A method for optimizing a snow removal path of a complex terrain snow removal machine

Abstract

The application belongs to the field of path planning, and particularly relates to a snow sweeping path optimization method for a snow sweeper in complex terrain. The application discloses a snow sweeping path optimization method for a snow sweeper in complex terrain, aiming to solve the problems of high energy consumption of snow sweeping operation and poor adaptability to complex environment. The method comprises the following steps: constructing a snow distribution thermal map and a volume model by using a depth perception device; establishing an energy consumption cost evaluation model coupled with terrain slope, snow characteristics and motion parameters; performing non-uniform topological decomposition by using a weighted Voronoi subdivision algorithm based on the snow load weight, so that the unit density in the high load area is greater than that in the low load area; planning an optimal operation sequence and path with the minimum total energy consumption as the target and introducing a gravity auxiliary component; collecting motor current feedback signals in real time to inverse the snow density change, and dynamically correcting the model parameters and adjusting the operation state according to the snow density change. The application improves the operation efficiency and endurance, and reduces the mechanical loss.

Description

Technical Field

[0001] This application belongs to the field of path planning, specifically involving a method for optimizing the snow sweeping path of a snow sweeper in complex terrain. Background Technology

[0002] With the deep integration of automation technology and construction machinery, intelligent snow removal equipment plays a crucial role in ensuring winter transportation safety and urban operational efficiency. Traditional mechanized snow removal operations are gradually evolving towards intelligent systems, aiming to reduce human intervention and improve operational accuracy through automated path planning and operation control. In urban streets, airport runways, and large work sites, efficient path planning solutions are the core technological support for ensuring the stable operation of snow removal equipment under complex working conditions.

[0003] Among these, path optimization technology for complex terrain conditions is a key research focus for snow removal robots and automated snow removal machinery. This involves real-time modeling of the working environment, perception of snow distribution, and precise control of mechanical energy consumption. Such technologies typically require comprehensive consideration of the geographical features and snow conditions of the work area, using appropriate logical algorithms to guide the equipment in completing comprehensive snow removal operations. To achieve efficient snow removal in variable environments, the system needs to comprehensively evaluate environmental obstacles, terrain undulations, and workload to construct the optimal work sequence.

[0004] However, traditional path planning methods often employ pre-defined fixed-coverage algorithms, which are based on the ideal assumption of uniform snow distribution and struggle to cope with significant fluctuations in snow thickness caused by wind or building shadows in complex terrain. Existing planning logic often leads to drive system overload or drive wheel slippage when entering areas with high-density snow cover due to a sudden increase in operating resistance, while significant path redundancy exists in areas with thin snow, resulting in low overall energy efficiency. Furthermore, conventional solutions lack the ability to predict dynamic changes in mechanical loads; constant-parameter operating modes are prone to severe mechanical shocks when facing sudden load changes, accelerating fatigue wear in the transmission system. In addition, existing technologies fail to effectively integrate terrain slope information and cannot utilize gravity components to offset snow-push resistance, leading to a significant increase in energy consumption on slopes and severely restricting operational continuity in complex terrain. Therefore, a snow-plowing path optimization scheme for snowplows in complex terrain is desired. Summary of the Invention

[0005] The purpose of this invention is to provide a method for optimizing the snow removal path of a snowplow in complex terrain, which can effectively solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for optimizing the snow removal path of a snowplow in complex terrain includes the following specific steps: Step 1: Collect multi-dimensional feature data of the snow environment: Use a depth sensing device to perform real-time scanning of the working area within a preset range in front of the snowplow to obtain raw environmental information including snow surface height data and reference ground height data, and construct a snow distribution heat map and a two-dimensional gridded snow volume model of the working area based on coordinate transformation and data fusion processing; Step 2: Construct a multi-factor coupled energy consumption cost assessment model: Integrate terrain slope data, snow physical property parameters, and snowplow motion state parameters to establish an energy consumption evaluation model describing the functional relationship between power output and snow depth, snow density, driving speed, and terrain slope, used to assess the expected energy consumption of the snowplow under different working conditions; Step 3: Perform non-uniform topology decomposition of the working area: Based on the load weight distribution in the snow distribution heat map, apply additive... The Quanvironnoi partitioning algorithm dynamically divides the target operation area into multiple operation units with different geometric scales, making the unit density of high snow load areas greater than that of low snow load areas. Step 4 generates the optimal operation sequence and path planning instructions: with the minimization of total energy consumption as the objective function, the gravity auxiliary component generated by the terrain slope is introduced as the path optimization constraint. The heuristic search algorithm is used to plan the operation sequence across different operation units and the coverage path within each operation unit, and to generate motion control instructions for the underlying actuators. Step 5 implements real-time feedback adjustment and parameter correction of the operation status: during the operation of the snow sweeper, the current feedback signal of the drive motor is collected in real time. The real-time disturbance characteristics of snow density are inverted through the change in current, and the relevant parameters in the energy consumption cost assessment model are dynamically corrected accordingly. When the load exceeds the preset threshold, local path densification and speed coordination adjustment are immediately executed.

[0007] Preferably, in step 1, an environmental scan is performed using a depth camera mounted on the front support of the snow sweeper. The effective detection distance of the depth camera is set to a predetermined detection distance, and the sampling frequency is set to a preset sampling frequency. By emitting infrared structured light pulses and receiving reflected signals, the three-dimensional spatial coordinates of each sampling point relative to the sensor coordinate system are calculated.

[0008] Preferably, in step 1, the process of constructing a two-dimensional gridded snow volume model includes: projecting the collected three-dimensional point cloud data onto a horizontal reference plane, dividing the work area into subdivided grids with predetermined sizes, and obtaining the real-time snow thickness value of each grid unit by calculating the average height of the point cloud in each grid and subtracting the preset ground reference height.

[0009] Preferably, in step 2, the energy consumption cost assessment model is described by a function P equal to a function f, where P represents the main drive power output of the snowplow, and the variables include snow depth h, snow density estimate ρ, driving speed v, and terrain slope θ.

[0010] Preferably, in step 2, a correction coefficient is introduced for the terrain slope θ. When θ is greater than the reference slope value, i.e., when the system is in an uphill condition, the system compensates for the additional load generated by gravity resistance by increasing the power weighting factor. When θ is less than the reference slope value, i.e., when the system is in a downhill condition, the component of gravity in the driving direction is used to offset part of the snow pushing resistance generated by the snow depth h, thereby reducing the power prediction value.

[0011] Preferably, in step 2, the snow density estimate ρ is a variable initialized based on historical climate data of the work area. It is set to a preset initial density range during the system startup phase and is iteratively updated using sensor data as the work progresses.

[0012] Preferably, in step 3, when performing non-uniform topological decomposition using the weighted Veronoy algorithm, the snow load of each grid cell is used as the weight center point. The larger the load, the smaller the corresponding Veronoy cell area, thereby ensuring that more detailed operation instructions can be generated in areas with thick snow.

[0013] Preferably, in step 3, after completing the non-uniform topology decomposition, the system assigns a corresponding job attribute label to each job unit based on the geometric center coordinates and boundary features of each job unit. The job attribute label includes a low-speed high-torque mode label or a high-speed coverage mode label.

[0014] Preferably, in step 4, when planning the work sequence across units, a search strategy based on potential energy difference is adopted to prioritize planning the path from the work unit with higher terrain to the work unit with lower terrain.

[0015] Preferably, in step 4, for the location of the drainage outlet or snow disposal point in the work area, the algorithm assigns a very low potential energy weight to the specific coordinate to guide the path planning system to generate a shortest resistance path directly from the high-thickness snow area to the disposal point, so as to reduce the secondary transportation of snow.

[0016] Preferably, in step 4, during the internal path planning process, for units identified as low-speed, high-torque mode, a dense reciprocating path with a first predetermined overlap rate is adopted; for units identified as high-speed coverage mode, a wide path with a second predetermined overlap rate is adopted.

[0017] Preferably, in step 5, the real-time feedback adjustment process is implemented by a current sensor integrated inside the drive controller, and the sampling period is set to a predetermined sampling period.

[0018] Preferably, in step 5, by calculating the difference between the real-time current of the motor and the no-load current, and combining it with the current driving speed v, the real-time force on the front end of the snowplow is derived, and then the actual snow density parameter ρ at the current working position is calculated in reverse.

[0019] Preferably, in step 5, when the monitored real-time current value continues for a predetermined duration for a predetermined period of time exceeding a predetermined proportion of the safe rated current, the system determines that the physical hardness of the current snow exceeds the model's expectation, immediately triggers a local path correction program, reduces the width of a single snow push by a predetermined proportion, and simultaneously reduces the driving speed.

[0020] Preferably, the depth sensing device further includes a global positioning system receiver and an inertial measurement unit installed on the top of the vehicle body, used to acquire the absolute position and attitude information of the snowplow in the geographic coordinate system in real time, and the attitude angle measurement accuracy is better than the preset accuracy requirement.

[0021] Preferably, the update frequency of the snow distribution heat map is synchronized with the sampling frequency of the depth camera. The system uses a Kalman filter algorithm to smooth the multi-frame scan data to eliminate noise interference caused by light fluctuations or snow cover.

[0022] Preferably, in step 4, the objective function for minimizing total energy consumption also considers steering energy consumption. By limiting the radius of curvature of the path to be no less than a preset radius of curvature threshold, the drive loss and mechanical wear of the snowplow during the steering process are reduced.

[0023] Preferably, in step 1, environmental feature extraction further includes using image recognition algorithms to identify fixed obstacles in the work area, including curbs, fire hydrants and streetlights, and marking the corresponding grids as impassable areas in the gridded model.

[0024] Preferably, in step 3, when performing topology decomposition, a minimum unit area threshold is set to prevent the generation of overly fragmented work units, thereby ensuring the continuity of path planning and the smoothness of mechanical execution.

[0025] Preferably, in step 5, the corrected snow density parameter ρ is uploaded to the cloud server in real time to update the global snow characteristics database of the geographical area, providing prior parameter support for other snow removal equipment operating in the area.

[0026] Preferably, the method further includes performing a snow removal effect evaluation at the end of the operation, scanning the ground again with a depth camera to calculate the residual snow height, and automatically triggering a secondary snow removal path if the residual height is greater than a preset residual height threshold.

[0027] Preferably, the current feedback signal of the drive motor is processed by a low-pass filter of a preset order to filter out the instantaneous impact noise caused by uneven ground.

[0028] Preferably, the calculation formula for the gravity-assisted component fully considers the overall mass of the snowplow, including the unloaded mass and the dynamic mass generated by the real-time snow-pushing load.

[0029] Preferably, the heuristic search algorithm evaluates multiple alternative path branches in each iteration and selects the path with the lowest cost score as the final execution instruction.

[0030] Preferably, the snowplow at the front end of the snowplow is equipped with a pressure sensor array to help verify the accuracy of the snow distribution heat map, and the range of the pressure sensor is a preset force range.

[0031] Preferably, the energy consumption cost assessment model established in step 2 also introduces an ambient temperature correction term to compensate for the impact of decreased battery discharge efficiency and increased hydraulic oil viscosity on total energy consumption under low temperature conditions.

[0032] Preferably, in step 4, for extreme terrain where the absolute value of the slope θ is greater than a preset slope threshold, the planning algorithm will restrict its operation direction to only downhill, in order to ensure the driving safety and operational stability of the equipment.

[0033] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention utilizes depth sensing equipment and gridded modeling technology to achieve non-uniform perception of snow distribution within the working environment, breaking the assumption of uniform snow distribution in traditional algorithms. By dividing the working area into working units with different weights, the snowplow can perform high-torque, high-density snow removal actions in areas with thick snow, while adopting a high-speed, low-overlap working mode in areas with thin snow. This on-demand power allocation strategy effectively eliminates path redundancy, significantly improving overall work efficiency. Simultaneously, by constructing a multi-factor coupled energy consumption cost assessment model, the system can predict energy loss under different paths, thereby selecting the optimal path, significantly reducing the energy consumption per unit area in a single snow removal operation and significantly extending the equipment's operating time.

[0034] 2. This invention introduces a real-time feedback adjustment mechanism for motor current to construct a closed-loop control link from environmental perception to mechanical execution. The system can automatically adjust operating parameters according to changes in real-time load, avoiding motor overload, drive wheel slippage, and transmission system impact fatigue caused by sudden increases in resistance when the snowplow enters areas with high-density snow accumulation. Through local path densification and adaptive speed adjustment, the stress fluctuation range of the mechanical structure is significantly reduced, greatly reducing the failure rate of key transmission components, extending the service life of the entire machine, and lowering long-term maintenance costs.

[0035] 3. This invention fully utilizes the gravity-assisted component generated by terrain slope to convert natural potential energy into operational power. Through a path search algorithm guided by potential energy difference, the snowplow can prioritize downhill operations or path planning from high to low, maximizing the offsetting of snow-pushing resistance by the gravity component. In complex terrain conditions involving slopes, this invention can effectively reduce energy consumption compared to traditional methods. Furthermore, specific planning strategies for drainage outlets and disposal points reduce the number of secondary snow handling operations within the site, further optimizing the operational process. 4. The non-uniform topology decomposition and heuristic search algorithm employed in this invention enable the snowplow to make autonomous decisions in response to complex and changing environments. A real-time parameter correction mechanism allows the system to dynamically adjust its energy consumption prediction model based on different snow physical characteristics, ensuring planning accuracy under various climatic conditions. Through cloud data interaction and multi-sensor fusion, this invention not only improves the intelligent operation level of individual machines but also lays a solid technical foundation for the collaborative operation of large-scale automated snowplow fleets, realizing a leap from traditional mechanized operations to data-driven intelligent engineering machinery. Attached Figure Description

[0036] Figure 1 A schematic diagram of the overall technical solution for optimizing the snow removal path of snowplows in complex terrain; Figure 2 This is a schematic diagram illustrating the core principle of a multi-factor coupled energy consumption cost assessment model. Figure 3 This is a flowchart of the non-uniform topology decomposition logic based on a snow distribution heatmap; Figure 4 A schematic diagram of the multi-level interaction relationship and data flow between path planning instruction generation and real-time status feedback adjustment; Figure 5 This is a schematic diagram illustrating the principle of how the terrain slope correction factor affects energy consumption assessment. Detailed Implementation

[0037] Example 1 To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with the appendix. Figure 1 To be continued Figure 5 The present invention will be further described in detail below with reference to specific embodiments.

[0038] In the above-mentioned method for optimizing the snow removal path of a snowplow in complex terrain, it is implemented according to the following steps, as follows: Step 1, collecting multi-dimensional feature data of the snow-covered environment, is the foundational perception stage for the entire path planning process. Specifically, a depth camera mounted on the front support of the snowplow performs an environmental scan. The installation height of the depth camera is set between 1.5 meters and 2.2 meters above the ground, depending on the overall size of the snowplow, with a viewing angle set between 15 and 30 degrees to ensure coverage of a fan-shaped working area 5 to 10 meters in front of the snowplow. The effective detection distance of the depth camera is set to a predetermined detection distance, which is 10 meters in this embodiment, and its sampling frequency is set to a preset sampling frequency of 30Hz to 60Hz.

[0039] During the scanning process, the depth camera calculates the three-dimensional spatial coordinates of each sampling point relative to the sensor coordinate system by emitting infrared structured light pulses or receiving reflected signals using the time-of-flight principle. To acquire accurate geospatial data, the depth sensing device also includes a GPS receiver and an inertial measurement unit (IMU) mounted on the top of the vehicle, used to acquire the snowplow's absolute position and attitude information in the geographic coordinate system in real time. The IMU contains a three-axis gyroscope and a three-axis accelerometer, with an attitude angle measurement accuracy better than 0.1 degrees. The system uses a coordinate transformation matrix to transform the local coordinate system point cloud data acquired by the depth camera to a unified global coordinate system in real time.

[0040] After acquiring the raw point cloud data, the process of constructing a two-dimensional meshed snow volume model in step 1 is executed. Specifically, the system projects the collected three-dimensional point cloud data onto a horizontal reference plane, dividing the work area into subdivided grids with a predetermined size of 0.1 meters by 0.1 meters. By calculating the average point cloud height within each grid and subtracting the preset ground reference height, the real-time snow thickness value of each grid unit is obtained. To eliminate noise interference caused by vibrations from the snowplow or obstruction by flying snow, the update frequency of the snow distribution heatmap is synchronized with the sampling frequency of the depth camera. The system uses a Kalman filter algorithm to smooth the multi-frame scan data. At the same time, image recognition algorithms are used to identify fixed obstacles in the work area, including curbs, fire hydrants, and streetlights, and the corresponding grids are marked as impassable areas in the meshed model, with weights set to infinity.

[0041] In the above method, step 2, constructing a multi-factor coupled energy consumption cost assessment model, is crucial for realizing the core decision-making basis for path optimization. This model integrates terrain slope data, snow physical property parameters, and snowplow motion state parameters to establish an energy consumption evaluation model that describes the functional relationship between power output and snow depth, snow density, driving speed, and terrain slope. Specifically, the energy consumption cost assessment model is described by the following function: Here, P represents the main drive power output of the snowplow, and the variables include snow depth h, estimated snow density ρ, driving speed v, and terrain slope θ. In actual engineering calculations, the function f comprehensively considers snow-pushing resistance, driving friction resistance, and the gravity component generated by the slope.

[0042] For terrain slope θ, the model introduces a dynamic correction coefficient. When θ is greater than the reference slope value (i.e., 0 degrees) and the system is operating uphill, the system compensates for the additional load caused by gravity resistance by increasing the power weighting factor. When θ is less than the reference slope value and the system is operating downhill, the component of gravity in the travel direction is used to offset part of the snow-pushing resistance caused by snow depth h, thereby reducing the predicted power value. The calculation of the gravity-assisted component fully considers the overall mass of the snowplow, which consists of the unloaded mass and the dynamic mass generated by the real-time snow-pushing load. In addition, the model also introduces an ambient temperature correction term to compensate for the impact of decreased battery discharge efficiency and increased hydraulic oil viscosity on total energy consumption in low-temperature environments below -20 degrees Celsius. The snow density estimate ρ is set to a preset initial density range (e.g., 150 kg to 300 kg per cubic meter) based on historical climate data of the operating area during the system startup phase and is iteratively updated using sensor data as the operation progresses.

[0043] In the above method, step 3 involves performing non-uniform topological decomposition of the work area, aiming to transform the complex work environment into computable discrete units. Based on the load weight distribution in the snow distribution heatmap generated in step 1, a weighted Veronoi partitioning algorithm is applied to dynamically divide the target work area into multiple work units with different geometric scales. During algorithm execution, the system uses the snow load (i.e., the product of snow depth and density) of each grid unit as the weight center point. The larger the load, the smaller the corresponding Veronoi unit area, which ensures that more detailed work instructions can be generated in areas with thick snow cover.

[0044] To ensure the continuity of path planning and the smoothness of mechanical execution, a minimum unit area threshold is set during topology decomposition to prevent the generation of overly fragmented work units. After completing the non-uniform topology decomposition, the system assigns a corresponding work attribute label to each work unit based on its geometric center coordinates and boundary features. Work attribute labels include a low-speed, high-torque mode label (suitable for units with high snow load) or a high-speed coverage mode label (suitable for units with low snow load). In low-speed, high-torque mode, the system's preset execution parameters are that the motor speed is controlled at 40% to 60% of the rated speed, and the torque output is maintained above 80%. In high-speed coverage mode, the motor speed can be increased to over 90% of the rated speed to improve coverage efficiency.

[0045] In the above method, step 4, generating the optimal job sequence and path planning instructions, is the specific execution stage for energy-saving operations. The system uses minimizing total energy consumption as the objective function and introduces a gravity-assisted component generated by terrain slope as a path optimization constraint. A heuristic search algorithm is used to plan the job sequence across different job units. During the planning process, a search strategy guided by potential energy difference is adopted, prioritizing the planning of paths moving from higher-altitude job units to lower-altitude job units, thereby maximizing the utilization of gravity.

[0046] For drainage outlets or snow absorption points within the work area, the algorithm assigns a very low potential energy weight to these specific coordinates (similar to a sink in a gravitational field model), guiding the path planning system to generate the shortest resistance path directly from the thick snow accumulation area to the absorption point, thus reducing secondary snow transport within the site. In intra-unit path planning, for units identified as low-speed, high-torque modes, a dense reciprocating path with a first predetermined overlap rate (e.g., 30% to 50%) is used; for units identified as high-speed coverage modes, a wide path with a second predetermined overlap rate (e.g., 5% to 10%) is used. Simultaneously, the total energy consumption minimization objective function also considers turning energy consumption, reducing drive losses and mechanical wear during turning by limiting the path's radius of curvature to a preset radius of curvature threshold (e.g., 1.2 times the minimum turning radius of the snowplow). For extreme terrain with an absolute slope θ greater than a preset slope threshold (e.g., 15 degrees), the planning algorithm restricts its working direction to downhill only and outputs braking coordination commands to ensure equipment safety.

[0047] In the above method, step 5, which implements real-time feedback adjustment and parameter correction of the operating status, constitutes the closed-loop control link of the system. During the operation of the snow sweeper, the current sensor integrated inside the drive controller collects the current feedback signal of the drive motor in real time, with a sampling period set to 10 milliseconds. The current feedback signal is processed by a second-order low-pass filter of a preset order to filter out the instantaneous impact noise caused by uneven ground.

[0048] The system calculates the difference between the motor's real-time current and no-load current, and combines this with the current travel speed v to deduce the real-time force on the snowplow's tip. Based on this force, it uses a preset resistance model to calculate the actual snow density parameter ρ at the current working location. When the monitored real-time current value exceeds 1.2 times the preset safe rated current for more than 500 milliseconds, the system determines that the physical hardness of the current snow exceeds the model's expectations and immediately triggers a local path correction program. This program controls the actuator to reduce the width of each snowplow pass by a predetermined 20%, while simultaneously reducing the travel speed and increasing the path density in real time. The corrected snow density parameter ρ is uploaded to a cloud server in real time to update the global snow characteristics database for that geographical area.

[0049] Furthermore, in this embodiment, the method also includes performing a snow removal effect evaluation at the end of the operation. After completing the preset path, the snowplow scans the ground again using a depth camera to calculate the residual snow height. If the residual height in a certain area is greater than a preset residual height threshold of 3 cm, a secondary snow removal path is automatically triggered. This path will re-execute the planning logic of steps 2 to 4 based on the distribution of residual snow.

[0050] To aid in verifying the accuracy of the snow distribution heat map, a pressure sensor array is also equipped on the snowplow at the front of the snowplow. The pressure sensor's range is set from 0 to 5000 Newtons. During operation, the contact stress data acquired by the pressure sensor is cross-compared in real time with the volume model constructed in step 1. If the deviation between the two exceeds 15%, the online calibration program of the depth camera is triggered to ensure the reliability of the sensing data.

[0051] Example 2 Based on Example 1, this example further refines the application details in complex micro-topographical environments such as urban residential communities. To make the technical solution of the present invention more detailed, the parameter setting and processing logic for this specific application scenario are described below.

[0052] In step 1, during the acquisition of multi-dimensional feature data of the snow-covered environment, the detection distance of the depth camera was adjusted to 5 meters for narrow passages and corner areas within the residential community to improve near-field resolution. The point cloud data acquired by the depth camera was voxelized using a high-performance embedded processing unit on the vehicle, with the voxel grid size set to 0.05 meters cubic. The system uses a multi-sensor fusion algorithm, combining GPS signals with attitude data provided by the inertial measurement unit, to maintain a dynamic local map of 200 meters by 200 meters in memory.

[0053] When constructing a two-dimensional meshed snow volume model, the system particularly enhances edge detection for low obstacles such as curbs. The snow height gradient is calculated using the Sobel operator or the Canny edge detection algorithm. When a sudden height change is detected and matches the geometric characteristics of a curb, the mesh is automatically marked as a boundary protection zone. In these areas, path planning reserves a 0.15-meter avoidance margin to prevent snowplows from making hard contact with stone structures.

[0054] In step 2, the energy consumption cost assessment model is constructed by introducing a terrain roughness factor to address common micro-topography features in the community, such as speed bumps and manhole covers. This factor is estimated using the root mean square value of vertical acceleration measured by the inertial measurement unit during driving. As the roughness factor increases, the model automatically increases the prediction weight of driving friction resistance. The specific form of the energy consumption assessment model, based on Example 1, adds a term to describe the dynamic losses caused by frequent start-stop cycles. in, Here, k represents the steady-state operating power, m represents the loss coefficient, m represents the total mass of the machine and its load, and a represents the instantaneous acceleration. This allows the system to more accurately estimate energy reserves when performing frequent reversing operations in confined areas.

[0055] In step 3, when performing non-uniform topology decomposition, the system employs an adaptive weighting mechanism. In pedestrian walkways within residential areas, the system incorporates "operational smoothness" as another weighting dimension into the weighted Veronoi partitioning algorithm. This means that even in areas with thin snow cover, smaller work units are generated to ensure that the cleared ground is free of residual ice. The system also adds a "quiet operation mode" label to the generated operation attribute tags. In this mode, the system limits the maximum current change rate of the drive motor to reduce mechanical noise.

[0056] In step 4, the optimal work sequence generation stage, due to the dense distribution of drainage pipe networks within the residential area, the system imports the precise latitude and longitude coordinates of drainage wells from the property management database. A heuristic search algorithm sets these drainage well coordinates as "energy efficiency traps," meaning the cost score of snow-plowing paths near these points is significantly reduced. The algorithm calculates the optimal slope path from any work unit to the nearest drainage well. During path generation, considering the mobility requirements of snowplows within the residential area, the system uses multi-order Bézier curves to smooth the planned polygonal path, ensuring that the underlying actuators have continuous first and second derivatives during movement, avoiding severe vibrations in the mechanical structure at turns.

[0057] In the real-time feedback adjustment in step 5, the system incorporates current surge monitoring logic to detect potential ice layers (snow density ρ sudden changes). If the motor current rises at a rate exceeding 200% per second of the rated current within 50 milliseconds, the system determines that the blade has encountered ice or a foreign object. In this case, the real-time feedback adjustment program issues an "emergency retreat" command, causing the snowplow to retreat 0.5 meters and re-scanning the obstacle ahead with a high-resolution depth camera. If the ice is confirmed to be solid, the local path encryption program from step 5 is automatically invoked, switching to a shallow cutting depth, high-frequency ice-breaking operation mode.

[0058] Furthermore, the corrected snow accumulation characteristics parameters are synchronized in real time to the property management cloud platform via the vehicle's 4G / 5G communication module. If multiple snowplows operate collaboratively, the snowplows at the rear can adjust their preset power output values ​​in advance based on the density parameters ρ and terrain slope θ uploaded in real time by the machines at the front, achieving clustered energy efficiency optimization. During the evaluation phase after the operation, the system will also generate a digital operation report, recording the initial snow thickness, residual height after clearing, and total power consumption for each operation unit, providing data support for subsequent operation strategies.

[0059] Example 3 Based on the above embodiments, this embodiment describes a specific implementation method for use in large open-air parking lots or airport feeder runways, focusing on how to achieve ultra-long range and high coverage in large open areas through this invention.

[0060] In step 1, due to the vast operating area, in addition to the front-end camera, the depth sensing device also includes an ultrasonic radar array around the vehicle body to assist in detecting the operating environment within a 360-degree range. The snow distribution heatmap was constructed using multi-resolution mesh technology: a high-resolution mesh of 0.1 meters was used within 20 meters of the snowplow's current location, while a coarse mesh of 0.5 meters was used in the far-field region from 50 to 100 meters. This multi-scale modeling strategy effectively reduced computational overhead in large-scale scenarios while ensuring accuracy in near-field operations.

[0061] In step 2, the energy consumption cost modeling stage, considering the impact of wind on energy consumption in large areas of flat terrain, an aerodynamic drag term is introduced into the model. The system obtains wind speed and direction data from local weather stations online and combines this data with the snowplow's travel vector to calculate the impact of wind resistance on the total power output P. When operating with the wind, the system increases the travel speed v to reduce motor power consumption by utilizing the wind component; when operating against the wind, the speed is appropriately reduced to maintain energy consumption within the optimal range.

[0062] In step 3, the non-uniform topology decomposition, since runways or parking lots typically have a significant drainage slope (e.g., a 1% to 2% cross slope), the weighted Veronoi algorithm forces the boundary lines to be aligned with the terrain contour lines to some extent when calculating the cell boundaries. This results in work cells whose major axes are usually parallel to the contour lines, facilitating long-distance straight-line snow-plowing actions by the snowplow within the cell and reducing ineffective empty-running transfers.

[0063] In step 4, during path sequence generation, for scenarios with extremely high time sensitivity, such as airport runways, the objective function is adjusted from purely "minimizing total energy consumption" to "energy- and time-weighted optimality." The heuristic search algorithm calculates a comprehensive cost factor when evaluating path branches: Where E represents the expected energy consumption and T represents the operation time. and This is an adjustable weighting coefficient. The system will automatically increase it when the snow removal window is short. At this point, the planned path will be straighter. Although the utilization rate of the gravity component may decrease slightly, the overall work efficiency will be guaranteed.

[0064] In the closed-loop control of step 5, to address the temperature rise issue caused by the continuous high-load operation of the motor during large-area operations, the real-time feedback adjustment system adds temperature control feedback. When the temperature of the drive motor winding exceeds the 85-degree Celsius warning line, the system automatically sends a restriction command to the path planning module. In the following work units, it prioritizes low-load areas (thin snow areas) for "cooling operations," reducing the single snow-pushing load to allow the mechanical system to cool down naturally, thus avoiding work interruptions caused by triggering hardware thermal protection.

[0065] Meanwhile, for large-area snow disposal points, the system planned a "relay" snow-plowing path. The first snowplow pushes the snow to the edge of the unit, and then the second snowplow takes over, pushing it to the final storage area. This multi-machine collaborative path optimization relies on the corrected parameter ρ and heat map data uploaded to the cloud in real time in step 5. The system performs global coordination in the cloud to ensure that the energy consumption model of each snowplow is synchronized in real time.

[0066] During the post-operation evaluation phase, in addition to residual height detection, the system also calculates the path overlap rate using GPS trajectory data. If an abnormally high overlap rate is detected, it indicates a potential blind spot or algorithm failure in that area. The system automatically records the coordinates of this point in the "Difficult Area Library" for subsequent manual maintenance reference. In this way, this method not only optimizes energy efficiency in large-scale scenarios but also improves the system's robustness under long-term, high-intensity operations.

[0067] In any of the above embodiments, when the snowplow executes path planning commands, its underlying actuators (including the steering drive, travel motor controller, and blade lifting hydraulic valve) all interact with each other at high speed via the CAN bus. The CAN bus baud rate is set to 500kbps, and the transmission period of critical control frames is 5 milliseconds, ensuring that the response delay from the path planning layer to the mechanical execution layer is less than 20 milliseconds.

[0068] When handling the gravity-assisted component generated by the terrain slope θ, the system maintains a high-precision digital elevation model (DEM). This model is loaded with survey data before operation or dynamically generated during operation using the snowplow's own GPS and IMU. When the snowplow is traveling on a cross slope (the slope direction is perpendicular to the travel direction), the path planning algorithm automatically calculates the risk of sideslip. If the sideslip risk value exceeds a threshold, the system will forcibly correct the path, causing the snowplow to travel at a certain crabbing angle. The lateral support force of the tracks or wheels is used to counteract the lateral component of gravity, ensuring that the working trajectory does not deviate from the preset path.

[0069] Furthermore, the process of correcting the snow density parameter ρ in step 5 also includes a logic for determining the physical state of the snow. The system determines whether the current snow is powdery snow, wet snow, or icy snow based on the slope of the force-velocity curve. For wet snow, due to its high viscosity, an additional 5% power reserve is added during path planning in step 4 to prevent shutdown due to excessive snow accumulation at the end of the snow-pushing path.

[0070] In summary, this invention achieves intelligent path planning in complex, non-uniform snow-covered environments through deep coupling of deep perception and multi-factor energy consumption modeling. From multi-dimensional environmental feature extraction to topology decomposition based on load weights, to heuristic path search utilizing terrain energy, and real-time closed-loop correction based on motor current feedback, each step closely revolves around the core engineering goals of "efficiency improvement" and "energy efficiency optimization." This not only enhances the snowplow's operational capabilities in harsh terrain but also provides solid technical support for its application in the fields of automated and unmanned engineering machinery.

[0071] In practical applications, users can set task priorities via mobile terminals or a central control interface, such as "energy saving priority," "speed priority," or "clearance priority." The system will dynamically adjust various coefficients in the energy consumption cost assessment model, as well as overlap rate and curvature constraint parameters in path planning, based on the user's selection. This high degree of flexibility and adaptability enables the invention to be widely applied in various complex terrain scenarios such as urban roads, residential areas, airports, and parking lots, demonstrating extremely high practical value and broad market prospects.

[0072] To further ensure system stability, at the software implementation level, all sensing data processing and path planning algorithms run on an industrial-grade controller with hardware redundancy. The controller employs dual-core lockstep technology; once a computational logic anomaly is detected, it immediately switches to a safe operating mode (i.e., stops in place or performs a straight-line backoff) and sends an error log to the management backend. At the data communication level, all critical sensor data and control commands undergo redundancy check (CRC) to prevent erroneous commands in environments with strong electromagnetic interference (such as near large power distribution facilities or communication towers).

[0073] Regarding the coordination of the mechanical structure, the snowplow's blade angle can be adjusted in real time according to path planning instructions. The path instructions generated in step 4 include a control sequence for the blade angle. For example, when pushing snow downhill, the blade angle is adjusted to the maximum snow-throwing angle to utilize gravity to assist in rapid snow removal; when transferring between units, the blade automatically rises to a safe height to reduce drag during idle driving. These subtle mechanical coordination actions are all uniformly driven by the motion control instruction set generated in step 4, ensuring the continuity of the entire machine's operation.

[0074] Through the aforementioned series of high-precision sensing, dynamic modeling, non-uniform spatial decomposition, and closed-loop feedback adjustment, this invention effectively solves the technical bottlenecks of traditional snow removal path planning, such as poor adaptability in complex terrain, high energy consumption, and large mechanical losses, providing a complete and feasible technical solution for intelligent snow removal operations. The basic principles, main features, and advantages of this invention have been shown and described above. Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the snow removal path of a snowplow in complex terrain, characterized in that, Includes the following steps: The depth sensing device installed on the front bracket of the snow sweeper performs real-time scanning of the work area to obtain raw environmental information including snow surface height data and reference ground height data. Based on coordinate transformation and multi-source data fusion processing, a snow distribution heat map and a two-dimensional gridded snow volume model of the work area are constructed. By integrating terrain slope data, snow physical property parameters, and snowplow motion state parameters, an energy consumption cost assessment model is established to describe the functional relationship between main drive power output and snow depth, snow density estimate, driving speed, and terrain slope, in order to assess the expected energy consumption of the snowplow under preset working conditions. Based on the load weight distribution in the snow distribution heatmap, the weighted Veronoi partitioning algorithm is applied to dynamically divide the target operation area into multiple discrete operation units with different geometric scales, so that the operation unit distribution density in the high snow load area is greater than the operation unit distribution density in the low snow load area. With the goal of minimizing total energy consumption, a gravity-assisted component generated by terrain slope is introduced as a path optimization constraint. A heuristic search algorithm is used to plan the global operation sequence across different work units and the coverage path within each work unit, and to generate motion control commands for the underlying actuators. During the operation of the snow sweeper, the current feedback signal of the drive motor is collected in real time. The real-time disturbance characteristics of snow density are inverted by the change in current, and the relevant parameters in the energy consumption cost assessment model are dynamically corrected accordingly. When the real-time load is detected to exceed the preset threshold, the local path encryption program and speed coordination adjustment strategy are executed.

2. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The steps for obtaining raw environmental information using a depth sensing device include: during the operation of the snowplow, performing an environmental scan using a depth camera installed at a predetermined height on the front support of the snowplow, wherein the depth camera covers the fan-shaped working area in front of the snowplow at a predetermined top-down angle, and the effective detection distance of the depth camera is set to a predetermined detection distance and the sampling frequency is set to a preset sampling frequency range. By emitting infrared structured light pulses and receiving reflected signals, the three-dimensional spatial coordinates of each sampling point relative to the local coordinate system of the sensor are calculated. At the same time, using a global positioning system receiver installed on the top of the vehicle body and an inertial measurement unit containing a three-axis gyroscope and a three-axis accelerometer, the absolute position information and attitude angle information of the snow sweeper in the geographic coordinate system are obtained in real time. The attitude angle measurement accuracy is better than the preset accuracy requirement. Using a preset coordinate transformation matrix, the point cloud data acquired by the depth camera in the local coordinate system is transformed in real time to a unified world coordinate system, thereby obtaining the geospatial location of each sampling point in the working area.

3. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The process of constructing a two-dimensional gridded snow volume model includes: projecting the collected three-dimensional point cloud data onto a horizontal reference plane, dividing the work area into subdivided grids with predetermined sizes, and calculating the average height of the point cloud in each subdivided grid and subtracting the preset ground reference height to obtain the real-time snow thickness value of each grid cell. The update frequency of the snow distribution heat map is synchronized with the sampling frequency of the depth camera. The Kalman filter algorithm is used to smooth the multi-frame scan data to eliminate noise interference caused by light fluctuations or snow cover. Image recognition algorithms are used to identify fixed obstacles in the work area, including curbs, fire hydrants and streetlights. The corresponding grids are marked as impassable areas in the gridded model, and a preset maximum weight value is assigned to the corresponding grids.

4. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The construction and correction process of the energy consumption cost assessment model includes: when calculating the main drive power output of the snowplow, the snow pushing resistance, driving friction resistance and gravity component generated by the slope are comprehensively considered, and a dynamic correction coefficient is introduced for the terrain slope; when the terrain slope is greater than the reference slope value and the working condition is uphill, the additional load generated by gravity resistance is compensated by increasing the power weight factor. When the terrain slope is less than the reference slope value and the vehicle is in a downhill condition, the component of gravity in the driving direction is used to offset part of the snow pushing resistance caused by the snow depth, thereby reducing the power prediction value. The calculation process of the gravity-assisted component incorporates the overall mass of the snowplow, which includes the unloaded mass and the dynamic mass generated by the real-time snow-pushing load. An ambient temperature correction term is introduced into the energy consumption cost assessment model to compensate for the impact of decreased battery discharge efficiency and increased hydraulic oil viscosity on total energy consumption under preset low temperature conditions.

5. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The process of performing non-uniform topological decomposition of the work area includes: when performing decomposition using the weighted Veronoy partitioning algorithm, the snow load of each grid cell is used as the weight center point. The snow load is determined by the product of snow depth and density. The larger the load value, the smaller the corresponding Veronoy cell area. During the topology decomposition process, the number of fragmented work units is limited by setting a minimum unit area threshold, thereby ensuring the continuity of path planning and the smoothness of mechanical execution. After completing the non-uniform topology decomposition, each work unit is assigned a corresponding work attribute label based on its geometric center coordinates and boundary features. The work attribute labels include low-speed high-torque mode labels or high-speed coverage mode labels.

6. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The process of planning the global work sequence across different work units includes: adopting a search strategy guided by potential energy difference, prioritizing the planning of paths from work units with higher elevations to work units with lower elevations, thereby utilizing natural potential energy to reduce energy consumption; For the locations of drainage outlets or snow dissipation points within the work area, by assigning a preset extremely low potential energy weight to these specific coordinates, the path planning system is guided to generate the shortest resistance path from the high-thickness snow area to the dissipation point, thereby reducing the secondary transport of snow. During the planning process, multiple alternative path branches are evaluated in real time, and the path with the lowest cost score is selected as the final execution instruction. The cost score is determined by a weighted combination of expected energy consumption and operation time.

7. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 5, characterized in that, The coverage path planning process within each work unit includes: for work units identified as low-speed, high-torque mode, a dense reciprocating path with a first predetermined overlap rate is adopted, where the first predetermined overlap rate corresponds to the working condition where the motor speed is within a first preset range and the torque output is within a second preset range. For work units identified as high-speed coverage mode, a wide path with a second predetermined overlap rate is adopted, where the second predetermined overlap rate is less than the first predetermined overlap rate, and the motor speed is within a third preset range. During the internal path planning process, the curvature radius of the path is limited to be no less than a preset curvature radius threshold, thereby reducing the drive loss and mechanical wear of the snow sweeper during the turning process. For extreme terrains where the absolute value of the slope is greater than the preset slope threshold, the operation direction is restricted to downhill, and a braking coordination command is output.

8. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The process of implementing real-time feedback adjustment of the working status includes: using a current sensor integrated inside the drive controller to collect the current feedback signal of the drive motor at a predetermined sampling period, and inputting the current feedback signal into a low-pass filter of a preset order for processing, thereby filtering out the instantaneous impact noise caused by uneven ground. By calculating the difference between the real-time current and the no-load current of the motor, and combining it with the current driving speed, the real-time force on the snowplow at the front of the snow sweeper is derived, and then the actual snow density parameters at the current working position are calculated in reverse using the preset resistance model. The actual snow density parameter is set to a preset initial density range based on historical climate data during the system startup phase and is iteratively updated as the operation progresses.

9. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 8, characterized in that, The process of implementing parameter correction and local path adjustment includes: when the monitored real-time current value lasts for a longer period of time and exceeds the preset safe rated current ratio, it is determined that the physical hardness of the current snow exceeds the model expectation, and the local path correction procedure is immediately triggered. The local path correction program reduces the width of a single snowplowing operation by a predetermined ratio by controlling the actuator, while simultaneously reducing the driving speed and increasing the path distribution density. The corrected actual snow density parameters are uploaded to the cloud server in real time to update the global snow cover characteristics database for the geographical area, providing prior parameter support for other snow removal equipment operating in the area.

10. The method for optimizing the snow removal path of a snowplow in complex terrain according to claim 1, characterized in that, The method also includes a snow removal effect evaluation and multi-sensor verification process: at the end of the operation, the ground after the operation is scanned again using a depth camera to calculate the residual snow height value. If the residual snow height value is greater than the preset residual height threshold, a secondary cleaning path is automatically triggered. Using an array of pressure sensors installed on the snowplow at the front of the snowplow, contact stress data during the operation is acquired, and the contact stress data is cross-compared with the snow volume model in real time. When the deviation between the contact stress data and the snow volume model exceeds a preset deviation threshold, the online calibration program of the depth camera is triggered to ensure the reliability of the sensing data. The underlying actuator of the snow sweeper performs data interaction through the controller local area network bus. The baud rate of the bus is set to a preset baud rate, and the transmission period of the key control frame is set to a preset period.

Images