. This mechanism balances operational efficiency, energy consumption control, and coverage in real time based on time urgency, energy constraints, and pollution levels. The system divides the work area into grid cells, utilizes an improved ant colony algorithm based on three-dimensional objective comprehensive benefit values ​​to plan the optimal path, and combines a cost matrix and task load variance to achieve multi-vehicle collaborative task allocation and balanced optimization. The execution module dynamically adjusts the parameters of the operating devices based on the degree of stain identification. This invention effectively coordinates the multi-objective conflict issues in sanitation operations, significantly reducing system energy consumption while ensuring operational coverage and improving the intelligence level of municipal cleaning.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving and control strategy technology, and in particular to an automatic identification and control system for sanitation vehicles operating on the road. Background Technology

[0002] With the acceleration of urbanization, the demand for municipal sanitation operations is increasing daily, placing higher demands on the intelligence and operational efficiency of medium-sized sanitation sweepers. Against the backdrop of rapid development in autonomous driving and new energy commercial vehicle technologies, sanitation operations are gradually shifting from traditional manual driving and control to intelligent automated operations. To improve the quality of sanitation operations and reduce operating costs, modern sweepers need to possess precise environmental perception and autonomous decision-making capabilities, automatically adjusting sweeping device parameters and planning the optimal path based on real-time road pollution levels, obstacle distribution, and the vehicle's own condition.

[0003] In existing technologies, images are primarily acquired using cameras, and stereo matching algorithms are employed to identify the distribution of road debris, thereby controlling the sweeping speed. However, existing solutions suffer from several significant drawbacks in practical applications: First, current recognition systems rely excessively on a single camera, making them susceptible to interference in complex lighting conditions, inclement weather, or uneven road surfaces. Second, they fail to comprehensively collect key environmental data such as road surface smoothness, temperature, humidity, and multi-dimensional obstacles, resulting in poor recognition accuracy and system robustness. Third, existing operation control logic often presets fixed parameters, failing to dynamically adjust operational strategies based on real-time environmental data. This often leads to conflicts between operational efficiency, energy consumption control, and coverage in actual operations. Finally, existing systems' path planning largely considers only travel distance, neglecting factors such as road surface contamination levels, deduction of repetitive work areas, and coverage rate in the gridded decision-making process.

[0004] Therefore, there is an urgent need for a system that can intelligently balance and coordinate the three dynamically conflicting goals of operation efficiency, energy consumption, and coverage based on multi-dimensional sensing data, so as to achieve true precision and intelligence in cleaning operations. Summary of the Invention

[0005] The main objective of this invention is to provide an automatic identification and control system for sanitation vehicles operating on the road, which solves the technical problem of the lack of dynamic decision-making mechanisms in the prior art. It dynamically adjusts the parameters of the operating device based on environmental data to accurately match the cleaning needs of different stains.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is to provide an automatic identification and control system for sanitation vehicles on the road, including a data acquisition module, a data processing module, a feature extraction module, a model building module, a stain identification module, a decision control module, and an execution module;

[0007] The decision control module includes an operation planning unit. Based on the road operation environment model and the stain information output by the stain recognition module, the operation planning unit introduces a dynamic target weight matrix to build a conflict coordination mechanism. By balancing operation efficiency, energy consumption control and coverage through dynamic target weights, and combining constraints such as driving speed, the operation planning unit plans the operation path and sequence.

[0008] Preferably, the decision control module includes a job planning unit, which has a built-in dynamic target weight matrix construction module. This module is used to construct a dynamic target weight matrix with three-dimensional targets: job time, energy consumption, and coverage rate. The module dynamically adjusts the weight coefficients of each target based on the characteristic parameters of the current job scenario to coordinate the priorities between job efficiency, energy consumption control, and cleaning coverage rate under different scenarios.

[0009] The task planning unit also includes a pheromone update and path planning module, which divides the task area into grids, uses the ant colony algorithm for path planning, updates the pheromone concentration based on the three-dimensional target comprehensive benefit value, and strengthens the optimal path through global pheromone updates.

[0010] The operation planning unit also includes a regional segmentation and task allocation module, which is used to divide the operation area into different priority areas according to the degree of pollution, construct a cost matrix for task allocation, and realize multi-vehicle collaborative scheduling by solving the minimum cost matching.

[0011] The region segmentation and task allocation module is also used to calculate the task load variance of each vehicle and to reallocate tasks when the load variance exceeds a preset threshold.

[0012] Preferably, the decision control module also includes a control strategy generation unit, used to generate control commands based on the operation planning results and the sanitation vehicle's own status data; Control commands include instructions to adjust the parameters of the working device based on the type and extent of the stain, as well as driving system commands generated based on obstacles or path planning results.

[0013] Preferably, the data acquisition module includes cameras installed at the front, sides and rear of the vehicle, lidar installed at the front and rear of the vehicle, humidity and temperature sensors installed at the bottom of the vehicle, vehicle speed sensors installed on the wheels, gyroscopes installed inside the vehicle, and ultrasonic sensors and millimeter-wave radar installed around the vehicle.

[0014] Preferably, the data processing module includes a data preprocessing unit, which is used to remove salt-and-pepper noise from camera image data using a median filtering algorithm, calibrate LiDAR data, and add a unified timestamp to all sensor data and perform interpolation alignment to achieve spatiotemporal synchronization of multi-source data.

[0015] Preferably, the road operation environment model established by the model building module includes a road condition model for describing smoothness, roughness and stain distribution, an obstacle distribution model for displaying the location and distribution pattern of surrounding obstacles, and a sanitation operation area model for determining the operation scope and area.

[0016] Preferably, the execution module includes a working device and a driving system; The working device includes a sweeping brush, a vacuum fan, and a water spraying device. The sweeping brush speed, vacuum fan power, and water spraying flow rate are adjusted according to control commands. The driving system adjusts the sanitation vehicle's speed, direction, and braking according to control commands.

[0017] Beneficial effects: 1. Achieve dynamic and precise balance and multi-dimensional decision optimization of work objectives. By introducing a three-dimensional dynamic target weight matrix and conflict coordination mechanism, it can automatically trigger weight adjustment according to different real-time scenarios such as morning peak, insufficient power or deep pollution.

[0018] 2. Significantly improves the scientific rigor and coverage effectiveness of path planning. Compared to traditional full-coverage path search, this invention employs an improved ant colony algorithm combined with 5m×5m grid cells for planning, and utilizes a three-dimensional target comprehensive benefit value. The guidance pheromone update enables path planning to consider not only distance, but also energy consumption cost and pollution priority.

[0019] 3. To improve the robustness and accuracy of multi-source data fusion recognition, the system utilizes multi-dimensional sensing methods such as cameras, LiDAR, and humidity / temperature sensors to construct a three-dimensional environmental model. By applying median filtering to images and performing spatiotemporal alignment calibration on radar data, the system effectively solves the interference of environmental noise such as changes in lighting, shadows, and road surface reflections on stain recognition. It can not only identify stain types but also dynamically adjust the power of the vacuum fan and the speed of the sweeping disc based on factors such as road surface roughness and humidity, achieving closed-loop control of operational intensity tailored to local conditions.

[0020] 4. To optimize load fairness and system endurance in multi-vehicle collaborative operations, a task allocation algorithm based on the "maximum-minimum fairness" principle is introduced into the system for multi-vehicle operation scenarios. This is achieved through a cost matrix. Solving for the minimum cost matching problem prevents battery over-discharge caused by excessive operation of a single vehicle, and achieves this by monitoring the task load variance in real time. Once a load imbalance is detected, task redistribution is triggered, which greatly improves the turnover efficiency and overall service life of the entire sanitation vehicle fleet. Attached Figure Description

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1This is a structural diagram of an automatic identification system for sanitation road operations according to the present invention; Figure 2 This is a flowchart of the system control strategy execution of the present invention. Detailed Implementation

[0022] Example 1 like Figure 1 As shown, an automatic identification and control system for sanitation vehicles operating on the road includes the following modules: data acquisition module, data processing module, feature extraction module, model building module, stain recognition module, decision control module, and execution module.

[0023] In embodiments of the present invention, the data acquisition module is used to collect relevant data on the road working environment, including road condition data, sanitation vehicle status data, and surrounding environment data: Multiple high-definition industrial cameras (1920×1080 resolution, 30fps) are installed on the front, sides, and rear of the sanitation vehicle. Different focal length lenses (100° wide-angle, 50° standard, and 30° telephoto) are used to achieve 360° coverage without blind spots. The cameras are treated with anti-glare coating and equipped with infrared fill lights to ensure that road images and video information from different angles are collected under low light conditions such as night, rain, and fog, and to monitor the road surface stains, traffic signs, and markings in real time. LiDAR is installed at the front and rear of the vehicle body in a vertical mounting manner to form a two-way scanning pattern, ensuring accurate detection of the position of obstacles in front and behind. The LiDAR data is processed by point cloud algorithms to generate a high-precision three-dimensional road surface model, which is used to identify road surface unevenness, potholes and obstacle height. A humidity sensor (measurement range 0-100%RH, accuracy ±2%RH) and a temperature sensor (measurement range -40~85℃, accuracy ±0.5℃) are installed on the bottom of the vehicle body, evenly distributed at the front, middle and rear positions of the vehicle body, to monitor the road surface humidity and temperature in real time, providing key information for judging the difficulty of road surface cleaning; Vehicle speed sensors are installed on the wheels to obtain the driving speed. Non-contact Hall sensors are used to obtain the rotation speed of each wheel in real time to calculate the actual driving speed of the vehicle and determine whether slippage has occurred. A 3-axis gyroscope is installed inside the vehicle body to detect changes in vehicle acceleration, tilt angle and direction, providing motion state reference for path planning. Multiple ultrasonic sensors (detection range 0-5m, accuracy ±1cm) and millimeter-wave radar (detection range 0-150m, accuracy ±0.1m) are evenly distributed around the vehicle body to monitor surrounding vehicles, pedestrians, and static obstacles. The millimeter-wave radar uses FMCW (Frequency Modulated Continuous Wave) technology, which can penetrate adverse weather conditions such as rain and fog, ensuring reliable perception in complex environments.

[0024] In embodiments of the present invention, the data processing module processes and analyzes the data acquired by the data acquisition module, including a data preprocessing unit that performs filtering, noise reduction, and calibration preprocessing operations on the raw data. The data preprocessing unit uses a median filtering algorithm to remove salt-and-pepper noise from the camera image data. It calibrates the LiDAR data to eliminate LiDAR installation errors, then uses the RANSAC algorithm to remove outliers from the ground point cloud, and finally uses a point cloud clustering algorithm to separate the obstacle point cloud from the ground point cloud, generating accurate obstacle contours. Kalman filtering is applied to data such as humidity, temperature, and vehicle speed to eliminate random noise and ensure data stability. All the above data are stamped with a unified timestamp and interpolated and aligned based on the timestamp to ensure the spatiotemporal synchronization of multi-source data. Interpolation and alignment are performed on sensor data of different frequencies using a cubic spline interpolation algorithm to ensure the continuity of time-series data.

[0025] The feature extraction module extracts features from the preprocessed data, including road surface smoothness, roughness, stain type and severity, and obstacle shape and location. Based on LiDAR point cloud data, it calculates the standard deviation of road surface elevation changes to generate smoothness feature values. Using an image texture analysis algorithm based on the gray-level co-occurrence matrix, it extracts road surface texture features and calculates the roughness index, thereby increasing the sweeper brush speed. Combining image features and LiDAR data, stains are classified into six categories: oil, paper scraps, fallen leaves, and garbage. Each category is further categorized into three levels—light, moderate, and heavy—based on coverage area and thickness, with stain severity quantified by stain coverage rate. The module extracts the size, shape, height, and location of obstacles from the LiDAR point cloud data to generate obstacle feature vectors. For dynamic obstacles, the system also extracts their direction of motion and velocity features. The output of the feature extraction module is transmitted to both the model building module and the stain recognition module, which process the data in parallel.

[0026] The model building module, based on the feature information output by the feature extraction module, establishes a comprehensive and accurate road operation environment model, providing an environmental foundation for decision-making and control. The specific model construction is as follows: The road surface condition model describes smoothness, roughness, and stain distribution. The smoothness model establishes a smoothness distribution heat map based on the road surface elevation map generated by LiDAR point cloud. The roughness model establishes a roughness distribution model based on image texture analysis results, reflecting the road surface material and wear degree. The stain distribution model establishes a distribution map of stain type and degree based on stain identification results, accurate to 1m×1m grid unit. The obstacle distribution model, based on LiDAR and millimeter-wave radar data, establishes a three-dimensional obstacle distribution model, including the real-time location, size, and movement trajectory of static obstacles (streetlights, trash cans, roadblocks) and dynamic obstacles (pedestrians, vehicles), providing obstacle avoidance basis for path planning; The sanitation operation area model, based on GPS positioning and GIS map data, establishes a boundary model of the operation area, dividing the operation area into 5m×5m grid units. Each grid unit stores characteristic information such as its flatness, roughness, stain type and degree, providing a basis for route planning and task allocation.

[0027] The stain recognition module has a built-in trained stain recognition model. Road surface images captured by the camera are input into this model, which outputs the type and extent of road stains, providing a basis for determining subsequent work methods and intensity. Through multi-source sensor fusion, refined feature extraction, and accurate model building, the system can comprehensively perceive the working environment, providing a reliable basis for intelligent decision-making and effectively solving problems such as insufficient environmental perception, low recognition accuracy, and fixed work strategies in existing technologies.

[0028] Example 2 like Figure 2 As shown, the decision control module generates operation control instructions based on the road operation environment model established by the model building module, the stain information output by the stain recognition module, and the preset operation rules. The decision control module includes an operation planning unit and a control strategy generation unit.

[0029] The operation planning unit is based on the road operation environment model and preset operation objectives. It introduces a dynamic objective weight matrix to construct a conflict coordination mechanism. The dynamic objective weights balance operation efficiency, energy consumption control, and coverage, and plan the operation path and sequence in combination with constraints such as driving speed. The operation time is the weighted sum of driving time and operation duration. The road condition and operation difficulty correction factors are introduced as shown in the following formula (1): (1) in, , These are the weighting coefficients, default. , In severe weather Then adjust it appropriately. The response will be adjusted upwards. The estimated time from the current position to the end of the operation is calculated using the following formula (2): (2) in, The straight-line distance between two points. For real-time driving speed, the default speed is 20 km / h on main roads and 12 km / h on alleys. Congestion correction factor (smooth traffic = 1.0, slow traffic = 1.5, congestion = 2.0).

[0030] The actual cleaning time for the work area is calculated using the following formula (3): (3) in, The sub-area operating area (㎡) The sweeping width for vehicles is (m, 2.5m for small vehicles and 4.0m for large vehicles). Operating speed (km / h, default 5km / h). The road surface difficulty coefficient is calculated as follows: cement road = 1.0, asphalt road = 1.2, muddy road = 1.8.

[0031] An energy consumption model is established based on vehicle mileage, load, and road conditions, as shown in equation (4): (4) in, The mileage energy consumption coefficient (kWh / km) is 0.85 by default and increases to 1.5 when fully loaded. The load energy consumption coefficient (kWh / ton) is 0.3 by default, and increases to 0.45 when the full load of clean water / garbage is >80%. The drag energy consumption coefficient (kWh / (km·N)) is 0.02 by default. Road surface resistance (N, slope) hour, , (Total vehicle mass). For example, the environmental correction factor is: 25℃ = 1.0, <0℃ = 0.9, >35℃ = 1.1, and rainy day = 1.2. The battery degradation coefficient is 1.0 for a new battery and 0.85 for a remaining capacity of less than 70%.

[0032] The coverage rate is calculated as the ratio of the actual working area to the total area of ​​the area to be worked, and the overlapping working areas need to be deducted, as shown in the following formula (5): (5) in, The actual cleaning area (㎡) is calculated based on the cleaning width. The area (㎡) of overlapping operations by multiple vehicles was obtained through GIS spatial overlay analysis. This represents the total area (㎡) of the sub-region. This is the quality coefficient.

[0033] The dynamic target weight matrix construction module constructs a target weight matrix that is a three-dimensional vector. ,satisfy ,in Assigning weights based on task time. As energy consumption weight, This is the coverage weight.

[0034] Scene parameters include time urgency Energy consumption constraints and pollution level Each weighting coefficient is calculated according to the following formula (6): (6) in, Values ​​range from 0 to 1, with higher values ​​indicating higher priority. For example, in a main road scenario before the morning rush hour: there is a high degree of time urgency. Energy consumption constraints Pollution level ,but , , .

[0035] The pheromone update and path planning module uses an improved ant colony algorithm for grid path planning. The working area is divided into 5m×5m grids, and each grid serves as a pheromone update unit. The pheromone update is divided into local update and global update. The local update formula for each iteration is shown in equation (7) below: (7) After finding the optimal path, the global update formula for reinforcing the optimal path is shown in equation (8) below: (8) in, Let pheromone concentration be the pheromone concentration from grid i to grid j at time t, initially... , The pheromone evaporation coefficient is set to a default value of 0.1, and 0.05 for complex regions. The default value for the pheromone constant is 100. The comprehensive benefit value of the three-dimensional target is shown in the following formula (9): (9) A global update is performed every 20 iterations or when a path that is at least 5% better than the current optimal solution is found. Factors are used to select the grid generation operation path on the grid map based on the objective function conditions and constraints such as driving speed.

[0036] The area segmentation and task allocation module divides the work area into several sub-regions. Based on the current location, remaining battery power, and operational capacity of each sanitation vehicle, sub-regions are allocated using the "maximum-minimum fairness" principle. These sub-regions are defined using LiDAR and camera data, including: "Must-scan areas" (contamination level) Areas such as oil stains and accumulated garbage are marked as "high priority grids"; "Optional Area" (Pollution Level) ), marked as "medium priority grid"; "Non-mandatory areas" (contamination level) Networks marked as "low priority" can be temporarily skipped to ensure full coverage of core areas.

[0037] At the same time, construct a cost matrix , indicating vehicle Assigned to sub-region The cost is shown in equation (10): (10) in, For vehicles Assigned to sub-region distance, For the maximum distance, To estimate energy consumption, For maximum energy consumption, To estimate coverage, we solve for the minimum cost match to ensure the total cost is closest.

[0038] Calculate the variance of task load for each vehicle As shown in equation (11): (11) in, For average load, if If so, the task will be reassigned.

[0039] The control strategy generation unit generates control commands based on the operation planning results and the sanitation vehicle's own status data. It adjusts the parameters of the operating devices according to the type and severity of the stains, increasing the sweeping brush speed and suction fan power for heavy oil stains. It also generates driving system commands, including deceleration, avoidance, and direction adjustment, based on obstacles or path planning results.

[0040] During the control delivery process, the control strategy unit receives the job planning results (path + target parameters), compares them with real-time data from onboard sensors (soil level, obstacle distance), calls matching commands from the preset command library, and sends them to the vehicle ECU via the 4G / 5G network. After the vehicle executes the command, it provides feedback on the execution status (such as speed and power). The platform verifies whether the requirements are met; if not, adjustments are made.

[0041] The execution module receives the operation control instructions generated by the decision control module and controls the sanitation vehicle's operating devices and driving system to perform the corresponding operation actions.

[0042] The operating equipment includes sweeping brushes, a vacuum fan, and a water spraying device. Operating parameters such as the sweeping brush speed, vacuum fan power, and water flow rate are adjusted according to control commands. The driving system adjusts the sanitation vehicle's speed, direction, and braking according to control commands, achieving automatic driving and operational control of the sanitation vehicle.

[0043] For example, the automatic driving command is: speed 12 km / h, drive straight along the edge of section #5. The operation control command is: sweeping disc high speed, vacuum fan full power, sweeping disc water spraying.

[0044] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A road surface operation automatic identification control system for a sanitation vehicle, characterized by, It includes a data acquisition module, a data processing module, a feature extraction module, a model building module, a stain recognition module, a decision control module, and an execution module; The decision control module includes an operation planning unit. Based on the road operation environment model and the stain information output by the stain recognition module, the operation planning unit introduces a dynamic target weight matrix to build a conflict coordination mechanism. By balancing operation efficiency, energy consumption control and coverage through dynamic target weights, and combining constraints such as driving speed, the operation planning unit plans the operation path and sequence.

2. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The decision control module includes a job planning unit, which has a built-in dynamic target weight matrix construction module. This module is used to construct a dynamic target weight matrix with three-dimensional targets: job time, energy consumption, and coverage rate. It dynamically adjusts the weight coefficients of each target based on the characteristic parameters of the current job scenario to coordinate the priorities between job efficiency, energy consumption control, and cleaning coverage rate under different scenarios.

3. The automatic identification and control system for sanitation vehicle road operations according to claim 2, characterized in that: The task planning unit also includes a pheromone update and path planning module, which divides the task area into grids, uses the ant colony algorithm for path planning, updates the pheromone concentration based on the three-dimensional target comprehensive benefit value, and strengthens the optimal path through global pheromone updates.

4. The automatic identification and control system for road operation of a sanitation vehicle according to claim 2, characterized in that: The operation planning unit also includes a regional segmentation and task allocation module, which is used to divide the operation area into different priority areas according to the degree of pollution, construct a cost matrix for task allocation, and realize multi-vehicle collaborative scheduling by solving the minimum cost matching.

5. The automatic identification and control system for road operation of a sanitation vehicle according to claim 4, characterized in that: The region segmentation and task allocation module is also used to calculate the task load variance of each vehicle and to reallocate tasks when the load variance exceeds a preset threshold.

6. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The decision control module also includes a control strategy generation unit, which generates control commands based on the operation planning results and the sanitation vehicle's own status data; Control commands include instructions to adjust the parameters of the working device based on the type and extent of the stain, as well as driving system commands generated based on obstacles or path planning results.

7. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The data acquisition module includes cameras installed at the front, sides, and rear of the vehicle; lidar installed at the front and rear of the vehicle; humidity and temperature sensors installed at the bottom of the vehicle; speed sensors installed on the wheels; gyroscopes installed inside the vehicle; and ultrasonic sensors and millimeter-wave radar installed around the vehicle.

8. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The data processing module includes a data preprocessing unit, which uses a median filtering algorithm to remove salt-and-pepper noise from camera image data, calibrates LiDAR data, adds a unified timestamp to all sensor data and performs interpolation alignment to achieve spatiotemporal synchronization of multi-source data.

9. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The model building module creates road operation environment models including a road condition model to describe smoothness, roughness, and stain distribution; an obstacle distribution model to display the location and distribution patterns of surrounding obstacles; and a sanitation operation area model to determine the operation scope and area.

10. The automatic identification and control system for road operation of a sanitation vehicle according to claim 1, characterized in that: The execution module includes the working device and the driving system; The working device includes a sweeping brush, a vacuum fan, and a water spraying device. The sweeping brush speed, vacuum fan power, and water spraying flow rate are adjusted according to control commands. The driving system adjusts the sanitation vehicle's speed, direction, and braking according to control commands.

Images