Unmanned sanitation vehicle path planning method based on artificial intelligence

Abstract

The method comprises the following steps: firstly, acquiring environment information, capturing surrounding information at the frequency of not less than 20Hz through a laser radar and other sensors, constructing a real-time updated three-dimensional high-precision environment map by fusing data through a deep learning algorithm, analyzing task requirements, receiving instructions or determining operation types and evaluating task priorities according to preset conditions, and performing real-time task management on the environment map. Locking a key area, pre-estimating the operation duration and generating a preliminary plan, finally constructing a path planning model, customizing a reward function and matching a weight based on a reinforcement learning algorithm and setting the driving efficiency as an optimization target, and training by using massive samples, so that the model masters an optimal path strategy. According to the invention, 24-hour efficient operation, precise environment sensing, rapid path planning, energy consumption reduction, cleaning quality improvement, labor cost reduction and urban environmental sanitation intelligence assistance can be effectively realized for the sanitation vehicle according to intelligent planning.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning for sanitation vehicles, and in particular to a path planning method for an unmanned sanitation vehicle based on artificial intelligence. Background Art

[0002] With the acceleration of urbanization, the scale and complexity of sanitation work are increasing. Traditional sanitation operations rely on manually driven sanitation vehicles, which has many disadvantages. On the one hand, manual operation is limited by working hours and energy, and it is difficult to achieve long-term, high-intensity continuous operation, resulting in low efficiency in the execution of sanitation tasks, especially when facing large-scale cleaning or garbage removal needs, and unable to meet the requirements of urban environmental maintenance in a timely manner. For example, in the bustling commercial areas of some large cities, the amount of garbage generated increases dramatically during peak hours in the morning and evening. Manually driven sanitation vehicles are often difficult to reach key areas in time for cleaning due to traffic congestion and parking difficulties, resulting in garbage accumulation that affects the city appearance.

[0003] On the other hand, manually driven sanitation vehicles lack scientificity and accuracy in route planning. Drivers rely on experience to judge routes and cannot fully consider the dynamic changes in real-time road conditions, traffic rules, and garbage distribution. This not only easily causes vehicles to drive empty, take detours, and waste energy, but may also cause accidents due to violations of traffic rules. Moreover, it is difficult for humans to fully and accurately perceive the surrounding environment information of the vehicle. For example, in narrow streets with dim light and many obstacles, the driver's field of vision is limited, which is prone to safety hazards such as collisions.

[0004] Furthermore, modern cities have increasingly stringent requirements for environmental protection standards and operation quality, and traditional sanitation operation models are difficult to meet the needs of refined management. Different areas have different cleaning priorities. For example, areas around schools and hospitals need to focus on ensuring hygiene, and traditional models cannot quickly and flexibly allocate resources. At the same time, with the continuous increase in labor costs, the human expenditure burden of the sanitation department has increased, and intelligent means are urgently needed to reduce costs and improve efficiency. In summary, it is urgent to develop an artificial intelligence-based path planning method for unmanned sanitation vehicles to fill the gaps in traditional sanitation operations and adapt to the needs of urban development. Summary of the invention

[0005] The present invention proposes an artificial intelligence-based unmanned sanitation vehicle path planning method to solve the problems mentioned in the above-mentioned prior art.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a path planning method for an unmanned sanitation vehicle based on artificial intelligence, comprising the following steps: first, environmental information is collected, and road boundaries, obstacles, and garbage distribution surrounding information are captured at a frequency of not less than 20Hz with the help of laser radar, high-definition camera, and millimeter-wave radar sensor, with an accuracy of centimeters. A deep learning algorithm is used to fuse multi-source data to build a real-time updated three-dimensional environmental map, followed by task demand analysis, receiving instructions from the management center or according to preset plans, clarifying the types of sanitation operations such as cleaning and transportation, and comprehensively assessing the task priority based on the urgency of regional cleaning, the degree of garbage pollution, and the impact of traffic flow. Combined with the constructed map, the key operation area is locked, the operation time is estimated, and a preliminary task plan is generated. Finally, a path planning model is constructed. Based on the reinforcement learning algorithm, driving efficiency, energy consumption, operation quality, and compliance with traffic rules are set as optimization goals, multi-sub-item reward functions are customized and weighted reasonably, and environmental maps and task plans are input. Training with massive samples allows the model to master the optimal path strategy for complex scenarios to ensure generalization and decision-making accuracy.

[0007] Furthermore, a method for path planning of an unmanned sanitation vehicle using the artificial intelligence includes a vehicle status monitoring step: during the entire operation, the power, tire pressure, and operating conditions of mechanical components of the sanitation vehicle are monitored in real time at a frequency of not less than 5 Hz. If an abnormality occurs, an immediate warning is issued and the path planning model is linked to plan a safe path to the nearest maintenance station based on the vehicle's position. When monitoring the power, a power estimation model is used to judge the remaining mileage based on the vehicle's current driving status, including speed, load, remaining battery power, and battery charging and discharging characteristics. The formula is: Remaining_range = B*(1-D*Load*Speed)*exp(-E*Speed)+F, where B, D, E, and F are coefficients obtained by fitting a large amount of actual driving data.

[0008] Furthermore, it also includes remote assisted decision-making steps: establishing a real-time communication link with the remote control center. When the path planning model encounters complex scenarios, including chaotic construction sites and extreme weather conditions, it automatically uploads on-site data and task details. Professionals give auxiliary decision-making suggestions within 30 seconds. After the vehicle receives the suggestions, the path is optimized. The remote control center adopts a distributed computing architecture, and the computing nodes process data in parallel to improve data processing speed and ensure timeliness of decision-making. At the same time, redundant backup technology is used to ensure the stability of the communication link. When the main link fails, it automatically switches to the backup link, and the switching time does not exceed 5 seconds.

[0009] Furthermore, it also includes the step of multi-vehicle collaborative operation: when multiple sanitation vehicles are working in the same area, they share their location, task progress, and environmental perception information in real time through the Internet of Vehicles, and dynamically adjust their respective paths every 10 seconds based on the collaborative algorithm to avoid overlapping operations, waste, and improve efficiency. The collaborative algorithm adopts an allocation strategy based on the auction mechanism, dividing the operating area into multiple sub-areas. Each vehicle bids for the sub-area of ​​interest, and determines the bid value based on the vehicle's own status, including the remaining power, the remaining life of the cleaning equipment, the difficulty of the sub-area operation, and the distance from the current location. The bidding formula is:

[0010] Bid=G*(1-Battery_level)*(1-Equipment_lifetime)+H*Distance 2 *Difficulty+I, G, H, and I are weight coefficients, which are determined through simulation optimization. The winning vehicle obtains the right to operate in the sub-area, achieving efficient collaboration.

[0011] Furthermore, it also includes historical data learning optimization steps: regularly collect and organize the path planning, environment, task completion, and historical data of the sanitation vehicles' past operations, use big data analysis to discover patterns and optimization points every week, and feed back to the path planning model for targeted training and optimization to improve the model's ability to handle similar scenarios. In big data analysis, cluster analysis algorithms are used to classify historical scenes and extract features for different categories of scenes.

[0012] Furthermore, in the S1 environmental information collection step, in addition to the above features, the high-definition camera also uses an intelligent exposure adjustment algorithm to dynamically adjust the exposure time according to the ambient light intensity, further improving the image quality and ensuring the accuracy of data collection.

[0013] Furthermore, in the S2 task demand analysis step, the system has the function of dynamically adjusting tasks. When the environment or instructions change, the system re-evaluates the priority, adjusts the operation sequence and time within 10 seconds to ensure flexible and efficient operations. When receiving a new emergency report of garbage accumulation, the system immediately recalculates the priority, adjusts the operation route, and responds quickly.

[0014] Furthermore, in the S3 path planning model construction step, transfer learning technology is introduced to reuse model parameters of similar scenarios and reduce training time and sample requirements.

[0015] Furthermore, in the S4 real-time path dynamic adjustment step, the re-planned path is evaluated using a risk probability model, taking into account obstacles, road conditions, and conflicts between people and vehicles, and the results are visualized for monitoring personnel. The visualization uses augmented reality (AR) technology to directly mark risk points on the real-life image seen by monitoring personnel.

[0016] Furthermore, it also includes a path backtracking and experience accumulation module, which tracks the execution of the plan after the task is completed, summarizes the lack of experience, and feeds back data to the model for subsequent training to improve adaptability and performance. During the backtracking analysis, a comparative analysis method is used to compare the actual path with the optimal theoretical path to find out the cause of the deviation.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] First of all, it greatly improves operational efficiency. Unmanned sanitation vehicles can operate 24 hours a day without interruption. Based on real-time task requirements and environmental conditions, they can quickly plan the optimal route and quickly reach the operating area. For example, they can efficiently complete the cleaning of main roads in the early morning when there are few people, thus avoiding interference during peak traffic hours.

[0019] Secondly, it optimizes energy utilization and accurately plans routes to reduce unnecessary trips and energy consumption, which is in line with the concept of environmental protection. Furthermore, it ensures the quality of operations by using high-precision sensors and intelligent algorithms to fully perceive the environment and ensure that there are no blind spots, such as accurately locating garbage in street corners. From a safety perspective, it reduces the risk of accidents, avoids human driving errors, and can quickly replan routes to avoid risks when dealing with emergencies.

[0020] It can also realize intelligent resource allocation and dynamically arrange vehicles according to regional cleaning priorities and garbage distribution to meet the needs of different scenarios.

[0021] At the same time, it will reduce labor costs, ease the employment pressure of the sanitation department, promote the intelligent and modern transformation of the sanitation industry, and create a cleaner and more livable environment for the city. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a schematic block diagram of an artificial intelligence-based unmanned sanitation vehicle path planning method proposed in the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0025] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, and it can be the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0026] Reference Figure 1 :A method for path planning of an unmanned sanitation vehicle based on artificial intelligence, comprising the following steps:

[0027] S1. Environmental information collection steps: Use a variety of high-precision sensors such as laser radar, high-definition camera, millimeter-wave radar, etc. The laser radar adopts a rotating multi-beam design, with a horizontal viewing angle of up to 360° and a vertical viewing angle of not less than 120°, and can accurately detect objects of different heights and orientations. The image resolution of the high-definition camera is not less than 1080P, with backlight compensation and wide dynamic adjustment functions. It uses an image sensor based on HDR technology and can adapt to changes in light intensity of 0-100000lux. The operating frequency band of the millimeter-wave radar is 77GHz-81GHz, with a detection distance of up to 200 meters and an accuracy of ±0.5 meters. It continuously collects environmental information around the sanitation vehicle at a frequency of not less than 20Hz, covering road boundaries, obstacle locations, traffic signs and garbage distribution, etc. The sensor accuracy reaches the centimeter level to ensure accurate and comprehensive information.

[0028] With the help of deep learning-based fusion algorithms, such as deep convolutional neural networks (DCNN) for feature extraction, combined with the Kalman filter algorithm to denoise and remove redundancy from multi-source data, a three-dimensional high-resolution environmental map is constructed in real time. The DCNN model structure contains 10 convolutional layers, 5 pooling layers, and 3 fully connected layers. The model parameters are optimized through training with a large amount of labeled data, so that it can efficiently extract environmental features. The Kalman filter algorithm continuously corrects the errors of sensor data through the prediction-correction steps to ensure the accuracy of map construction. The map update frequency is not less than 1Hz, providing real-time and accurate basis for subsequent decision-making.

[0029] S2. Task demand analysis steps: Immediately receive task instructions from the management center or according to the preset plan, clarify the type of sanitation operation, such as cleaning, transportation, cleaning, etc., and determine the task priority and operation time window. The task priority is comprehensively evaluated based on three factors: the urgency of regional cleaning (C), the degree of garbage pollution (P), and the impact of traffic flow (T). The priority evaluation formula is: Priority = (C*ω1+P*ω2+T*ω3) / (C+P+T+ε), where ω1, ω2, and ω3 are weight coefficients, which are determined through statistical analysis of historical operation data in different regions. For example, in a busy commercial area, ω1 = 0.4, ω2 = 0.4, ω3 = 0.2; in a residential area, ω1 = 0.3, ω2 = 0.5, ω3 = 0.2, and ε is a minimum value (such as 0.01) to prevent the denominator from being zero, ensuring that the priority evaluation is reasonable and accurate, and can flexibly adapt to the needs of different scenarios.

[0030] Combined with the environmental map, we can lock in key operation areas, such as garbage-accumulated sections and crowded cleaning areas, estimate the operation time, and generate a preliminary task plan. When estimating the operation time, we consider factors such as the area (S), garbage density (D), and cleaning equipment efficiency (E), and use the operation time estimation formula: Time = S*D / (E*(1-k1*S-k2*D)), where k1 and k2 are correction coefficients obtained based on statistical analysis of different types of cleaning equipment and past operation data. For example, for a certain type of sweeper, k1 = 0.05, k2 = 0.1 (units are uniformly adapted), so as to generate a more realistic operation plan.

[0031] S3. Steps for building a path planning model: Build a model based on the reinforcement learning algorithm, with driving efficiency (measured by the distance traveled per unit time), energy consumption (energy consumption per unit operating distance), operation quality (assessed by the cleanliness compliance rate of the operating area), and traffic rule compliance (the inverse of the ratio of traffic rule violations to the total number of driving times) as optimization goals, and set a multi-item reward function:

[0032] Reward=α*Efficiency*(1-Energy)+β

[0033] β*Quality*Compliance / (1+Compliance)+γ, where α, β, and γ are weight coefficients, which are set according to the importance the sanitation department attaches to each target and the actual operating cost analysis. For example, α=0.4, β=0.4, and γ=0.2, so that the model training is more in line with the actual operation needs.

[0034] The real-time environment map and task planning are used as input, and no less than 100,000 sets of samples and 1,000 iterations are used for training, so that the model can master the optimal path strategy for complex scenarios and ensure generalization and decision-making accuracy. During the training process, the ε-greedy strategy based on exploration-utilization balance is adopted, with a certain probability of ε randomly selecting actions (path exploration), and a probability of 1-ε selecting the current optimal action (using existing experience). The ε value gradually decreases with the progress of training, and the initial value is set to 0.5, and it decreases by 0.05 every 100 iterations to ensure that the model can fully explore new paths and converge quickly using the learned knowledge.

[0035] S4. Real-time path dynamic adjustment steps: When the sanitation vehicle is driving, the model is called within 0.5 seconds to generate the optimal path based on real-time information and task progress to ensure continuous and efficient driving.

[0036] In case of emergencies, such as temporary obstacles, congestion, and emergency tasks, re-planning is done within 1 second, giving priority to safety and reducing impact, and flexibly adjusting operation strategies. When re-planning the route, a risk probability model is used to assess the risk of the new route, taking into account factors such as obstacle density (ρ), road condition coefficient (κ, based on a comprehensive assessment of road flatness, slope, etc.), and the possibility of conflict with surrounding vehicles and pedestrians (λ). The risk assessment formula is: Risk = ρ*(a1*κ+a2*λ)+b1*ρ 2 +b2*κ 2 +b3*λ 2 +c, a1, a2, b1, b2, b3, c are weight coefficients, which are determined based on historical accident data and simulation tests, such as a1=0.3, a2=0.3, b1=0.1, b2=0.1, b3=0.1, c=0.1. The risk assessment results are visualized to the monitoring personnel to facilitate real-time monitoring and decision-making.

[0037] The present invention also includes a vehicle status monitoring step: during the entire operation, the key status information such as the power, tire pressure, and operating status of mechanical components of the sanitation vehicle are monitored in real time at a frequency of not less than 5Hz. Once an abnormality occurs, an immediate warning is issued and the path planning model is linked to plan a safe path to the nearest maintenance station according to the vehicle position. When monitoring the power, a power estimation model is used to estimate the remaining mileage based on factors such as the current driving state of the vehicle (including speed, load, etc.), the remaining battery power, and the battery charging and discharging characteristics. The formula is: Remaining_range = B*(1-D*Load*Speed)*exp(-E*Speed)+F, where B, D, E, and F are coefficients obtained by fitting a large amount of actual driving data. For example, for a certain type of battery, B = 120, D = 0.02, E = 0.03, and F = 20, to ensure that power risks can be discovered in a timely manner and to ensure the safety of vehicle operation.

[0038] The present invention also includes a remote assisted decision-making step: establishing a real-time communication link with the remote control center. When the path planning model encounters complex scenarios, such as a chaotic construction site or extreme weather, the site data and task details are automatically uploaded. Professionals give auxiliary decision-making suggestions within 30 seconds, and the vehicle optimizes the path after receiving them. The remote control center adopts a distributed computing architecture, with multiple computing nodes processing data in parallel to improve data processing speed and ensure the timeliness of decision-making. At the same time, redundant backup technology is used to ensure the stability of the communication link. When the main link fails, it automatically switches to the backup link, and the switching time does not exceed 5 seconds.

[0039] The present invention also includes a multi-vehicle collaborative operation step: when multiple sanitation vehicles are working in the same area, they share information such as location, task progress, and environmental perception in real time through the Internet of Vehicles, and dynamically adjust their respective paths every 10 seconds based on the collaborative algorithm to avoid overlapping operations, waste, and improve efficiency. The collaborative algorithm adopts an allocation strategy based on an auction mechanism, dividing the operating area into multiple sub-areas. Each vehicle bids for the sub-area of ​​interest, and determines the bid value based on factors such as the vehicle's own status (such as remaining power, remaining life of cleaning equipment, etc.), the difficulty of sub-area operations (assessed based on the amount of garbage, road complexity, etc.), and the distance from the current location. The bidding formula is: Bid = G*(1-Battery_level)*(1-Equipment_lifetime)+H*Distance 2 *Difficulty+I, G, H, and I are weight coefficients determined through simulation optimization, such as G=0.3, H=0.4, and I=0.3. The winning vehicle obtains the right to operate in the sub-area, thereby achieving efficient collaboration.

[0040] The present invention also includes a historical data learning optimization step: regularly collect and organize historical data such as the path planning, environment, and task completion of the sanitation vehicle's past operations, use big data analysis to mine patterns and optimization points every week, and feed back to the path planning model for targeted training and optimization to improve the model's processing capabilities for similar scenes. When analyzing big data, a cluster analysis algorithm is used to classify historical scenes, and features are extracted for scenes of different categories. For example, for scenes with narrow roads and a lot of garbage, features such as road width and garbage density are extracted, and these features are used as model inputs for targeted training to improve the training effect.

[0041] In the present invention, in the S1 environmental information collection step, in addition to the above-mentioned characteristics, the high-definition camera also adopts an intelligent exposure adjustment algorithm to dynamically adjust the exposure time according to the ambient light intensity, further improving the image quality and ensuring the accuracy of data collection.

[0042] In the present invention, in the S2 task demand analysis step, the task dynamic adjustment function is provided. When the environment or instructions change, the priority is re-evaluated within 10 seconds, and the operation sequence and time are adjusted to ensure flexible and efficient operation. When a new emergency report of garbage accumulation is received, the priority is immediately recalculated, the operation route is adjusted, and a quick response is made to ensure urban environmental sanitation.

[0043] In the present invention, in the S3 path planning model construction step, the transfer learning technology is introduced to reuse the model parameters of similar scenarios, with a reuse rate of not less than 30%, reducing the training time and sample requirements. The model parameters trained in the flat commercial area are migrated to the new similar commercial area, and only a small amount of sample fine-tuning is required to quickly adapt to the new environment and accelerate model deployment.

[0044] In the present invention, in the S4 real-time path dynamic adjustment step, the re-planned path is evaluated using a risk probability model, taking into account obstacles, road conditions, conflicts between people and vehicles, etc., and the results are visualized for monitoring personnel. The visualization uses augmented reality (AR) technology to directly mark the risk points on the real-life image seen by the monitoring personnel, which is clear at a glance and facilitates timely decision-making.

[0045] The present invention also includes a path backtracking and experience accumulation module. After the task is completed, the execution of the planning is backtracked, the lack of experience is summarized, and the data is fed back to the model for subsequent training to improve adaptability and performance. During the backtracking analysis, the comparative analysis method is used to compare the actual path with the optimal theoretical path to find out the cause of the deviation, such as the path deviation caused by temporary traffic control, record the control information, and optimize the path in advance when encountering similar situations next time.

[0046] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A path planning method for an unmanned sanitation vehicle based on artificial intelligence, characterized in that: The following steps are involved: Environmental information collection: Use lidar, high-definition cameras, and millimeter-wave radar sensors to capture road boundaries, obstacles, and information about garbage distribution, and use deep learning algorithms to fuse multi-source data to build a real-time updated three-dimensional environmental map; Task demand analysis: Receive instructions from the management center or follow preset plans to identify types of sanitation operations such as cleaning and transportation, and assess task priorities based on the urgency of regional cleaning, the degree of garbage pollution, and the impact of traffic flow; Combine the constructed map to lock the operation area, estimate the operation duration, and generate a preliminary task plan; Path planning model construction: Based on the reinforcement learning algorithm, driving efficiency, energy consumption, operation quality, and compliance with traffic rules are set as optimization goals, multi-sub-item reward functions are customized and weights are reasonably assigned, environmental maps and task plans are input, and samples are used for training to allow the model to calculate the optimal path strategy for the scenario.

2. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: It also includes a vehicle status monitoring step: throughout the operation, the battery level, tire pressure, and operating status of mechanical components of the sanitation vehicle are monitored in real time at a frequency of not less than 5Hz; if an abnormality occurs, an immediate warning is issued and the path planning model is linked to plan a safe path to the nearest maintenance station based on the vehicle's location; when monitoring the battery level, a battery estimation model is used to determine the remaining mileage based on the vehicle's current driving status, including speed, load, remaining battery power, and battery charging and discharging characteristics. The formula is: Remaining_range = B*(1-D*Load*Speed)*exp(-E*Speed)+F, where B, D, E, and F are coefficients.

3. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: It also includes remote assisted decision-making steps: establishing a real-time communication link with the remote control center, and automatically uploading site data and task details when the path planning model encounters complex scenarios, including chaotic construction sites and extreme weather conditions; After receiving the vehicle, the route is optimized. The remote control center adopts a distributed computing architecture, with computing nodes processing data in parallel. At the same time, redundant backup technology is used to automatically switch to the backup link when the main link fails.

4. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: It also includes the steps of multi-vehicle collaborative operation: when multiple sanitation vehicles are working in the same area, they share their location, task progress, and environmental perception information in real time through the Internet of Vehicles, and dynamically adjust their respective paths every 10 seconds based on the collaborative algorithm; The collaborative algorithm adopts an allocation strategy based on the auction mechanism, dividing the operating area into multiple sub-areas. Each vehicle bids for the sub-area of ​​interest, and determines the bid value based on the vehicle's own status, including the remaining power, the remaining life of the cleaning equipment, the difficulty of operating in the sub-area, and the distance from the current location. The bidding formula is: Bid = G*(1-Battery_level)*(1-Equipment_lifetime)+H*Distance 2*Difficulty+I, where G, H, and I are weight coefficients determined through simulation optimization. The winning vehicle obtains the right to operate in the sub-area.

5. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: It also includes historical data learning and optimization steps: regularly collect and organize the path planning, environment, task completion, and historical data of the sanitation vehicles' past operations, use big data analysis to explore patterns and optimization points every week, and feed back to the path planning model for targeted training and optimization. During big data analysis, a clustering analysis algorithm is used to classify historical scenes and extract features for different categories of scenes.

6. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: In the S1 environmental information collection step, the high-definition camera also uses an intelligent exposure adjustment algorithm to dynamically adjust the exposure time according to the ambient light intensity to improve image quality.

7. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: In the S2 task requirement analysis step, the system has the function of dynamically adjusting tasks. When the environment or instructions change, the priority level is re-evaluated, the operation sequence and time are adjusted within 10 seconds. When a new emergency report of garbage accumulation is received, the priority level is immediately recalculated and the operation route is adjusted.

8. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: In the S3 path planning model construction step, transfer learning technology is introduced to reuse model parameters of similar scenarios, reducing training time and sample requirements.

9. The path planning method for an unmanned sanitation vehicle based on artificial intelligence according to claim 1 is characterized in that: In the S4 real-time path dynamic adjustment step, the re-planned path is evaluated using a risk probability model, taking into account obstacles, road conditions, and conflicts between people and vehicles. The results are visualized for monitoring personnel. The visualization uses augmented reality (AR) technology to mark risk points directly on the real-life image seen by the monitoring personnel.

10. The method for path planning of an unmanned sanitation vehicle based on artificial intelligence according to claim 1, characterized in that: It also includes a path backtracking and experience accumulation module, which traces back the execution of the plan after task completion, summarizes the lack of experience, and feeds back data to the model for subsequent training. During the backtracking analysis, a comparative analysis method is used to compare the actual path with the optimal theoretical path to find out the cause of the deviation.

Images