A full-process automatic cleaning and transporting control method and system of an intelligent garbage cleaning and transporting vehicle

Abstract

The application relates to the technical field of automatic garbage collection and transportation, and discloses a full-process automatic garbage collection and transportation control method and system for an intelligent garbage collection and transportation vehicle. The method obtains a collection and transportation position and a current vehicle position in response to a collection and transportation instruction, plans and travels along a collection and transportation route; after arriving at the collection and transportation position, a collection and transportation target is identified, a collection and transportation amount is extracted and verified, and single-point collection and transportation is started after verification; when an abnormality occurs, the abnormality can be ignored to continue collection and transportation or the position can be skipped. When single-point collection and transportation is performed, the posture of the collection and transportation target is identified and matched with preset control logic, a mechanical arm or a collection and transportation inlet is controlled to complete collection and transportation work; after completion, the remaining amount of collection and transportation is identified, a travel direction is determined in combination with the collection and transportation amount of the next position, and if the remaining amount is insufficient, the starting point is returned to plan a new route; the new route can be executed by the vehicle itself or sent to an idle cooperative collection and transportation vehicle within a preset range around the starting point for execution. The method realizes full-process automatic garbage collection and transportation, improves collection and transportation work efficiency and rationality of resource scheduling.

Description

Technical Field

[0001] This application relates to the technical field of automated waste collection and transportation, and in particular to a fully automated waste collection and transportation control method and system for an intelligent waste collection vehicle. Background Technology

[0002] As core equipment for ensuring urban environmental sanitation, garbage trucks are responsible for collecting and transporting domestic waste, commercial waste, and some industrial waste. They are crucial infrastructure for maintaining a clean urban environment and ensuring residents' quality of life. Their operational efficiency directly impacts the smooth operation of the entire waste management chain and is of great significance for reducing waste accumulation and pollution, and improving the level of refined urban management. Currently, garbage trucks are widely used in various scenarios such as residential communities, commercial districts, and industrial parks, providing fundamental support for urban environmental governance.

[0003] At present, the operation mode of garbage collection vehicles in China is still mainly based on manual operation of mechanized equipment. During operation, at least two people are required: a driver and an operator. The driver is responsible for driving and parking the vehicle, while the operator manually controls the mechanical arm, lifting device and other components to complete actions such as grabbing, lifting, dumping and resetting the garbage container.

[0004] The efficiency and stability of manually operated mechanized waste collection equipment are highly dependent on the professional skills and practical experience of the operators. Skilled operators can quickly adapt to complex scenarios and precisely control the equipment's movements, while novice operators are prone to errors in judgment and lack of operational familiarity, leading to longer work times and accelerated equipment wear and tear. Furthermore, manual operation is significantly affected by factors such as fatigue, emotional state, and environmental visibility, which can easily result in problems such as inaccurate robotic arm positioning and incomplete dumping, impacting collection efficiency. Summary of the Invention

[0005] To achieve higher waste collection efficiency, this application provides a fully automated waste collection control method and system for intelligent waste collection vehicles.

[0006] Firstly, this application provides a fully automated waste collection and control method for intelligent waste collection vehicles, employing the following technical solution: A fully automated waste collection and control method for intelligent waste collection vehicles includes the following steps: Obtain a collection instruction, respond to the collection instruction by calling a collection vehicle and obtaining collection address information, obtain multiple collection locations from the collection address information, obtain the current location, calculate the collection route based on the current location and multiple collection locations, and proceed based on the collection route; After passing the cleaning location, the system acquires on-site image data, identifies the cleaning target from the image data, and extracts the amount of material to be cleaned based on preset sensors. The system verifies the cleaning target and the amount of material to be cleaned. If the verification is successful, single-point cleaning is initiated; otherwise, an on-site cleaning anomaly is displayed, and the system waits for instructions. Based on the on-site cleaning anomaly display, if an instruction to ignore the anomaly is received, single-point cleaning is initiated; if an instruction to skip the site is received, the system proceeds to the next cleaning location. The steps for initiating single-point cleaning include: identifying the posture of the cleaning target; matching the preset cleaning control logic from the preset control logic database based on the posture of the cleaning target; controlling the preset robotic arm to grab the cleaning target and dump the cleaning contents according to the cleaning control logic, or controlling the cleaning inlet to open to receive the cleaning contents according to the cleaning control logic. After completing a single-point cleanup, the current remaining cleanup volume is identified, and the cleanup volume for the next cleanup location is obtained. If the remaining cleanup volume is greater than the cleanup volume for the next cleanup location, the vehicle proceeds to the next cleanup location; otherwise, it returns to the preset starting point and regenerates the cleanup route based on the unvisited cleanup location. After reaching the starting point, the vehicle proceeds based on the regenerated cleanup route, or the regenerated cleanup route is sent to other cleanup vehicles located within a preset range around the starting point that are in an idle state or have returned to the starting point.

[0007] By adopting the above technical solutions and through the fully automated waste collection control logic, intelligent planning of waste collection routes, accurate identification and attitude adaptation of waste collection targets, automated execution of mechanical actions, and dynamic scheduling of waste collection capacity can be achieved. Efficient waste collection operations can be completed without relying on human operating experience, significantly reducing personnel collaboration costs and operational errors. It effectively avoids the impact of factors such as human fatigue and environmental visibility on operational efficiency. At the same time, through anomaly verification mechanisms and cross-vehicle route collaborative scheduling, the stability and resource utilization of the entire waste collection chain are improved.

[0008] Furthermore, the step of calculating the waste collection route based on the current location and multiple waste collection locations includes the following steps: The current location and multiple cleaning locations are encoded to generate an initial population. Each element in the initial population corresponds to a cleaning path. All cleaning paths start from the current location, pass through all cleaning locations, and then return to the current location. Construct a fitness function to calculate the total cleaning distance of each cleaning route based on the current location and multiple cleaning locations, calculate the route feasibility of each cleaning route based on a preset map database, calculate the cleaning priority of each cleaning route based on a preset priority database, and use the fitness function to calculate the route population based on the total cleaning distance, route feasibility and cleaning priority. The route population is iteratively updated through selection, crossover, and / or mutation operations, retaining the individual with the best fitness in the route population; Obtain the set iteration termination conditions, including the iteration count reaching a preset iteration threshold or the individual with the best fitness remaining unchanged, and output the corresponding individual as the cleaning route.

[0009] By adopting the above technical solutions, the optimal or near-optimal waste disposal routes can be efficiently converged, which not only ensures the shortest distance characteristic of the routes to reduce transportation costs, but also takes into account the path accessibility and waste disposal task priority ranking requirements in actual operations, effectively avoiding ineffective detours and path conflicts.

[0010] Furthermore, the step of calculating the waste collection route based on the current location and multiple waste collection locations includes the following steps: Initialize the distance matrix between the current location and multiple cleaning locations, and set the set of unvisited nodes and the set of visited nodes; Update the current location and the arrival attributes from the current location to each unvisited node in real time, and select nodes whose arrival attributes meet the preset filtering conditions to add to the visited set; Iteratively update the arrival attributes of unvisited nodes, obtain the congestion level corresponding to the unvisited nodes based on the preset map database, and adjust the weight of the arrival attributes of unvisited nodes according to the congestion level. After traversing all nodes in the unvisited node set, a cleaning route is generated from the current position through all cleaning locations based on the arrival attributes and corresponding weights of all nodes in the visited set. The overall congestion value is calculated based on the obtained congestion level. If the overall congestion value is less than the preset reference congestion value, the arrival attribute is the current location and the temporary distance from the current location to each unvisited node. Otherwise, the arrival attribute is the current location and the temporary arrival time from the current location to each unvisited node.

[0011] By adopting the above technical solutions, we can ensure the complete traversal of multiple waste collection nodes and the orderly generation of routes, and realize dynamic strategy adaptation in congestion scenarios. When there is no congestion, we prioritize the shortest distance to reduce energy consumption, and when there is congestion, we prioritize the shortest time to avoid delays. This effectively improves the real-time performance, flexibility and scenario adaptability of waste collection route planning, and significantly reduces path time and unnecessary waiting.

[0012] Furthermore, the step of calculating the collection route based on the current location and multiple collection locations includes the following steps: obtaining temporary instructions in real time, extracting temporary additional locations from the temporary instructions, and inserting the temporary additional locations into the collection route; obtaining the current collection volume loaded on the collection vehicle, and calculating the collection ratio based on the current collection volume and the total collection volume of the collection vehicle; and adjusting the interval between the temporary additional locations in the collection route and the current location of the collection vehicle based on the negative correlation of the collection ratio.

[0013] By adopting the above technical solution, when the garbage truck has sufficient loading capacity, priority is given to handling temporary locations with large amounts of garbage, avoiding secondary round trips due to loading limitations. This not only improves the dynamic adjustment capability of the garbage collection route and the response efficiency of temporary tasks, but also optimizes the loading utilization rate of the garbage truck, reducing ineffective transportation and resource waste.

[0014] Furthermore, the method also includes the following steps: Within a preset statistical period, obtain the maximum material quantity at each collection location during collection, and calculate the average material quantity corresponding to each collection location based on the maximum material quantity. The average of the average material quantities at all collection locations is then calculated as the comprehensive material quantity. The ratio of average material quantity to total material quantity is calculated as the material comparison value. Based on the material comparison value, the weight of the arrival attribute corresponding to the cleaning location is adjusted positively with the first adjustment parameter.

[0015] By adopting the above technical solutions, we can make full use of historical data to improve the targeting and scientific nature of route planning, optimize the utilization rate of the loading space of the waste collection vehicles, avoid wasting space due to loading a small amount of materials first, which would result in a large amount of materials not being able to be loaded later, reduce ineffective round-trip transportation, and further enhance the efficiency and loading rationality of the fully automated waste collection process.

[0016] Furthermore, the method also includes the following steps: Within a preset statistical period, obtain the longest time required to complete the collection at each collection location, and calculate the average time for each collection location based on the longest time. The average of the average times for all collection locations is then calculated as the comprehensive time. The ratio of average time to total time is calculated as the time comparison value. Based on the time comparison value, the weight of the arrival attribute corresponding to the collection location is adjusted negatively with the second adjustment parameter.

[0017] By adopting the above technical solutions, the time consumption dimension of the cleaning and sorting is optimized, allowing cleaning locations with shorter operation times to receive higher cleaning priority, effectively reducing the accumulation of time consumption in the entire cleaning and sorting operation, and avoiding the overall cleaning and sorting rhythm being delayed due to prioritizing the handling of long-time locations.

[0018] Furthermore, when the weights of the arrival attributes are adjusted simultaneously for both the material comparison value and the time comparison value, the first adjustment parameter is greater than the second adjustment parameter.

[0019] By adopting the above technical solutions, the optimization priority of the material quantity dimension is higher than that of the operation time dimension, which precisely meets the core needs of actual waste collection and transportation operations.

[0020] Furthermore, the step of identifying the target posture for waste removal and matching the preset waste removal control logic from the preset control logic database based on the target posture also includes the following sub-steps: Upon arrival at the collection location, the vehicle enters the pre-set collection bay and acquires on-site image data of the collection target. Based on a preset box template, box graphics are obtained from on-site image data, and the number of boxes is calculated based on the box graphics; Based on the collection location, the system retrieves the configured quantity and the corresponding quantity range from the preset configuration database. If the difference between the number of boxes and the configured quantity is calculated, and the difference is outside the quantity range, an abnormal box quantity prompt is generated. Otherwise, the system extracts the grasping structure and the corresponding grasping position based on the box image, guides the robotic arm to the grasping position, and docks with the grasping structure. After docking with the capture structure, the amount of remaining material in the target to be cleared is obtained in real time. The remaining material comparison value is calculated based on the remaining material amount and the total amount of material in the target to be cleared. The speed of dumping and clearing the contents is adjusted according to the positive correlation of the remaining material comparison value.

[0021] By adopting the above technical solutions, the success rate and positioning accuracy of the robotic arm's grasping are improved, and damage such as equipment collisions is reduced; it adapts to the dumping needs of different loading capacities, avoiding the problems of material spillage due to dumping too fast and time consumption due to dumping too slow.

[0022] Furthermore, the step of extracting the grasping structure and the corresponding grasping position based on the box graph also includes the following sub-steps: If no grab structure is extracted from the box graph, an error message will be displayed regarding the box graph. The box type, reference point, and orientation are calculated based on the box graphic. The simulated grabbing position is then generated based on the reference point, orientation, and box type using a pre-defined box dataset. If the simulated gripping position and the actual gripping position are outside the preset range or fail to connect successfully to the gripping structure, a prompt for manual intervention will be generated; otherwise, the robotic arm will be guided to the simulated gripping position and connect to the gripping structure.

[0023] By adopting the above technical solution, the problem of automated operation without a clear grasping position under abnormal posture is effectively solved; at the same time, by verifying the grasping simulated position with the preset position, problems such as invalid movements of the robotic arm due to position deviation and equipment collisions are avoided.

[0024] Secondly, this application provides a fully automated waste collection and transportation control system for intelligent waste collection vehicles, employing the following technical solution: A fully automated waste collection and transportation control system for an intelligent waste collection vehicle includes a processor, wherein the processor executes the steps of the fully automated waste collection and transportation control method for the intelligent waste collection vehicle as described in any of the preceding claims. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the steps of a fully automated waste collection and control method for intelligent waste collection vehicles.

[0026] Figure 2 It is a sub-step diagram that calculates the waste removal route based on the current location and multiple waste removal locations. Detailed Implementation

[0027] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0028] This application discloses a fully automated waste collection control method for an intelligent waste collection vehicle. The intelligent waste collection vehicle is equipped with an onboard central control system, and is equipped with an onboard positioning module, visual acquisition equipment, various sensors, a multi-joint robotic arm, a waste collection entrance opening and closing mechanism, a wireless communication module, and an automatic driving actuator. All components are electrically connected to the onboard central control system, which uniformly completes data reception, logical operations, and motion control. The entire process control of this method is centered on the onboard central control system. Figure 1 Specifically, it includes the following steps: The first step is to obtain collection instructions and plan the collection route. The vehicle-mounted central control system receives collection instructions from the background dispatch system via a wireless communication module, and can also receive collection requests reported by on-site terminals to generate collection instructions. After responding to the collection instructions, it calls the corresponding intelligent garbage collection vehicle according to the instruction content, and parses the collection address information from the collection instructions. It extracts multiple collection locations to be collected from the collection address information through a geographic information parsing algorithm. At the same time, it obtains the current location of the collection vehicle through the vehicle positioning module (GPS / BeiDou dual-mode positioning module). This current location can be the preset starting point of the collection vehicle, or the real-time location of the collection vehicle during the collection operation. Based on the obtained current location and multiple collection locations, the vehicle-mounted central control system calculates the collection route in combination with preset electronic map data, and issues advance instructions to the autonomous driving actuator to control the collection vehicle to move to the collection location according to the planned collection route. If the collection vehicle is not equipped with autonomous driving function, the vehicle-mounted central control system can also send the collection route to the vehicle display terminal for the driver to follow.

[0029] The second step is to complete on-site data collection and clearance verification upon arrival at the clearance location. After the garbage truck arrives at any designated collection location according to the planned route, the onboard central control system controls the vehicle to park in the preset work area and activates the visual acquisition equipment (high-definition industrial camera, visual recognition sensor) to acquire images of the collection location. The onboard central control system then calls a preset image recognition algorithm to process the image data and identify the collection target from the image. In this embodiment, the collection target is various sizes of garbage bins. Simultaneously, preset sensors are activated to extract the amount of waste to be collected at that location. These sensors can be distance sensors or weight sensors, or a combination of both. The distance sensor is a laser rangefinder or ultrasonic rangefinder, installed inside the garbage bin lid or at the robotic arm's execution end. It calculates the amount of waste collected by detecting the height of the accumulated garbage inside the bin. The weight sensor is a pressure sensor, installed at a preset placement point on the garbage bin. It obtains the amount of waste collected by detecting the overall weight difference of the garbage bin. If a combination of both is used, the garbage volume is calculated from the garbage height detected by the distance sensor, and the garbage weight is calculated by combining this with preset garbage density parameters. After calibration of the detected values, the amount of waste collected is obtained. After completing the identification of the waste collection target and the extraction of the waste collection volume, the vehicle-mounted central control system performs dual verification. The verification criteria are whether the waste collection target is a preset waste bin to be collected, and whether the amount of waste collected reaches a preset waste collection threshold. In this embodiment, the collection threshold is set by the background scheduling system according to the management needs of urban waste collection, generally a waste bin loading rate ≥70%. If the verification result is passed, it indicates that the waste bin at that collection location meets the collection conditions, and the vehicle-mounted central control system directly triggers a single-point collection command; if the verification result is... If the request fails, it indicates that there are no garbage bins available at the collection location or that the garbage bins contain too little garbage, and the collection conditions are not met. The vehicle's central control system immediately generates a collection anomaly alert, which can be simultaneously pushed to the vehicle's display terminal and the back-end dispatch system. At the same time, the collection vehicle enters a waiting state. If it receives an "ignore anomaly" instruction from the back-end dispatch system or on-site operators within the preset waiting time, the vehicle's central control system will initiate single-point collection. If it receives an "skip on-site" instruction, it will directly issue a progress instruction to the autonomous driving actuator to control the collection vehicle to move to the next collection location.

[0030] The third step is to initiate single-point collection and execute automated collection operations. Upon receiving a single-point collection command, the vehicle-mounted central control system identifies the target's posture based on on-site image data. It analyzes the placement of the trash can using image contour extraction and posture detection algorithms, including normal and abnormal postures such as upright, angular deviation, and tipping. The system then matches the collection control logic corresponding to the target's posture from a pre-set control logic database. This database stores the robotic arm's motion parameters, collection inlet opening and closing parameters, and action timing logic for collection targets of different sizes and postures. Based on the matched collection control logic, the vehicle-mounted central control system selects the appropriate collection execution method to complete the single-point collection operation. If the target requires a robotic arm... For mobile trash cans grasped by robotic arms, the pre-set multi-joint robotic arm is controlled to perform a series of actions such as extension, grasping, lifting, and rotation according to the motion parameters. After grasping the target, the robotic arm is controlled to adjust to the preset tilting angle to dump the contents of the target into the trash bin of the collection vehicle. After dumping, the robotic arm is controlled to reset. If the target is a trash can at a fixed location, it can be directly docked with the collection vehicle. The collection vehicle's collection entrance opening and closing mechanism is controlled to open the collection entrance according to the preset logic, allowing the contents of the trash can to be directly transported into the trash bin. After receiving, the collection entrance is controlled to close, thus completing the single-point collection operation at that location.

[0031] Step 4: Detect the remaining capacity for waste collection and complete the scheduling of the subsequent travel direction. After completing the single-point waste collection operation, the on-vehicle central control system activates the remaining capacity detection device配套 to the waste bin to identify the current remaining capacity for waste collection of the waste collection vehicle. The remaining capacity detection device is an ultrasonic distance sensor or a weight sensor installed in the waste bin. The ultrasonic distance sensor calculates the remaining loading volume of the waste bin by detecting the stacking height of the waste in the waste bin, and the weight sensor calculates the remaining loading weight of the waste bin by detecting the overall load of the waste bin. The remaining capacity for waste collection is characterized by the remaining loading volume or the remaining loading weight. At the same time, the on-vehicle central control system obtains the pre-calculated waste collection volume of the next waste collection location from the background scheduling system through the wireless communication module. This waste collection volume is obtained by the background scheduling system by combining the historical waste collection data and the real-time material quantity detection data of this location. The on-vehicle central control system compares the current remaining capacity for waste collection with the waste collection volume of the next waste collection location numerically. If the remaining capacity for waste collection is greater than the waste collection volume of the next waste collection location, it means that the waste bin still has enough loading space to receive the waste of the next waste collection location. The on-vehicle central control system directly issues a travel instruction to control the waste collection vehicle to travel to the next waste collection location. If the remaining capacity for waste collection is less than or equal to the waste collection volume of the next waste collection location, it means that the waste bin no longer has enough loading space and cannot承接 the subsequent waste collection tasks. The on-vehicle central control system immediately issues a return instruction to control the waste collection vehicle to return to the预设 starting point location, which is a预设 waste unloading and scheduling point such as a waste transfer station or a waste collection vehicle parking point. At the same time, the on-vehicle central control system extracts all the unvisited waste collection locations and regenerates the waste collection route in combination with the starting point location.

[0032] Step 5: Complete the route execution at the starting point and the multi-vehicle collaborative scheduling. After the waste collection vehicle arrives at the starting point location according to the return instruction, the on-vehicle central control system executes two waste collection route processing methods according to the actual state of the waste collection vehicle and the background scheduling requirements: First, if the waste bin恢复 full loading capacity after the waste collection vehicle completes waste unloading at the starting point location and there are no other waste collection tasks with higher priorities, the on-vehicle central control system directly controls the waste collection vehicle to出发 again to执行 the remaining waste collection tasks based on the regenerated waste collection route. Second, if the waste collection vehicle has not completed waste unloading at the starting point location or needs to承接 other waste collection tasks issued by the background scheduling system, the on-vehicle central control system sends the regenerated waste collection route to the background scheduling system through the wireless communication module. The background scheduling system进行 multi-vehicle scheduling. The background scheduling system筛选出 other intelligent waste collection vehicles located within the预设 range around the starting point location and in the idle state or the state of returning to the starting point location according to the coverage range of this waste collection route. The预设 range is设定 by the background according to the urban area division and the operation radius of the waste collection vehicle. The regenerated waste collection route is下发 to the on-vehicle central control systems of the筛选出 other waste collection vehicles, and the other waste collection vehicles承接并执行 the uncompleted part of this waste collection task to实现 multi-vehicle collaborative waste collection scheduling.

[0033] The fully automated waste collection control method of the intelligent waste collection vehicle in this embodiment, through the linkage of various supporting equipment by the on-board central control system, completes the entire process control from waste collection command response and route planning, to on-site data collection and waste collection verification, to single-point automated waste collection, surplus scheduling and multi-vehicle collaboration. The entire process does not rely on the professional skills and practical experience of operators, effectively reducing the cost of manual collaboration and operational errors. At the same time, by setting up a waste collection volume verification mechanism and an abnormal command handling mechanism, the impact of factors such as human fatigue and environmental visibility on waste collection operations is avoided. Combined with the dynamic scheduling of surplus waste and multi-vehicle collaboration strategy, the operational stability and resource utilization rate of the entire waste collection chain are greatly improved, providing reliable technical support for the refined and intelligent management of urban waste collection.

[0034] In this embodiment of the application, the step of calculating the waste collection route based on the current location and multiple waste collection locations specifically includes the following sub-steps: Sub-step 1: Encode the locations to generate the initial population for the genetic algorithm. The vehicle-mounted central control system encodes the current location of the waste collection vehicle and all waste collection locations to be collected using geographic coordinates. Each geographic location is mapped to a unique numerical code value using natural number encoding, ensuring a one-to-one correspondence between the code and the location. Using the current location of the waste collection vehicle as the starting and ending codes, the encoded values ​​of multiple waste collection locations are randomly arranged and combined to generate multiple different waste collection paths. All generated waste collection paths strictly follow the path rule of starting from the current location → sequentially passing through all waste collection locations → returning to the current location, without omissions or repetitions in passing through any waste collection location. All the waste collection paths generated above together constitute the initial population of the genetic algorithm. Each waste collection path in the initial population is a population individual. The number of individuals is preset by the vehicle-mounted central control system based on the number of waste collection locations. The more waste collection locations there are, the more individuals the initial population will have, thus ensuring the diversity of the initial path schemes.

[0035] Sub-step 2: Construct a multi-objective fitness function and calculate and select the route population. The vehicle-mounted central control system constructs a multi-objective optimization fitness function, which serves as the core quantitative basis for evaluating the merits of individuals along each path in the initial population. This fitness function integrates three evaluation indicators: total transport distance, route feasibility, and transport priority. Through weighted calculations, it achieves synergistic optimization of multiple indicators. The specific calculation and selection process is as follows: Calculate the total cleaning distance: The vehicle-mounted central control system calculates the total cleaning distance of each cleaning path in the initial population using the Euclidean distance formula based on the geographical coordinates of each location. This is used as the primary evaluation indicator, and the shorter the total distance, the better the performance of this indicator. Calculate route feasibility: Retrieve a preset map database containing information such as urban road traffic conditions, height and weight restrictions, rules for prohibiting access to work areas, and dedicated lanes for garbage collection operations. Use a route traversal algorithm to detect the road accessibility of each collection route. Deduct points for routes with restrictions or obstacles. Quantitatively calculate a route feasibility score of 0-100 as the second evaluation indicator. The higher the score, the stronger the route feasibility. Calculate collection priority: Retrieve the preset priority database. This database presets the collection priority weight of each collection location based on the type of garbage, the risk of garbage accumulation, and the importance of the area (such as business districts, residential areas, and industrial parks). The vehicle central control system calculates the overall collection priority of each collection route in sequence through a weighted summation algorithm, which serves as the third evaluation index. The higher the value, the stronger the adaptability of the route to the collection task. Comprehensive calculation and selection: The vehicle-mounted central control system assigns preset weight coefficients to the three evaluation indicators, performs negative normalization on the total transportation distance, and positive normalization on the route feasibility and transportation priority. Then, it substitutes these values ​​into the fitness function for comprehensive calculation to obtain the fitness value of each individual on each path in the initial population. Based on the fitness value from high to low, the individuals in the initial population are selected, and the individuals on each path whose fitness value meets the preset threshold are retained. These individuals together constitute the route population.

[0036] Sub-step 3: Sequentially execute selection, crossover, and mutation operations to iteratively update the route population. The vehicle-mounted central control system sequentially performs selection, crossover, and mutation operations on the selected route population. Through the coordination of these three operations, the route population is iteratively updated, continuously selecting individual paths with better fitness. All three operations are based on the calculation rules preset by the vehicle-mounted central control system. The specific operation method and the logic adapted to the waste collection scenario are as follows: Selection Operation: A roulette wheel selection algorithm combined with an elite retention strategy is used to perform the selection operation. The fitness values ​​of individuals on each path in the route population are normalized and converted into corresponding selection probabilities. Individuals with higher fitness values ​​have a greater probability of being selected. At the same time, the top N high-quality individuals with the highest fitness values ​​in the route population are directly retained and do not participate in subsequent random selection, achieving a precise "survival of the fittest" selection effect. In the cleaning scenario, if the route population contains 100 path individuals, among which 20 are high-quality individuals with short total distances, high feasibility, and high priority, the selection operation will directly retain these 20 high-quality individuals. The remaining 80 individuals will be randomly selected according to the selection probability. All the individuals ultimately retained will serve as the base population for the crossover operation. The core purpose of this operation is to ensure that the characteristics of high-quality paths are not lost, so that subsequent optimizations can be carried out based on high-quality path solutions, avoiding starting the calculation from scratch.

[0037] Crossover operation: A single-point or multi-point crossover operation is performed on the base population obtained after the selection operation. The vehicle-mounted central control system randomly selects two path individuals from the base population as parent individuals, randomly sets one or more crossover points, and swaps the encoding sequences of the two parent individuals after the crossover point to generate two new child path individuals. Simultaneously, the vehicle-mounted central control system verifies the validity of the generated child individuals, filtering out invalid child individuals that repeatedly pass through the same collection location or have not passed through all collection locations, retaining only valid child individuals that meet the path rule of "passing through all collection locations without repetition." In a collection scenario, if parent individual 1 is "current location → collection point A → collection point C → collection point B → current location" and parent individual 2 is "current location → collection point B → collection point C → collection point A → current location," the encoding sequences are swapped at collection point C to generate valid child individuals that meet the rules. The core purpose of this operation is to integrate the local path features of different high-quality parent individuals to explore new, more adaptive path schemes.

[0038] Mutation Operation: A bit mutation operation is performed on the valid offspring individuals obtained after the crossover operation. The vehicle-mounted central control system randomly fine-tunes the coding sequence of each offspring individual with a preset low mutation probability. Specifically, it randomly selects two different collection location codes from the offspring individual's coding sequence, swaps their order, and completes the mutation operation. The mutation probability is preset to low to avoid excessive mutation that could damage the local features of high-quality paths. Simultaneously, the vehicle-mounted central control system re-verifies the validity of the mutated individuals, retaining only valid ones. In a collection scenario, if the offspring individual's path is "current location → A → C → B → current location," the mutation operation can adjust it to "current location → A → B → C → current location," slightly adjusting the path structure. The core purpose of this operation is to increase the diversity of path solutions, prevent the genetic algorithm from getting trapped in local optima, and ensure that the globally optimal collection path can be explored.

[0039] The vehicle-mounted central control system treats the above selection-crossover-mutation operation as a complete iterative process, and iteratively updates the route population. After each iteration, it retains the individual with the best fitness in the population, continuously optimizes the path plan, and continuously improves the overall fitness value of the route population.

[0040] Sub-step four: Determine the iteration termination condition and output the optimal waste collection route. The vehicle-mounted central control system pre-sets the iteration termination condition for the genetic algorithm. This termination condition includes two judgment dimensions, and the iteration is determined to be terminated if either dimension is met: First, the number of iterations reaches a preset iteration threshold. This threshold is dynamically preset according to the number of waste collection locations. The more waste collection locations there are, the higher the iteration threshold is, ensuring that there are enough iterations to complete the path optimization; Second, after multiple consecutive iterations, the individual with the best fitness in the route population remains unchanged, that is, the fitness value of the optimal path does not increase and the path scheme does not optimize at all, indicating that the algorithm has converged to the optimal solution.

[0041] After each iteration update, the vehicle-mounted central control system synchronously determines the two termination conditions mentioned above. If either termination condition is met, the iterative update of the route population is immediately stopped, and the path individual with the best fitness in the current route population is taken as the final calculation result. The encoding sequence of the best individual is decoded, and the digital code is converted into the actual geographical coordinate route, which is the optimal transportation route for this transportation. If neither termination condition is met, the selection-crossover-mutation iterative operation is continued until the termination condition is met.

[0042] This embodiment uses a genetic algorithm to achieve intelligent planning of waste collection routes. It incorporates the total waste collection distance, path feasibility, and waste collection priority into the fitness function for multi-objective optimization. By using encoding to generate the initial population, iterative updates through selection-crossover-mutation, and a clear design of iterative termination conditions, the algorithm can efficiently converge to the optimal or near-optimal waste collection route.

[0043] In another embodiment of this application, the step of calculating the waste collection route based on the current location and multiple waste collection locations employs an improved Dijkstra algorithm to achieve intelligent path planning for multiple waste collection nodes, specifically including the following sub-steps: Sub-step 1: Initialize the distance matrix and set up sets of unvisited and visited nodes. The vehicle-mounted central control system uses the current location of the waste collection vehicle and each waste collection location as independent nodes for route planning. Based on the geographic coordinate information of each node and combined with the road network topology data in the map database, it calculates the actual travel distance between any two nodes and constructs an N-order distance matrix containing the current location and all waste collection locations, where N is the total number of nodes (N = number of waste collection locations + 1). This distance matrix stores the basic travel distance between any two nodes, providing basic data support for subsequent route iteration calculations. At the same time, the vehicle-mounted central control system sets up two data sets according to the geographic coordinate encoding of the nodes: an unvisited node set and a visited node set. The unvisited node set initially contains nodes corresponding to all waste collection locations, while the visited node set initially only contains the starting node corresponding to the current location of the waste collection vehicle. The two sets are dynamically updated mutually exclusive sets; any node can only belong to one set, and all nodes must be included in the set management.

[0044] Sub-step 2: Update the current location and arrival attributes in real time, and add the filtered nodes to the visited set. The vehicle-mounted central control system first defines the arrival attribute, which is a core quantitative indicator for evaluating the quality of the path from the current location to each unvisited node. Initially, it includes two basic dimensions: temporary distance and temporary arrival time, both retrieved from the map database. The temporary distance is the actual travel distance from the current location to the unvisited node on the road network, and the temporary arrival time is the congestion-free travel time calculated based on real-time basic traffic conditions in the map database. The vehicle-mounted central control system uses the last added node in the visited node set as the current location and updates the original values ​​of the arrival attribute from this current location to each unvisited node in real time, completing the real-time refresh of the attribute values. Next, the vehicle-mounted central control system judges the quality of the arrival attribute of each unvisited node according to preset filtering conditions. In this embodiment, the filtering condition is the optimal value after normalization of the arrival attribute, i.e., a shorter path or less travel time. An unvisited node that meets this condition is selected, removed from the unvisited node set, and added to the visited node set, completing a single node filtering and update. This filtering logic aligns with the greedy optimization core of the Dijkstra algorithm, ensuring that the optimal path node is selected at each step.

[0045] Sub-step 3: Iteratively update the arrival attributes of unvisited nodes, and adjust attribute weights based on congestion levels. The in-vehicle central control system's node selection and set update process is an iterative cycle. After each new node is added to the set of visited nodes, the arrival attributes of all unvisited nodes are iteratively updated. That is, based on the newly added node as a transit node, the arrival attribute values ​​from the current location to each unvisited node via the transit node are recalculated. If the attribute value of the transit path is better than the original attribute value, the original attribute value is replaced, thus achieving iterative optimization of arrival attributes. At the same time, this step is the core optimization point for improving the Dijkstra algorithm. The in-vehicle central control system will synchronously retrieve the real-time congestion levels of the corresponding road segments for each unvisited node from the map database. This map database is linked in real time with the urban traffic management system and stores the congestion levels of each road segment. The system collects real-time traffic data such as congestion level (mild / moderate / severe), congestion coefficient (quantized from 0 to 1, where 0 represents no congestion and 1 represents complete congestion), and congestion duration. The vehicle's central control system converts the congestion level into a standardized congestion coefficient of 0 to 1. Following the rule that the higher the congestion level, the lower the weight, the more dynamically the weight of the arrival attribute corresponding to each unvisited node is adjusted. The specific adjustment formula is: Adjusted weight = Basic weight × (1 - Standardized congestion coefficient), where the basic weight is the initial weight of the arrival attribute preset by the vehicle's central control system. The adjusted weight will serve as the core basis for subsequent judgment of the quality of arrival attributes. Through this weight adjustment, the proportion of arrival attributes corresponding to road segments with high congestion levels in the route evaluation is reduced.

[0046] Sub-step four: Calculate the overall congestion value and intelligently switch to the arrival attribute type based on the congestion status. During the iterative update of the arrival attribute, the vehicle-mounted central control system calculates the overall congestion value for this route planning in real time based on the standardized congestion coefficients corresponding to each unvisited node. To adapt to the traffic statistics needs of different cities, the overall congestion value supports two calculation methods: one is the arithmetic mean, which is the arithmetic average of the standardized congestion coefficients corresponding to all unvisited nodes; the other is the weighted average, which is the weighted average calculated after assigning preset weights to each standardized congestion coefficient based on the length of the passageway corresponding to each collection location and the priority of the collection task. The vehicle-mounted central control system can select the calculation method according to the configuration instructions of the background dispatch system, and the calculated result of the overall congestion value is also a standardized value of 0-1. Simultaneously, the vehicle-mounted central control system presets... There is a reference congestion value, which is configurable from 0.3 to 0.5 and can be dynamically adjusted according to the daily congestion situation in the city. The real-time calculated comprehensive congestion value is compared with the reference congestion value. If the comprehensive congestion value is less than the preset reference congestion value, the current collection route coverage area is determined to be in a non-congested state. At this time, the temporary distance is used as the sole evaluation dimension of the arrival attribute, and the shortest distance is given priority as the core planning path. If the comprehensive congestion value is greater than or equal to the preset reference congestion value, the current collection route coverage area is determined to be in a congested state. At this time, the evaluation dimension of the arrival attribute is intelligently switched to the temporary arrival time, and the shortest travel time is given priority as the core planning path, so as to achieve dynamic adaptation of the route planning strategy and the road congestion status.

[0047] Sub-step 5: Traverse all unvisited nodes and generate a cleaning route based on attributes and weights. The vehicle-mounted central control system continuously executes the aforementioned node filtering, attribute iteration update, and congestion weight correction operations until the set of unvisited nodes is empty. At this point, all waste collection location nodes have been traversed, and the set of visited nodes contains the current location of the waste collection vehicle and all nodes corresponding to all waste collection locations. The order in which the nodes are added is the optimal path order obtained through greedy optimization based on the improved Dijkstra algorithm. The vehicle-mounted central control system then initiates a path backtracking algorithm, using the last waste collection location node added to the set of visited nodes as the backtracking starting point. Based on the corrected arrival attributes, corresponding weights, and iteration update records of each node, it traces back to the optimal preceding transfer node of each node until it backtracks to the initial current location node of the waste collection vehicle, forming a complete reverse path sequence. Finally, the vehicle-mounted central control system arranges this reverse path sequence in the forward direction to generate a complete waste collection route that starts from the current location, passes through all waste collection locations in sequence, and finally returns to the current location. At the same time, it converts this route into standardized path data containing geographic coordinate sequences, driving segment information, and turning instructions, which can be directly sent to the autonomous driving actuator or pushed to the vehicle display terminal for the driver's reference.

[0048] This embodiment uses an improved Dijkstra algorithm to plan waste collection routes. Based on the traditional algorithm's greedy search and ensuring path optimality, it achieves ordered traversal of multiple waste collection locations by initializing a distance matrix and a dynamic node set. It also dynamically corrects the arrival attribute weights based on real-time congestion levels and intelligently switches between temporary distance and temporary arrival time based on the comprehensive congestion value.

[0049] In this embodiment of the application, in response to sudden temporary collection needs during urban waste collection, a dynamic adjustment step for the collection route is added to the step of calculating the collection route based on the current location and multiple collection locations. This includes the following sub-steps: Sub-step 1: Obtain temporary cleaning instructions in real time, extract temporary added locations, and complete validity verification. Throughout the entire process of the garbage truck traveling along the planned route, the on-board central control system maintains real-time communication with the back-end dispatch system, on-site operation terminals, and the urban sanitation service platform via a wireless communication module. It continuously receives temporary garbage collection instructions from various channels. These instructions can be issued by the back-end dispatch system based on sudden garbage accumulation, reported by on-site staff through their operation terminals, or generated by citizens submitting garbage collection requests through the sanitation service platform. Upon receiving a temporary garbage collection instruction, the on-board central control system parses the instruction content, extracting core information about the temporary location, including its geographical coordinates, region, and collection request label. Simultaneously, it retrieves preset garbage truck operating radius data to verify the validity of the temporary location, determining whether it falls within the reasonable operating radius of the garbage truck and is a valid point for execution of the collection operation. If the verification passes, the temporary location is used as a pending collection node for further processing. If the verification fails, the on-board central control system generates an invalid instruction feedback message and simultaneously pushes it to the temporary instruction issuer, explaining the reason for the inability to execute.

[0050] Sub-step 2: Collect the current waste removal volume data and accurately calculate the removal ratio. After the vehicle-mounted central control system completes the validity check of the temporarily added location, it immediately activates the remaining amount detection device equipped with the waste bin to collect the actual loading data of the waste removal vehicle at present and obtain the current waste removal volume. This current waste removal volume is the actual amount of waste loaded by the waste removal vehicle, which can be characterized by the actual load detected by the weight sensor in the waste bin and the waste accumulation volume detected by the volume sensor, or can also be accurately collected by a combination of weight and volume; the vehicle-mounted central control system retrieves the pre-stored rated total waste removal volume of the waste removal vehicle, that is, the designed maximum loading capacity of the waste bin of the waste removal vehicle. This value is a fixed rated parameter, including the rated loading weight and the rated loading volume, which is consistent with the characterization dimension of the current waste removal volume; based on the collected current waste removal volume and the rated total waste removal volume, the vehicle-mounted central control system completes the calculation of the removal ratio through a pre-designed calculation formula. The specific formula is: removal ratio = current waste removal volume / rated total waste removal volume. The value range of this removal ratio is 0-1, and the value is distributed in a normalized manner. The smaller the removal ratio, the more sufficient the loading remaining amount of the waste bin of the waste removal vehicle is, and the more waste can be承接; the larger the removal ratio, the more tense the remaining loading space of the waste bin is, and the less waste can be承接. After the vehicle-mounted central control system completes the calculation, it will calibrate the data of the removal ratio to eliminate the微小误差detected by the sensor and ensure the accuracy of the value.

[0051] Sub-step 3: Based on the negative correlation adjustment interval of the removal ratio, complete the intelligent insertion of the route for the temporarily added location. The vehicle-mounted central control system clearly defines the interval between the temporarily added location and the current location of the waste removal vehicle. This interval is the sorting position of the temporarily added location in the waste removal route, that is, the number of waste removal position nodes between the route node corresponding to the current location of the waste removal vehicle. The smaller the value of the interval, the earlier the sorting of the temporarily added location in the waste removal route is, and the earlier the waste removal vehicle will arrive at this location to perform the waste removal operation; conversely, the later the sorting is, and the waste removal operation will be postponed.

[0052] The core of this step is the execution of the negative correlation adjustment logic. The vehicle-mounted central control system adaptively adjusts the above interval according to the calculated removal ratio according to the negative correlation relationship, and this adjustment logic is accurately adapted to the operation scenario where the waste amount of the temporarily added location is usually large. The specific adjustment rule is: the smaller the removal ratio, that is, the more sufficient the loading remaining amount of the waste removal vehicle is, the greater the adjustment amplitude of the interval, and the smaller the finally obtained interval value, the earlier the sorting of the temporarily added location in the waste removal route is, and the priority waste removal is realized; the larger the removal ratio, that is, the more tense the loading remaining amount of the waste removal vehicle is, the corresponding adjustment amplitude of the interval is reduced, and the larger the finally obtained interval value is, the later the sorting of the temporarily added location in the waste removal route is, and the waste removal is postponed.

[0053] To achieve this negative correlation adjustment, the vehicle-mounted central control system has a preset negative correlation adjustment formula and graded adjustment thresholds. It combines the current location of the waste collection vehicle, the node distribution of the initial waste collection route, and the geographical coordinates of any temporarily added locations to complete the quantitative calculation and position matching of the interval. For example, when the waste collection ratio is ≤0.3, it is determined that the waste collection vehicle has sufficient loading capacity, and the interval is adjusted to 0 or 1, meaning the temporarily added location is directly inserted into the next waste collection node after the current location of the waste collection vehicle, prioritizing the waste collection operation at that temporary location and making full use of the remaining loading space. When the waste collection ratio is between 0.3 and 0.7, it is determined that the waste collection vehicle has moderate loading capacity, and the interval is adjusted to 3-5 waste collection nodes after the current location in the initial route, inserting the temporarily added location into the later part of the route. When the waste collection ratio is ≥0.7, it is determined that the waste collection vehicle has limited loading capacity, and the interval is adjusted to after the last waste collection node of the initial route, or the information of the temporarily added location is temporarily stored and fed back to the background dispatch system, which then determines whether to postpone the execution of the temporary waste collection task.

[0054] After completing the negative correlation adjustment of the interval and calculating the optimal insertion position, the vehicle central control system will temporarily add a position to accurately insert into the corresponding position of the collection route, completing the dynamic update of the initial collection route. At the same time, the updated collection route will be simultaneously sent to the autonomous driving actuator and the vehicle display terminal. If the collection vehicle is driven manually, the driver will be prompted with the route update through the vehicle display terminal to ensure that the collection vehicle continues to travel according to the updated route.

[0055] In addition, if multiple temporary collection instructions are received, the vehicle-mounted central control system will estimate the amount of garbage and cluster the geographic coordinates of all temporary additional locations. Combined with the negative correlation adjustment logic of the collection ratio, the multiple temporary additional locations will be sorted and inserted in batches to ensure the rationality of route adjustments and the continuity of collection operations.

[0056] This embodiment achieves dynamic and intelligent adjustment of the waste collection route through the above steps. It can respond to various temporary waste collection instructions in real time and effectively insert temporary additional locations. It can also adaptively optimize the waste collection sequence of temporary locations based on the actual loading status of the waste collection vehicle and through negative correlation adjustment of the waste collection ratio.

[0057] In this embodiment of the application, to further optimize the rationality of the sorting of the waste collection routes and improve the utilization rate of the loading space of the waste collection vehicles, the following sub-steps are specifically included: Sub-step 1: Set the statistical period and collect and preprocess historical maximum material volume data. The vehicle-mounted central control system reads the preset statistical period parameters. This statistical period is a configurable time cycle that can be set by the background scheduling system according to the urban waste generation pattern and the frequency of collection operations, such as 1 month, 3 months, or 1 quarter, to ensure that the statistical data can cover the fluctuation characteristics of waste generation in different time periods. The vehicle-mounted central control system retrieves the maximum material volume data recorded by each collection location during each collection operation within the preset historical database through the wireless communication module. The historical database stores historical records such as the material volume, collection time, and equipment operating parameters of each collection location. The maximum material volume data is collected and uploaded in real time by sensors (distance sensor, weight sensor, or a combination of both) during the collection operation, and the data accuracy is consistent with the real-time collection standard.

[0058] To ensure data validity, the vehicle-mounted central control system preprocesses the collected maximum material quantity data: outliers (such as extreme data caused by sensor failure) are removed using the 3σ principle, and valid data that conforms to normal waste collection scenarios are retained; if the amount of valid data for a certain waste collection location within the statistical period is less than a preset threshold (such as less than 5 data), the average material quantity of similar waste collection locations in the same area is retrieved to supplement the data, or the valid data from the most recent statistical period is used to ensure that each waste collection location has sufficient sample data to support subsequent calculations.

[0059] Sub-step 2: Calculate the average material quantity and overall material quantity for each collection location. The loading control system quantifies the pre-processed effective maximum material quantity data. For each collection location, the arithmetic mean algorithm is used to calculate its corresponding average material quantity. The specific formula is: Average material quantity = (Sum of maximum material quantities within the statistical period) / Number of valid data entries. This value directly reflects the scale of waste generation at that collection location. The larger the value, the more urgent the waste loading demand at that location. For example, if the effective maximum material quantity data for collection location A within one month is [800kg, 750kg, 820kg, 780kg], then its average material quantity = (800+750+820+780) / 4 = 787.5kg.

[0060] After calculating the average material quantity at all collection locations, the vehicle-mounted central control system further calculates the comprehensive material quantity, which is the global average of the average material quantity at all collection locations. The specific formula is: Comprehensive material quantity = (sum of average material quantities at all collection locations) / total number of collection locations. This value is the benchmark value for measuring the material quantity level at all collection locations and is used to determine the relative level of the material quantity at a single collection location. For example, if the average material quantities at three collection locations are 787.5 kg, 650 kg, and 920 kg, then the comprehensive material quantity = (787.5 + 650 + 920) / 3 ≈ 785.83 kg.

[0061] Sub-step 3: Calculate the material comparison value and adjust the arrival attribute weight based on the positive correlation of the first adjustment parameter. For each collection location, the vehicle-mounted central control system calculates its material comparison value, which is the ratio of the average material quantity at that location to the overall global material quantity. The specific formula is: Material Comparison Value = Average Material Quantity at a Collection Location / Overall Material Quantity. The value range is (0, +∞). A ratio greater than 1 indicates that the material quantity at that location is higher than the global average, and the larger the ratio, the more material there is. A ratio less than 1 indicates that the material quantity at that location is lower than the global average, and the smaller the ratio, the less material there is. For example, the average material quantity at collection location A is 787.5 kg, and the overall material quantity is 785.83 kg, so its material comparison value is approximately 1.002, slightly higher than the global average. The average material quantity at collection location C is 920 kg, so its material comparison value is approximately 1.171, significantly higher than the global average.

[0062] The core of this step is the execution of the positive correlation adjustment logic. The vehicle-mounted central control system has a preset first adjustment parameter, which is a configurable weight coefficient (with a value range of 0.5-2.0). This parameter is used to quantify the adjustment range of the material comparison value on the arrival attribute weight. The value of the first adjustment parameter can be dynamically adjusted according to the actual collection needs (e.g., the parameter value can be appropriately increased in urban core areas to increase the priority weight of material quantity). Based on the material comparison value and the first adjustment parameter, the vehicle-mounted central control system dynamically corrects the arrival attribute weight corresponding to each collection location according to the positive correlation. The specific correction formula is: Corrected arrival attribute weight = basic weight × (1 + material comparison value × first adjustment parameter - first adjustment parameter), which can be simplified to: Corrected arrival attribute weight = basic weight × [1 + first adjustment parameter × (material comparison value - 1)].

[0063] The core effect of this positive correlation adjustment logic is that the larger the material comparison value, the higher the corrected arrival attribute weight, and the higher the priority of the collection location in route planning, making it easier to be prioritized at the front of the collection route. For example, if the first adjustment parameter is set to 1.0 and the base weight is 1.0, and the material comparison value of collection location C is 1.171, then its corrected arrival attribute weight = 1.0 × [1 + 1.0 × (1.171 - 1)] = 1.171; the material comparison value of collection location B is 650 / 785.83 ≈ 0.827, then its corrected arrival attribute weight = 1.0 × [1 + 1.0 × (0.827 - 1)] = 0.827. Through this adjustment, collection location C, with a material quantity significantly higher than the average level, obtains a higher arrival attribute weight and will be prioritized in route planning, achieving the sorting goal of "priority collection of multiple materials".

[0064] The corrected arrival attribute weights will be directly integrated into subsequent path planning algorithms (such as genetic algorithms and improved Dijkstra's algorithm) as core weight factors for path evaluation, participating in the calculation and ranking of the waste disposal routes. For example, in the improved Dijkstra's algorithm, high-weight arrival attributes make the corresponding waste disposal locations more likely to be included in the set of visited nodes, giving them priority as key nodes in the path. In the genetic algorithm, high-weight waste disposal locations will receive higher evaluation priority in path coding and fitness calculation, driving the algorithm to converge to the optimal route of "multiple materials first".

[0065] This embodiment fully utilizes the statistical value of historical material quantity data through the above steps, transforming the qualitative "multi-material priority" requirement into a quantitative weight adjustment logic. This not only improves the pertinence and scientific nature of route planning, but also accurately adapts to the loading space constraints of the waste collection vehicles.

[0066] In this embodiment of the application, in order to further improve the multi-dimensional planning system of the waste removal route from the perspective of operation time, based on the optimization of material quantity, the following sub-steps are specifically included: Sub-step 1: Set the statistical time period and collect and preprocess the longest historical time data. The vehicle-mounted central control system reads the preset statistical time period parameters. This statistical time period can be consistent with the statistical time period of the material quantity dimension, which facilitates the overall analysis and management of urban waste collection data. Alternatively, it can be independently set by the background scheduling system according to the time-specific characteristics of the collection operation, such as 7 days or 1 month, to ensure that the statistical data can truly reflect the actual operation time patterns of each collection location. The vehicle-mounted central control system retrieves the longest time data recorded for each collection location when completing a collection operation within the preset historical database through the wireless communication module. The longest time stored in the historical database is the total operation time of a single collection operation at each collection location, specifically covering the total time of operation links such as vehicle parking and positioning, robotic arm docking and grabbing, waste dumping, and equipment reset. This data is recorded in real time and automatically uploaded and stored by the vehicle-mounted central control system in each collection operation, with data collection accuracy down to the second level, perfectly matching the actual operation scenario.

[0067] To ensure the accuracy of subsequent calculations, the vehicle-mounted central control system performs standardized preprocessing on the collected longest time data: outliers in the data are removed using the 3σ principle, such as extreme values ​​of excessively long time caused by temporary equipment failures or unexpected on-site situations; if the amount of valid longest time data for a certain collection location within the statistical period is less than a preset threshold, time data from similar collection locations with the same administrative region and operation type are retrieved to supplement the data, or the valid time data from the most recent statistical period for that location is used, ensuring that each collection location has sufficient valid sample data to support subsequent quantitative calculations, thus guaranteeing the scientific nature of the weight adjustment of the time dimension from the source.

[0068] Sub-step 2: Calculate the average time and overall time for each collection location. The vehicle-mounted central control system performs an arithmetic average of the pre-processed valid historical longest time data, independently calculating the corresponding average time for each collection location. The specific calculation formula is: Average time = (Sum of the longest times within the statistical period) / Number of valid data entries. This value directly reflects the routine operation time characteristics of the corresponding collection location. The smaller the value, the higher the efficiency of the single-point collection operation at that location and the shorter the time; conversely, the larger the value, the more complex the operation process at that location and the longer the time.

[0069] After calculating the average time for all collection locations, the vehicle-mounted central control system further calculates the comprehensive time, which is the global average of the average time for all collection locations. The specific calculation formula is: Comprehensive time = (sum of average times for all collection locations) / total number of collection locations. This value is the global benchmark value for measuring the time level of single-point collection operations at all collection locations. It is used to determine the relative level of the operation time at a single collection location and provides a reference standard for the calculation of subsequent time comparison values.

[0070] Sub-step 3: Calculate the time comparison value and adjust the arrival attribute weight based on the negative correlation of the second adjustment parameter. The vehicle-mounted central control system calculates the time comparison value for each collection location, which is the ratio of the average time at that location to the overall global time. The specific formula is: Time Comparison Value = Average Material Quantity at a Collection Location / Overall Material Quantity. The value of this ratio ranges from (0, +∞). A ratio greater than 1 indicates that the single-point operation time at that collection location is higher than the global average, and the larger the ratio, the longer the relative time. A ratio less than 1 indicates that the single-point operation time at that location is lower than the global average, and the smaller the ratio, the shorter the relative time. This ratio will serve as the core quantitative basis for adjusting the weight of the time dimension.

[0071] The core of this step is the execution of the negative correlation adjustment logic. The vehicle-mounted central control system has a preset second adjustment parameter, which is a configurable weight adjustment coefficient with a value range of 0.3-1.5. It can be dynamically adjusted by the background scheduling system according to the overall operational efficiency requirements of urban waste collection. The value of this second adjustment parameter is less than the first adjustment parameter in the material quantity dimension, ensuring that the optimization of the time consumption dimension always conforms to the core requirement of utilizing the loading space of the waste collection vehicle. Based on the time consumption comparison value and the second adjustment parameter, the vehicle-mounted central control system dynamically corrects the basic weight of the arrival attribute corresponding to each waste collection location according to the negative correlation relationship. The specific correction formula is: Corrected arrival attribute weight = basic weight × [1 - second adjustment parameter × (time consumption comparison value - 1)].

[0072] The core effect of this negative correlation adjustment logic is as follows: the smaller the time comparison value, that is, the shorter the single-point operation time of the cleaning location, the higher the corrected arrival attribute weight, the higher the priority of the cleaning location in the route planning, and the easier it is to be prioritized for planning at the front of the cleaning route; the larger the time comparison value, that is, the longer the single-point operation time of the cleaning location, the lower the corrected arrival attribute weight, the lower the priority of the cleaning location in the route planning, and it will be planned later. For example, if the second adjustment parameter is set to 0.8 and the base weight is 1.0, and the time consumption comparison value of a certain collection location is 0.7, it indicates that its time consumption is significantly lower than the average level. Then, the corrected arrival attribute weight = 1.0 × [1 - 0.8 × (0.7 - 1)] = 1.24, and the weight is increased. If the time consumption comparison value of another collection location is 1.3, it indicates that its time consumption is significantly higher than the average level. Then, the corrected arrival attribute weight = 1.0 × [1 - 0.8 × (1.3 - 1)] = 0.76, and the weight is decreased. Through this quantitative adjustment, the sorting goal of prioritizing collection of locations with shorter time consumption is achieved.

[0073] The arrival attribute weights, corrected by the second adjustment parameter, will be superimposed with the weights corrected for the material quantity dimension to form the comprehensive arrival attribute weights for each collection location. This comprehensive weight will be directly integrated into subsequent collection route planning algorithms (genetic algorithm, improved Dijkstra algorithm), serving as a core weight factor in path evaluation and node selection during route calculation. For example, in the improved Dijkstra algorithm, collection locations with high comprehensive weights will be prioritized for inclusion in the visited node set; in the genetic algorithm, the comprehensive weight will serve as an important evaluation index for the fitness function, driving the algorithm to converge to the optimal collection route that prioritizes multiple materials and uses shorter processing times as a secondary factor, achieving synergistic optimization across both dimensions.

[0074] This embodiment transforms the qualitative optimization requirements of cleaning operation time consumption into quantitative weight adjustment logic through the above steps, fully explores the statistical value of historical operation time consumption data, realizes the optimization of the time consumption dimension of cleaning sorting, gives higher priority to cleaning locations with shorter operation time, effectively reduces the accumulation of time consumption in the entire cleaning operation, and avoids the overall cleaning rhythm being delayed due to prioritizing the handling of long time-consuming locations.

[0075] In this embodiment, when the material quantity dimension adjustment corresponding to the material comparison value and the operation time dimension adjustment corresponding to the time consumption comparison value are simultaneously applied to the arrival attribute weight, in order to ensure that the core needs and optimization goals of the entire waste collection process are orderly matched, the first adjustment parameter is explicitly set to be greater than the second adjustment parameter. Through the quantified parameter priority allocation, the core optimization of material quantity and the auxiliary optimization of time consumption are coordinated and unified, avoiding route planning deviations caused by conflicts in multi-dimensional adjustments. The specific implementation of this coordinated adjustment rule is as follows: I. Quantitative setting of adjustment parameters: The first and second adjustment parameters pre-stored in the vehicle's central control system are both configurable weighted coefficients. However, their value ranges and default values ​​strictly follow the priority rule of "first adjustment parameter > second adjustment parameter". The specific parameter setting standards are as follows: The first adjustment parameter (material quantity dimension) has a value range of 1.2-2.0, with a default value of 1.5. The design of this parameter fully considers the core constraints of waste collection and transportation; the limited loading space of the collection vehicle, by using a higher parameter value, ensures that the material quantity comparison value has a greater adjustment range on the weight of the arrival attribute, so that the optimization needs of the material quantity dimension take the lead in multi-dimensional collaboration.

[0076] The second adjustment parameter (task duration dimension): the value range is 0.3-1.0, and the default value is 0.6. The value of this parameter is lower than the first adjustment parameter, and the minimum difference between the two is not less than 0.5, i.e., 1.2-0.7=0.5. This ensures that the adjustment range of the task duration dimension is always less than that of the material quantity dimension, avoiding interference with the core optimization objective, while reserving a certain adjustment space to meet the time optimization requirements.

[0077] The above parameter values ​​can be dynamically adjusted by the back-end scheduling system according to the actual urban waste disposal scenario. However, the adjusted parameters must maintain the core rule of "first adjustment parameter > second adjustment parameter". For example, if the first adjustment parameter is adjusted to 1.8, the second adjustment parameter can be adjusted to a maximum of 1.0 (still satisfying 1.8 > 1.0); if the first adjustment parameter is adjusted to 1.2, the second adjustment parameter can only be adjusted to a maximum of 0.7 (ensuring a minimum difference of 0.5). The stability of the priority setting is ensured through quantitative constraints.

[0078] II. Collaborative Calculation Logic of Multi-Dimensional Weights: When both the material quantity dimension and the operation time dimension adjust the arrival attribute weight of the same collection location, the vehicle-mounted central control system adopts a calculation logic of "step-by-step correction + comprehensive superposition" to ensure that the priority rules are effectively implemented. The specific calculation steps are as follows: Basic weight initialization: The vehicle central control system retrieves the basic weight of the arrival attribute of the collection location (the default value is 1.0, which can be preset and adjusted according to the regional importance of the collection location) as the benchmark value for multi-dimensional adjustment.

[0079] Material quantity dimension priority correction: Based on the first adjustment parameter and the material comparison value, according to the preset positive correlation adjustment formula, the corrected weight 1 = basic weight × [1 + first adjustment parameter × (material comparison value - 1)] is used to first complete the weight correction of the material quantity dimension, thus obtaining the first stage correction weight. Since the first adjustment parameter has a higher value, the weight adjustment range in this stage will directly determine the core range of the attribute weight.

[0080] Secondary correction of the operation time dimension: Based on the first-stage correction weight, and using the second adjustment parameter and the time comparison value, the final correction weight is calculated according to the preset negative correlation adjustment formula: final correction weight = corrected weight 1 × [1 - second adjustment parameter × (time comparison value - 1)]. This completes the secondary correction of the operation time dimension, yielding the final comprehensive arrival attribute weight. This stage of adjustment prioritizes maintaining the core optimization results of the material quantity dimension, making only minor adjustments to the weights.

[0081] Example Explanation: Assume a certain collection point has a base weight of 1.0, a material comparison value of 1.3 (30% higher than the global average), and a time consumption comparison value of 1.2 (20% higher than the global average). Set the first adjustment parameter to 1.5 and the second adjustment parameter to 0.6 (satisfying 1.5 > 0.6): After material quantity dimension correction, the weight 1 = 1.0 × [1 + 1.5 × (1.3 - 1)] = 1.0 × 1.45 = 1.45; After secondary correction, the final weight of the task time dimension is 1.45 × [1 - 0.6 × (1.2 - 1)] = 1.45 × 0.88 = 1.276; Ultimately, the overall arrival attribute weight of the collection location was 1.276, which remains at a high level, ensuring its priority in route planning. At the same time, a slight adjustment to the time consumption dimension prevented the overall pace from being slowed down due to slightly longer operation time.

[0082] Conversely, if the material comparison value at a certain cleaning location is 0.8 (20% lower than the global average) and the time comparison value is 0.7 (30% lower than the global average): the weight after material quantity dimension correction is 1 = 1.0 × [1 + 1.5 × (0.8 - 1)] = 1.0 × 0.7 = 0.7; the final weight after secondary correction of the operation time dimension is 0.7 × [1 - 0.6 × (0.7 - 1)] = 0.7 × 1.18 = 0.826; Although the weight of the time consumption dimension has been increased, due to the core adjustment role of the material quantity dimension, the final weight is still less than 1.0, ensuring that its sorting priority is lower than the clearing location with more material quantity, which is in line with the core rule of "material quantity priority".

[0083] III. Priority settings, scenario adaptation, and technical support: The setting of "the first adjustment parameter being greater than the second adjustment parameter" precisely matches the actual needs of waste collection and transportation operations, and its technical guarantee logic is as follows: Prioritizing Core Constraints: The loading space of the waste collection vehicle is a key bottleneck to overall efficiency. Overemphasizing the shortest processing time by prioritizing areas with low material volumes can lead to situations where subsequent areas with higher material volumes cannot be cleared in one go due to insufficient loading space, resulting in multiple trips and increased overall collection time and energy consumption. By setting a high priority for the first adjustment parameter, we ensure that areas with higher material volumes occupy loading space first, thus avoiding this problem at its root.

[0084] Multi-dimensional demand balancing: While ensuring the core demand for material quantity, the non-zero value of the second adjustment parameter ensures that the optimization demand for the operation time dimension is not ignored. For multiple collection points with similar material quantities, the location with shorter operation time will receive higher weight through the adjustment of the time dimension, achieving refined optimization of "prioritizing shorter operation time under the same material quantity" and avoiding the one-sidedness of single-dimensional optimization.

[0085] Algorithm Co-adaptation: This priority setting is deeply integrated with the path planning logic of genetic algorithms and improved Dijkstra's algorithm. In the process of fitness calculation and node selection, the core influencing factor of the comprehensive arrival attribute weight is always the correction result of the material quantity dimension, while the operation time dimension is only used as an auxiliary factor to drive the algorithm to converge to the globally optimal route that "satisfies the core constraints and takes into account auxiliary needs".

[0086] This embodiment achieves a scientific weight allocation for the two optimization dimensions of material quantity and operation time by clearly setting the priority of the first adjustment parameter being greater than the second adjustment parameter. It firmly grasps the core constraints of garbage collection and transportation, ensures the utilization rate of loading space and prioritizes the collection of multiple material locations to avoid secondary round trips, and also takes into account the refined optimization needs of operation time. This makes the multi-dimensional collaborative route planning more in line with the actual operation scenario of urban garbage collection and transportation, and effectively improves the rationality, stability and overall operation efficiency of the collection route planning.

[0087] In this embodiment of the application, the step of identifying the posture of the cleaning target and matching it with the preset cleaning control logic specifically includes the following sub-steps: Sub-step 1: Precise positioning and acquisition of on-site image data of the waste collection target. After the waste collection vehicle arrives at the collection location according to the planned route, the onboard central control system activates the precise positioning control logic: combining the onboard positioning module (GPS / BeiDou dual-mode, positioning accuracy ±0.5m) and visual guidance sensors (high-definition cameras installed at the front and sides of the vehicle), it identifies the benchmark markers such as the marking lines and positioning stakes of the preset waste collection location in real time. Through the automatic driving actuator, it adjusts the vehicle's driving direction and speed, controlling the waste collection vehicle to smoothly drive into the preset waste collection location, ensuring that the relative positional deviation between the vehicle and the waste collection target (garbage bin) does not exceed ±10cm, providing a stable benchmark for subsequent image acquisition and robotic arm operations.

[0088] After the vehicle comes to a complete stop, the onboard central control system simultaneously activates the visual acquisition equipment, which includes three sets of high-definition industrial cameras (1920×1080 resolution, 30fps frame rate), deployed at different angles on the working side of the vehicle to achieve comprehensive coverage of the target area for cleaning. The cameras continuously acquire 3-5 frames of on-site image data, which are then transmitted to the onboard central control system for image preprocessing, including noise reduction (using Gaussian filtering algorithm), white balance correction, and image enhancement (histogram equalization). This process eliminates the impact of ambient light, dust, and other interference factors on image quality, ensuring the accuracy of subsequent box image recognition.

[0089] Sub-step two: Identify the bin graphic based on the bin template and quantify the number of bins. The vehicle-mounted central control system has a pre-set bin template database, which stores various bin specifications commonly used in urban waste collection, including bin shape feature data of different volumes and shapes such as 240L standard bins and 1200L transfer bins. Each template includes key recognition parameters such as bin outline dimensions, corner feature points, color thresholds, and identification patterns.

[0090] The vehicle-mounted central control system calls a template matching algorithm (using SIFT feature point matching combined with contour similarity comparison) to compare the preprocessed on-site image data with templates in the box template database one by one, and selects the regions that match the template features from the image, which are the box graphics. At the same time, the connected component analysis algorithm is used to remove duplicates and count the identified box graphics: a distance threshold (≥5cm) is set between adjacent box graphics. If the distance between two graphics is less than the threshold, it is determined to be an overlapping image of the same box and is counted as only 1; if the distance is greater than the threshold, it is determined to be an independent box. The cumulative count is used to obtain the number of boxes, and the counting accuracy error is ≤1.

[0091] Sub-step 3: Check the number of boxes and handle anomalies to accurately extract the grabbing structure and location.

[0092] (1) Box quantity verification and abnormal prompts: The vehicle-mounted central control system retrieves the corresponding configuration quantity (i.e., the number of garbage bins that should be placed at the current collection point, which is preset by the back-end dispatch system according to the amount of garbage generated in the area) and the quantity range (configuration quantity ± 1, which can be adjusted according to actual management needs) from the preset configuration database based on the current collection location code; it then calculates the difference between the currently identified bin quantity and the configuration quantity, i.e., bin difference = |identified bin quantity - configuration quantity|.

[0093] If the difference in the number of boxes is outside the preset range, such as when the configured quantity is 3, the quantity range is 2-4, and the number of boxes identified is 1 or 5, the vehicle-mounted central control system immediately generates an abnormal box quantity prompt. This prompt includes information such as the abnormality type (missing / excess boxes), the number of identified boxes, and the configured quantity, and is simultaneously pushed to the vehicle-mounted display terminal, the background dispatch system, and the on-site operation terminal. At the same time, the garbage truck enters the instruction waiting state, waiting for the background or on-site operators to issue subsequent instructions, such as replenishing boxes, clearing excess boxes and re-identifying, or ignoring the abnormality and continuing the operation.

[0094] If the difference between the bins is within the preset quantity range (including equal values), the bin quantity verification is deemed successful, and the process proceeds to the subsequent structure extraction process.

[0095] (2) Extraction of grabbing structure and grabbing location: The vehicle-mounted central control system performs refined feature extraction on each bin graphic, focusing on the bin's gripping structure (such as bin handles, bin hooks, and dedicated gripping interfaces): it extracts the outline edges of the bin graphic through edge detection algorithms (Canny operator), combines morphological operations (dilation, erosion) to highlight the feature areas of the gripping structure, and then uses key point detection algorithms (Harris corner detection) to locate the core coordinates of the gripping structure, such as the two ends of the handle and the load-bearing center point of the hook.

[0096] Based on the core coordinates of the gripping structure, the vehicle-mounted central control system calculates the gripping position, i.e., the precise coordinates of the robotic arm's execution end (gripper / suction cup) for docking. These coordinates are based on the vehicle coordinate system (X-axis along the vehicle's driving direction, Y-axis perpendicular to the driving direction, and Z-axis perpendicular to the ground), with a positioning accuracy of ±5mm. Simultaneously, according to the type of gripping structure (handle / hook / interface), the system retrieves the corresponding robotic arm docking posture parameters from the control logic database, such as the gripper opening and closing angle, suction cup negative pressure value, and docking force threshold, to ensure docking stability.

[0097] The vehicle-mounted central control system sends the gripping position coordinates and docking posture parameters to the robotic arm controller, which controls the multi-joint robotic arm to extend to the gripping position according to the preset motion trajectory (using trapezoidal speed planning to avoid sudden acceleration and deceleration) and accurately dock with the gripping structure. During the docking process, the force sensor at the end of the robotic arm provides real-time feedback on the contact force. If the contact force exceeds the threshold, the position is automatically fine-tuned to avoid hard collisions that could damage the equipment or the housing.

[0098] Sub-step four: Dynamically adjust the dumping speed based on the remaining material comparison value. After the robotic arm completes the docking of the grasping structure and fixes the container, the vehicle-mounted central control system initiates the remaining material detection logic: data is collected collaboratively by two types of sensors; one is a weight sensor installed at the docking end of the robotic arm, which detects the current total weight of the container in real time, including the remaining waste; the other is an ultrasonic sensor installed on top of the container, which detects the height of the accumulated waste inside the container; combining the data from the two sensors, the remaining material amount is calculated through a data fusion algorithm (weighted average, weight of the weight sensor 0.7, weight of the ultrasonic sensor 0.3), with a detection error ≤5%.

[0099] The vehicle-mounted central control system retrieves the total material volume of this type of container, i.e., the rated maximum loading capacity of the container, and stores it in the configuration database. It then calculates the remaining material comparison value = remaining material volume / total material volume. This ratio ranges from 0 to 1, which directly reflects the degree of remaining fullness of the garbage in the container.

[0100] Based on the remaining material comparison value, the vehicle-mounted central control system dynamically adjusts the tilting speed of the robotic arm according to a positive correlation adjustment rule. The specific adjustment logic is as follows: The preset tipping speed range is 0.5m / s-2.0m / s, and the speed adjustment step is 0.1m / s; When the residual material ratio is ≥0.8 (the bin is full of garbage), the dumping speed should be adjusted to 1.6m / s-2.0m / s to quickly complete the dumping and reduce operation time. When the residual material ratio is between 0.4 and 0.8 (meaning a moderate amount of waste in the bin), the dumping speed should be adjusted to 1.0 m / s to 1.5 m / s to balance dumping efficiency and stability. When the residual material ratio is less than 0.4 (the amount of waste in the bin is small), the dumping speed should be adjusted to 0.5m / s-0.9m / s to avoid material spillage caused by dumping too quickly due to the small amount of waste and the shift in the center of gravity.

[0101] During the dumping process, the vehicle's central control system monitors the dumping status of the garbage in real time through visual acquisition equipment. If material spillage is detected, the system will automatically reduce the dumping speed by 0.3 m / s based on the abnormal diffusion of the garbage outline in the image, until the spillage disappears. After dumping is completed, the system will maintain the dumping posture for 3-5 seconds to ensure that the garbage in the container is completely emptied. Then, the system will control the robotic arm to smoothly reset and place the container back to its original position or the preset recycling area.

[0102] This embodiment achieves refined control of the entire process of single-point cleaning operations, from placement, identification, and verification to grasping and dumping, through the above-mentioned detailed steps. By accurately extracting the grasping structure and position, the accuracy and safety of the robotic arm docking are improved. By dynamically adjusting the dumping speed through the remaining material comparison value, it adapts to the operational needs of different loading volumes, effectively avoiding the problems of material spillage and excessively long operation time, and comprehensively improving the standardization and intelligence level of single-point automated cleaning.

[0103] In this embodiment of the application, for scenarios where an abnormal situation occurs due to the target of the cleaning being deviated from its posture (such as tipping over or overturning), material spillage covering the gripping structure, or stains obscuring the surface of the container, resulting in the inability to extract an effective gripping structure based on the container graphic, an abnormal posture adaptation and virtual gripping position generation process is further added, which specifically includes the following sub-steps: Sub-step 1: Grabbing Structure Extraction Failure Judgment and Box Graphic Anomaly Prompt. When the vehicle central control system executes the grab structure and grab position extraction process, if it fails to identify a grab structure that meets preset features (such as the outline ratio of the handle, the geometry of the hook, and the feature point distribution of the interface) from the box graphic after calling edge detection, key point detection, and other algorithms three consecutive times (the number of times can be adjusted through backend configuration), it is judged that the grab structure extraction has failed, and the box graphic anomaly prompt mechanism is triggered at this time.

[0104] The anomaly alert includes specific information: the current collection location code, container type (if identified), cause of the anomaly (failed to extract the grabbing structure), and anomaly level (Level 2 warning). It is simultaneously pushed to the vehicle-mounted display terminal (audio-visual alarm + text prompt), the background dispatch system (system pop-up + SMS notification), and the on-site operation terminal to ensure that relevant personnel are aware of the anomaly status in real time. At the same time, the vehicle-mounted central control system suspends subsequent robotic arm movements and enters the anomaly handling process to avoid blind operation that could lead to equipment damage or safety accidents.

[0105] Sub-step two: Analyze the core features of the box graphic to determine the basic parameters for virtual grasping. After triggering an anomaly warning, the vehicle central control system initiates the box feature analysis logic under abnormal posture. Based on the identified box graphic, it accurately extracts three core basic parameters to provide data support for the virtual grasping position: Bin type determination: The bin template database is accessed, and the bin graphic's outline dimensions (length, width, height), corner curvature, color features, and other key information are compared with various preset bin templates in the database. A cosine similarity algorithm is used, and if the similarity threshold is ≥85%, the corresponding bin type is matched, such as a 240L standard garbage bin or a 1200L transfer garbage bin. If the similarity is below the threshold, the bin type is determined to be unknown, and the subsequent virtual process uses general parameters.

[0106] Graphic reference point calculation: Based on the pixel coordinates of the box graphic, the box outline is fitted using the minimum bounding rectangle method, and the geometric center of the rectangle is calculated as the graphic reference point (coordinates are represented as (X0, Y0), with the upper left corner of the image as the origin); at the same time, combined with the actual size of the box detected by the ultrasonic sensor, the pixel coordinates are converted into physical coordinates (X0', Y0') in the vehicle coordinate system. The conversion formula is: physical coordinates = pixel coordinates × pixel scale, pixel scale = actual size / size in the image, ensuring that the reference point position corresponds accurately with the actual box position.

[0107] Image orientation recognition: The direction of the major axis of the smallest bounding rectangle of the box image is detected by the Hough transform algorithm. The angle θ between the major axis and the X-axis (driving direction) of the vehicle coordinate system is calculated (the value range is 0°-180°). This angle is the image orientation, which is used to characterize the actual placement posture of the box (such as upright θ=0°, sideways θ=90°, tilted θ=45°, etc.).

[0108] After the above three core parameters are calculated, the vehicle central control system stores them in a temporary data cache area as the core input data for subsequent virtual location generation.

[0109] Sub-step 3: Virtually generate the grasping simulation position based on the box dataset. The vehicle central control system retrieves the preset box dataset, which is a structured database that stores the grasping structure parameters of various box types in the standard posture, including: the coordinate offset (ΔX, ΔY, ΔZ) of the standard grasping position relative to the geometric center of the box, the size specifications of the grasping structure, the grasping posture requirements, etc.; for unknown box types, the dataset presets general grasping offset parameters (based on the average size design of mainstream boxes).

[0110] The process of generating the virtual grasping simulation position uses coordinate transformation and posture adaptation logic, and the specific steps are as follows: Based on the determined bin type, the corresponding standard offsets (ΔX_std, ΔY_std, ΔZ_std) are retrieved from the bin dataset. The standard offsets are then rotated according to the angle θ between the graphic directions, calculated using the following formulas: ΔX_vir = ΔX_std × cosθ - ΔY_std × sinθ; ΔY_vir = ΔX_std × sinθ + ΔY_std × cosθ; ΔZ_vir = ΔZ_std, with the vertical direction unaffected by the horizontal posture. Here, ΔX_vir, ΔY_vir, and ΔZ_vir are virtual offsets adapted to the current bin posture. Using the physical coordinates (X0', Y0') of the graphic reference point as a baseline, the virtual offsets are superimposed to calculate the physical coordinates of the simulated grab position: X_vir = X0' + ΔX_vir; Y_vir = Y0' + ΔY_vir; Z_vir = Z0' + ΔZ_vir, where Z0' is the vertical coordinate of the reference point, detected by the height sensor. If the bin type is unknown, use a general offset (ΔX_gene, ΔY_gene, ΔZ_gene) and calculate the simulated position according to the same logic as above. At the same time, mark the "general virtual position" in the data to facilitate subsequent verification and traceability.

[0111] After generating the simulated gripping position, the vehicle-mounted central control system synchronously retrieves the virtual gripping posture parameters (such as gripper opening and closing angle, docking force threshold, and suction cup negative pressure range) corresponding to that position from the box data set to form a complete virtual gripping scheme.

[0112] Sub-step four: Verification of the simulated grasping position and determination of job execution / manual intervention. To ensure the reliability of the virtual grasping position, the vehicle-mounted central control system performs dual verification of the simulated grasping position. If the verification passes, the grasping operation is executed; if the verification fails, a manual intervention prompt is triggered. Position range verification: A preset position range threshold (±2cm, adjustable according to the positioning accuracy of the robotic arm) is set. This threshold represents the maximum allowable deviation between the standard grasping position (if extracted) and the virtual grasping position. If some valid grasping positions have already been extracted (e.g., some boxes are normal in a multi-box scenario), the distance difference between the simulated grasping position and the standard position is calculated. If the distance difference is ≤ the preset threshold, the position range verification passes. If no standard grasping position is extracted, the position range verification passes directly (the virtual position is within a reasonable range by default).

[0113] Docking feasibility verification: The vehicle-mounted central control system controls the robotic arm's execution end to move to a position 10cm above the simulated grasping position (safe distance). The vision sensor (miniature high-definition camera) at the end of the robotic arm collects local images in real time, while the distance sensor detects the actual distance to the surface of the box. If the vision sensor identifies a suspected grasping structure (such as an outline that matches preset features), and the distance detected by the distance sensor is ≤5cm (determined to be within the docking range), the docking feasibility verification passes; otherwise, the verification fails.

[0114] Based on the results of the double verification, perform the following operations: If both checks pass: the vehicle's central control system issues a command to control the robotic arm to move smoothly to the simulated grasping position according to the virtual grasping posture parameters and attempt to dock with the grasping structure; during the docking process, the force sensor provides real-time feedback on the contact force. If the contact force reaches a preset threshold (e.g., 5N) and remains stable (for 2 seconds), the docking is considered successful, and the subsequent tilting operation is performed according to the normal procedure; if the contact force is abnormal (e.g., exceeding 10N or continuously below 1N), the simulated grasping position is automatically fine-tuned (each fine-tuning is ±0.5cm, with a maximum of 3 fine-tunings) until the docking is successful.

[0115] If any verification fails, or if the docking still fails after three fine-tuning attempts: the vehicle-mounted central control system immediately generates a manual intervention prompt, which includes details of the anomaly (virtual location verification failure / docking failure), the coordinates of the captured simulated location, the container's attitude parameters, and a screenshot of the scene, and pushes them to the relevant terminals simultaneously; at the same time, it controls the robotic arm to reset to a safe position, and the garbage truck enters standby mode, waiting for the operator to handle the situation on-site (such as adjusting the container's attitude, clearing obstructions, manually specifying the grabbing position, etc.).

[0116] This embodiment constructs an automated grasping adaptation mechanism under abnormal postures through the above-described detailed steps: for scenarios where the grasping structure cannot be extracted, a precise grasping simulation position is first generated through box feature analysis, and then double verification is used to ensure operational safety, effectively solving the problem of automated operation interruption caused by abnormal situations such as posture deviation and material spillage; at the same time, manual intervention is triggered in a timely manner when verification fails, avoiding invalid operations and equipment damage, achieving a balance between automation efficiency and operational safety, greatly improving the adaptability of intelligent garbage collection vehicles to complex on-site environments, and further improving the technical system of fully automated garbage collection.

[0117] This application also discloses a fully automated waste collection and transportation control system for an intelligent waste collection vehicle, including a processor, wherein the processor executes the steps of the fully automated waste collection and transportation control method for an intelligent waste collection vehicle as described in any of the above embodiments.

[0118] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A fully automated waste collection and transportation control method for intelligent waste collection vehicles, characterized in that, Includes the following steps: Obtain a collection instruction, respond to the collection instruction by calling a collection vehicle and obtaining collection address information, obtain multiple collection locations from the collection address information, obtain the current location, calculate the collection route based on the current location and multiple collection locations, and proceed based on the collection route; After traversing the cleaning location, acquire on-site image data, identify the cleaning target from the on-site image data, and extract the amount of cleaning material based on preset sensors; Verify the collection target and the amount of materials to be collected. If the verification is successful, start single-point collection. Otherwise, display an on-site collection abnormality prompt and wait for instructions. Based on the on-site cleaning anomaly prompt, if an instruction to ignore the anomaly is received, single-point cleaning will be initiated; If a skip-site instruction is received, proceed to the next cleaning location; The steps for initiating single-point cleaning include: identifying the posture of the cleaning target; matching the preset cleaning control logic from the preset control logic database based on the posture of the cleaning target; controlling the preset robotic arm to grab the cleaning target and dump the cleaning contents according to the cleaning control logic, or controlling the cleaning inlet to open to receive the cleaning contents according to the cleaning control logic. After completing a single-point cleanup, the current remaining cleanup volume is identified, and the cleanup volume for the next cleanup location is obtained. If the remaining cleanup volume is greater than the cleanup volume for the next cleanup location, the vehicle proceeds to the next cleanup location; otherwise, it returns to the preset starting point and regenerates the cleanup route based on the unvisited cleanup location. After reaching the starting point, the vehicle proceeds based on the regenerated cleanup route, or the regenerated cleanup route is sent to other cleanup vehicles located within a preset range around the starting point that are in an idle state or have returned to the starting point.

2. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 1, characterized in that, The steps for calculating the waste removal route based on the current location and multiple waste removal locations include the following: The current location and multiple cleaning locations are encoded to generate an initial population. Each element in the initial population corresponds to a cleaning path. All cleaning paths start from the current location, pass through all cleaning locations, and then return to the current location. Construct a fitness function to calculate the total cleaning distance of each cleaning route based on the current location and multiple cleaning locations, calculate the route feasibility of each cleaning route based on a preset map database, calculate the cleaning priority of each cleaning route based on a preset priority database, and use the fitness function to calculate the route population based on the total cleaning distance, route feasibility and cleaning priority. The route population is iteratively updated through selection, crossover, and / or mutation operations, retaining the individual with the best fitness in the route population; Obtain the set iteration termination conditions, including the iteration count reaching a preset iteration threshold or the individual with the best fitness remaining unchanged, and output the corresponding individual as the cleaning route.

3. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 1, characterized in that, The steps for calculating the waste removal route based on the current location and multiple waste removal locations include the following: Initialize the distance matrix between the current location and multiple cleaning locations, and set the set of unvisited nodes and the set of visited nodes; Update the current location and the arrival attributes from the current location to each unvisited node in real time, and select nodes whose arrival attributes meet the preset filtering conditions to add to the visited set; Iteratively update the arrival attributes of unvisited nodes, obtain the congestion level corresponding to the unvisited nodes based on the preset map database, and adjust the weight of the arrival attributes of unvisited nodes according to the congestion level. After traversing all nodes in the unvisited node set, a cleaning route is generated from the current position through all cleaning locations based on the arrival attributes and corresponding weights of all nodes in the visited set. The overall congestion value is calculated based on the obtained congestion level. If the overall congestion value is less than the preset reference congestion value, the arrival attribute is the current location and the temporary distance from the current location to each unvisited node. Otherwise, the arrival attribute is the current location and the temporary arrival time from the current location to each unvisited node.

4. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 3, characterized in that, The steps for calculating the collection route based on the current location and multiple collection locations include the following: obtaining temporary instructions in real time, extracting temporary additional locations from the temporary instructions, and inserting the temporary additional locations into the collection route; obtaining the current collection volume loaded on the collection vehicle, and calculating the collection ratio based on the current collection volume and the total collection volume of the collection vehicle; and adjusting the interval between the temporary additional locations in the collection route and the current location of the collection vehicle based on the negative correlation of the collection ratio.

5. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 4, characterized in that, The method also includes the following steps: Within a preset statistical period, obtain the maximum material quantity at each collection location during collection, and calculate the average material quantity corresponding to each collection location based on the maximum material quantity. The average of the average material quantities at all collection locations is then calculated as the comprehensive material quantity. The ratio of average material quantity to total material quantity is calculated as the material comparison value. Based on the material comparison value, the weight of the arrival attribute corresponding to the cleaning location is adjusted positively with the first adjustment parameter.

6. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 4 or 5, characterized in that, The method also includes the following steps: Within a preset statistical period, obtain the longest time required to complete the collection at each collection location, and calculate the average time for each collection location based on the longest time. The average of the average times for all collection locations is then calculated as the comprehensive time. The ratio of average time to total time is calculated as the time comparison value. Based on the time comparison value, the weight of the arrival attribute corresponding to the collection location is adjusted negatively with the second adjustment parameter.

7. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 6, characterized in that, When both the material comparison value and the time comparison value are adjusted to adjust the weight of the arrival attribute, the first adjustment parameter is greater than the second adjustment parameter.

8. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 1, characterized in that, The step of identifying the target posture for waste removal and matching the preset waste removal control logic from the preset control logic database based on the target posture also includes the following sub-steps: Upon arrival at the collection location, the vehicle enters the pre-set collection bay and acquires on-site image data of the collection target. Based on a preset box template, box graphics are obtained from on-site image data, and the number of boxes is calculated based on the box graphics; Based on the collection location, the system retrieves the configured quantity and the corresponding quantity range from the preset configuration database. If the difference between the number of boxes and the configured quantity is calculated, and the difference is outside the quantity range, an abnormal box quantity prompt is generated. Otherwise, the system extracts the grasping structure and the corresponding grasping position based on the box image, guides the robotic arm to the grasping position, and docks with the grasping structure. After docking with the capture structure, the amount of remaining material in the target to be cleared is obtained in real time. The remaining material comparison value is calculated based on the remaining material amount and the total amount of material in the target to be cleared. The speed of dumping and clearing the contents is adjusted according to the positive correlation of the remaining material comparison value.

9. The fully automated waste collection and control method for intelligent waste collection vehicles according to claim 8, characterized in that, The steps of extracting the grasping structure and the corresponding grasping position based on the box graph also include the following sub-steps: If no grab structure is extracted from the box graph, an error message will be displayed regarding the box graph. The box type, reference point, and orientation are calculated based on the box graphic. The simulated grabbing position is then generated based on the reference point, orientation, and box type using a preset box dataset. If the simulated gripping position and the actual gripping position are outside the preset range or fail to connect successfully to the gripping structure, a prompt for manual intervention will be generated; otherwise, the robotic arm will be guided to the simulated gripping position and connect to the gripping structure.

10. A fully automated waste collection and transportation control system for an intelligent waste collection vehicle, characterized in that, The system includes a processor that performs the steps of the fully automated waste collection control method for intelligent waste collection vehicles as described in any one of claims 1-9.

Images