AGV (Automatic Guided Vehicle) two-way passage obstacle avoidance track generation method and system integrated with road condition perception

Abstract

The invention provides an AGV two-way passage obstacle avoidance track generation method and system fused with road condition perception, and relates to the technical field of track planning, and the method comprises the steps: scanning the ground condition in real time through a vehicle-mounted laser radar, recognizing the driving interference, fitting the safety speed, calculating the dynamic safety distance based on the real-time safety speed of two AGVs, and calculating the real-time safety speed of the two AGVs; if the conflict distance is smaller than the safe distance, triggering obstacle avoidance data acquisition, calculating asymmetric safe turning capability, positioning a shared avoidance safe area, generating a simulated avoidance route, performing role allocation according to energy consumption prediction, determining an avoidance vehicle, controlling a driving track of the avoidance vehicle, decelerating a non-avoidance vehicle, and ensuring safe avoidance of the two AGVs. And the vehicle leaves the shared avoidance area. According to the invention, the technical problems that the AGV is difficult to flexibly deal with the obstacle avoidance in a complex environment and the driving efficiency and the safety are reduced due to the fact that the two-way passing obstacle avoidance is realized by only depending on static path planning in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of trajectory planning technology, specifically to a method and system for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception. Background Technology

[0002] AGVs (Automated Guided Vehicles) play a crucial role in warehousing, assembly lines, and other similar scenarios. In workshop environments such as assembly plants, bidirectional movement of AGVs is particularly important because these environments often involve narrow passages and complex traffic conditions. AGVs need to be able to dynamically adjust their travel paths based on real-time road conditions and the surrounding environment to avoid collisions and ensure safe and efficient operation. However, current technologies rely solely on static path planning and preset turning radii for obstacle avoidance. This means that AGVs may not be able to adjust their travel strategies in time when facing complex ground conditions. This makes their path planning inflexible, unable to effectively avoid new obstacles or cope with environmental changes, increasing the risk of collisions and consequently reducing travel efficiency and safety. Summary of the Invention

[0003] To address the aforementioned deficiencies or improvement needs of existing technologies, this application provides a method and system for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception. This addresses the technical problem that existing technologies rely solely on static path planning to achieve bidirectional obstacle avoidance, which makes it difficult for AGVs to flexibly cope in complex environments, leading to a decrease in driving efficiency and safety.

[0004] To achieve the above objectives, this application provides a method and system for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception.

[0005] The first aspect of this application provides a method for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception, the method comprising:

[0006] Based on the first AGV's real-time ground conditions obtained through vehicle-mounted LiDAR scanning, driving interference is identified, and a first real-time safe speed is fitted and output. Based on the second AGV's real-time ground conditions obtained through vehicle-mounted LiDAR scanning, driving interference is identified, and a second real-time safe speed is fitted and output. A dynamic safe distance is calculated based on the first and second real-time safe speeds. If the real-time collision distance is less than the dynamic safe distance, the first and second AGVs are triggered to collect first obstacle avoidance correlation data and second obstacle avoidance correlation data, respectively. Based on the first and second obstacle avoidance correlation data, asymmetric safe turning capability is calculated, and a shared obstacle avoidance safety zone is located. Using the shared yield safety zone as a yield boundary constraint, a first simulated yield route and a second simulated yield route are fitted and output. With the yield direction as a constraint, the first and second yield energy consumptions of the first and second simulated yield routes are predicted respectively. By comparing the first and second yield energy consumptions, the roles of the first AGV and the second AGV are assigned, and the real-time yield route of the yielding vehicle and the original driving trajectory of the non-yielding vehicle are located. While controlling the real-time yielding vehicle to drive according to the real-time yield route, the non-yielding vehicle is controlled to slow down on the original driving trajectory until both the real-time yielding vehicle and the non-yielding vehicle leave the shared yield safety zone.

[0007] In one implementation, the following processing is also performed:

[0008] The first predetermined trajectory of the first AGV and the second predetermined trajectory of the second AGV are retrieved from the workshop logistics platform; the first real-time position of the first AGV and the second real-time position of the second AGV are obtained through the interactive vehicle UWB positioning system; the real-time conflict distance is calculated based on the first real-time position and the second real-time position, using the intersection segment of the first predetermined trajectory and the second predetermined trajectory as a constraint.

[0009] In one implementation, based on the first real-time ground conditions obtained by the first AGV through onboard LiDAR scanning, driving interference is identified, a first real-time safe speed is fitted and output, and the following processing is also performed:

[0010] The vehicle-mounted lidar of the first AGV scans the ground of the channel at a frequency of 10Hz to obtain the first reflection intensity point cloud data. Based on the reflectivity distribution characteristics, the first reflection intensity point cloud data is segmented using a dual threshold method to identify driving interference features, thereby obtaining the first oil stain distribution area and the first metal debris distribution area. After spatially superimposing the first oil stain distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safe speed is fitted and output.

[0011] In one implementation, after spatially superimposing the first oil spill distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, the first real-time safety velocity is fitted and output, and the following processing is also performed:

[0012] The system receives the first real-time transport load transmitted back by the load sensor of the first AGV; it receives the first real-time slope angle and the first real-time slope direction output by the IMU of the first AGV; it performs a safety adhesion attenuation analysis based on the first oil stain distribution area and the first metal debris distribution area to obtain the first oil stain attenuation coefficient and the first debris disturbance coefficient; it then constructs a dynamic adhesion model by spatially superimposing the first oil stain distribution area and the first metal debris distribution area, combined with the first real-time slope angle; based on the trajectory segment of the first predetermined trajectory at the first real-time position, it performs an adhesion safety analysis in the dynamic adhesion model and outputs the initial safety speed; it then compensates the initial safety speed with the first real-time slope direction and outputs the first real-time safety speed.

[0013] In one implementation, based on the reflectivity distribution characteristics, a dual-threshold segmentation method is used to identify driving interference features in the first reflectivity intensity point cloud data to obtain the first oil stain distribution area and the first metal debris distribution area, and the following processing is also performed:

[0014] Background noise in the first reflection intensity point cloud data is filtered out by removing invalid points with a reflection intensity of <5%; discrete point cloud clusters with a reflection intensity >85% are extracted by traversing the first reflection intensity point cloud data; isolated specular points are removed by traversing the discrete point cloud clusters using a preset point cloud spatial density threshold, and the first metal debris distribution area is output; continuous point cloud clusters with a reflection intensity of 5% to 30% are extracted by traversing the first reflection intensity point cloud data; micro-region removal is performed by traversing the continuous point cloud clusters using a preset point cloud projection area threshold, and morphological closing operations are performed on the continuous point cloud clusters to fuse adjacent oil stains, thereby generating the first oil stain distribution area.

[0015] In one implementation, the dynamic safety distance is calculated based on the first real-time safety speed and the second real-time safety speed, and the following processing is also performed:

[0016] Using the first real-time slope direction as a deceleration correction constraint, the first deceleration capacity is pre-calculated based on the first real-time transport load and the first real-time safe speed, and the second deceleration capacity of the second AGV is solved by analogy. A dominant safe speed is output by comparing the first real-time safe speed and the second real-time safe speed. A dominant deceleration is output by comparing the first deceleration capacity and the second deceleration capacity. A safe reaction distance is calculated based on a preset fixed response time and the dominant safe speed. A theoretical braking distance is calculated based on the dominant safe speed and the dominant deceleration. A first dynamic environmental margin is matched based on the first oil spill distribution area and the first metal debris distribution area. A second dynamic environmental margin is matched based on the second oil spill distribution area and the second metal debris distribution area. Based on the sum of the safe reaction distance and the theoretical braking distance, the maximum value of the first dynamic environmental margin and the second dynamic environmental margin is taken for safety distance compensation, and the dynamic safe distance is output.

[0017] In one implementation, based on the first obstacle avoidance association data and the second obstacle avoidance association data, the asymmetric safe turning capability is calculated, the shared obstacle avoidance safe zone is located, and the following processing is also performed:

[0018] When the first AGV is an uphill AGV and the second AGV is a downhill AGV: Front wheel load attenuation compensation is performed based on the first real-time transport load, the first real-time slope angle, and the first real-time slope direction, outputting a first minimum turning radius; rear wheel adhesion attenuation compensation is performed based on the second real-time transport load, the second real-time slope angle, and the second real-time slope direction, outputting a second minimum turning radius; based on the uphill expansion rule, the first minimum turning radius is mapped in the first real-time position space to generate a first dynamic envelope; based on the downhill expansion rule, the second minimum turning radius is mapped in the second real-time position space to generate a second dynamic envelope; the first and second dynamic envelopes are solved by union to obtain the conflict core area; after interactively obtaining the physical boundary of the transport channel, geometric Boolean operations are performed on the physical boundary of the transport channel and the core conflict area through convex polygon Boolean subtraction operations to output the shared avoidance safety zone.

[0019] In one implementation, the shared safety zone is used as a yield boundary constraint to fit and output a first simulated yield route and a second simulated yield route, and the following processing is also performed:

[0020] The shared obstacle avoidance safety zone space is used to divide the first predetermined trajectory and the second predetermined trajectory to obtain the first segmented trajectory and the second segmented trajectory. In the scenario where the first segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a cubic spline curve that satisfies the second minimum turning radius, which is used as the output of the first simulated yield route. In the scenario where the second segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a shortest obstacle avoidance Bezier curve that satisfies the first minimum turning radius, which is used as the output of the second simulated yield route.

[0021] A second aspect of this application provides an AGV bidirectional obstacle avoidance trajectory generation system that integrates road condition perception, the system comprising:

[0022] A first driving interference identification unit is used to identify driving interference based on the first real-time ground conditions obtained by the first AGV through vehicle-mounted LiDAR scanning, and to fit and output a first real-time safe speed; a second driving interference identification unit is used to identify driving interference based on the second real-time ground conditions obtained by the second AGV through vehicle-mounted LiDAR scanning, and to fit and output a second real-time safe speed; a dynamic safe distance calculation unit is used to calculate the dynamic safe distance based on the first real-time safe speed and the second real-time safe speed; an obstacle avoidance association data acquisition unit is used to trigger the first AGV and the second AGV to collect the first obstacle avoidance association data and the second obstacle avoidance association data respectively if the real-time conflict distance is less than the dynamic safe distance; an obstacle avoidance safety zone positioning unit is used to calculate the asymmetric safe turning capability based on the first obstacle avoidance association data and the second obstacle avoidance association data, and to position the obstacle avoidance safety zone. The system includes: a shared yielding safety zone; a simulated yielding route output unit, which uses the shared yielding safety zone as a yielding boundary constraint to fit and output a first simulated yielding route and a second simulated yielding route; a yielding energy consumption prediction unit, which uses the yielding direction as a constraint to predict the first yielding energy consumption and the second yielding energy consumption of the first simulated yielding route and the second simulated yielding route, respectively; a role allocation unit, which performs role allocation for the first AGV and the second AGV by comparing the first yielding energy consumption and the second yielding energy consumption, and locates the real-time yielding route of the yielding vehicle and the original driving trajectory of the non-yielding vehicle; and a vehicle control unit, which, while controlling the real-time yielding vehicle to drive according to the real-time yielding route, controls the non-yielding vehicle to slow down on the original driving trajectory until the real-time yielding vehicle and the non-yielding vehicle leave the shared yielding safety zone.

[0023] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0024] By using onboard LiDAR to scan the ground in real time, it can accurately identify driving obstacles such as oil stains and metal debris, and calculate the real-time safe speed of each AGV, ensuring the safe driving of AGVs in complex road conditions and providing a scientific basis for subsequent path planning, reducing potential safety hazards during driving. It calculates dynamic safe distances, automatically triggering an obstacle avoidance mechanism when two vehicles are about to collide, collecting obstacle avoidance correlation data for each vehicle, and providing information for subsequent path planning and decision-making. This mechanism ensures that AGVs can promptly perceive potential dangers and take corresponding avoidance actions. Combined with the asymmetric safe turning capability of each AGV, it accurately calculates the safe turning range of two AGVs in complex environments, thereby dynamically delineating a shared avoidance safety zone, enabling AGVs to navigate safely in traffic. This system effectively avoids collisions and improves the safety of two-way traffic. Based on a shared safety zone, it collaboratively plans an asymmetric smooth path to provide reasonable routes for both AGVs. During this process, energy consumption prediction determines the yielding vehicle, and path optimization ensures optimal energy efficiency. This not only improves driving efficiency but also further ensures the safety of the AGVs during the yielding process. By comparing the energy consumption of two simulated routes, roles are automatically assigned to determine which AGV should yield. The yielding vehicle is controlled in real time to travel along the predetermined yielding path, while the non-yielding vehicle slows down, ensuring that the two AGVs can pass each other smoothly in the same lane. This dynamic role allocation and control strategy effectively reduces conflicts when two vehicles meet, improving the work efficiency and safety of the workshop environment. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic diagram of the AGV bidirectional obstacle avoidance trajectory generation method integrating road condition perception provided in this application is shown.

[0027] Figure 2 The diagram illustrates the process of calculating the real-time conflict distance in the AGV bidirectional obstacle avoidance trajectory generation method that integrates road condition perception provided in this application.

[0028] Figure 3 The diagram illustrates the process of fitting and outputting the first real-time safe speed in the AGV bidirectional passage obstacle avoidance trajectory generation method with fused road condition perception provided in this application.

[0029] Figure 4A schematic diagram of the structure of the AGV bidirectional passage obstacle avoidance trajectory generation system with integrated road condition perception provided in this application is shown.

[0030] Explanation of reference numerals in the attached figures: 1. First driving interference identification unit; 2. Second driving interference identification unit; 3. Dynamic safe distance calculation unit; 4. Obstacle avoidance associated data acquisition unit; 5. Avoidance safety zone positioning unit; 6. Simulated yield route output unit; 7. Yield energy consumption prediction unit; 8. Role allocation unit; 9. Vehicle control unit. Detailed Implementation

[0031] This application provides a method and system for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception. This is intended to address the technical problem that existing technologies rely solely on static path planning to achieve bidirectional obstacle avoidance, which makes it difficult for AGVs to flexibly cope in complex environments, leading to a decrease in driving efficiency and safety.

[0032] Example 1: This embodiment of the invention provides a method for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception. See [link to example]. Figure 1 The method includes:

[0033] Based on the first real-time ground conditions obtained by the first AGV through onboard LiDAR scanning, driving interference is identified, and a first real-time safe speed is fitted and output.

[0034] The first AGV acquires real-time data of the road surface by scanning with an onboard LiDAR. The LiDAR scans the ground with high-frequency laser pulses, capturing reflected laser signals to generate point cloud data. This point cloud data contains reflection information from different objects on the ground, such as oil stains, metal debris, and obstacles. By analyzing the reflectivity distribution characteristics of the point cloud data, a dual-threshold segmentation method is used to identify characteristic areas of driving interference. This method sets upper and lower limits and classifies the point cloud data according to the reflection intensity value. Reflection intensity points above the upper limit are associated with hard objects such as metal debris, while reflection intensity points below the lower limit represent soft objects such as oil stains. This method can divide the ground conditions into multiple interference areas, primarily identifying oil stain areas and metal debris areas, both of which significantly affect the AGV's operation.

[0035] After identifying the interference areas, adhesion analysis is performed on these areas to determine the driving safety of the AGV in these areas. For example, oil stains can cause a decrease in tire adhesion, while metal debris can cause wheel damage. By combining the distribution of oil stains and metal debris, a dynamic adhesion model is constructed to calculate the driving safety in these interference areas and obtain the first initial safe speed. Then, by combining the slope direction and load information, speed compensation is performed on the first initial safe speed to output the first real-time safe speed, ensuring that the first AGV can drive safely and stably.

[0036] Based on the second real-time ground conditions obtained by the second AGV through onboard LiDAR scanning, driving interference is identified, and a second real-time safe speed is output.

[0037] The same operation is performed on the second AGV, including: acquiring point cloud data using the onboard LiDAR of the second AGV; identifying driving interference, including oil stains and metal debris, using a dual-threshold segmentation method; and performing dynamic adhesion analysis based on the load information of the second AGV to output a second real-time safe speed. Since the acquisition process of the second real-time safe speed is almost identical to that of the first real-time safe speed, the detailed process has already been explained in the previous step, and the description of the second real-time safe speed can be omitted and repeated.

[0038] The dynamic safety distance is calculated based on the first real-time safety speed and the second real-time safety speed.

[0039] The first and second real-time safe speeds represent the maximum safe speeds that the two AGVs can travel at under the current ground conditions. Based on the vehicle's load, gradient, and real-time safe speed, the deceleration capacity of each AGV is estimated. Deceleration capacity is typically related to factors such as vehicle weight, tire adhesion, and ground conditions. By comparing the safe speeds and deceleration capacities of the first and second AGVs, the dominant safe speed and dominant deceleration are determined—that is, which AGV's speed and deceleration have a greater influence and become the dominant factors in the calculation. Based on the dominant safe speed and deceleration, the safe reaction distance when the vehicle encounters an obstacle is calculated, i.e., the distance required for the vehicle to respond and begin deceleration. Furthermore, the theoretical braking distance, i.e., the distance required for the vehicle to come to a complete stop from the start of deceleration, is calculated based on the dominant safe speed and dominant deceleration. For each AGV, its dynamic environmental margin is calculated based on the oil and metal debris areas it is located in. The dynamic environmental margin refers to the vehicle's adhesion margin under given road conditions. The safe reaction distance and theoretical braking distance of the first and second AGVs are added together with the dynamic environmental margin to obtain the total dynamic safe distance, which is used to determine the safe distance between the two AGVs, thereby avoiding collisions.

[0040] If the real-time collision distance is less than the dynamic safety distance, the first AGV and the second AGV are triggered to collect the first obstacle avoidance association data and the second obstacle avoidance association data, respectively.

[0041] The system monitors the real-time collision distance between the first and second AGVs. This distance is dynamically determined by the vehicles' current relative positions and speeds, and is obtained through the vehicles' UWB positioning system or similar high-precision positioning technology. If the real-time collision distance is less than the dynamic safety distance, it means the two vehicles are very close and there is a risk of collision. In this case, an obstacle avoidance mechanism is triggered, and obstacle avoidance-related data is collected for subsequent path planning and adjustments. Specifically, for the first AGV, real-time data related to its movement is collected, including its position, speed, direction of travel, load, and slope information, to obtain first obstacle avoidance-related data for assessing potential obstacles and interference during the first AGV's movement. Similarly, for the second AGV, its corresponding real-time data is collected to obtain second obstacle avoidance-related data.

[0042] Based on the first obstacle avoidance association data and the second obstacle avoidance association data, the asymmetric safe turning capability is calculated, and the shared obstacle avoidance safety zone is located.

[0043] The asymmetric safe turning capability of two AGVs is calculated by analyzing their obstacle avoidance correlation data. Asymmetric safe turning capability refers to the different turning behaviors of the AGVs in confined spaces due to factors such as their dynamic characteristics, load, and turning radius. For example, the first and second AGVs may have different loads; increasing the load affects their turning radius and stability. Therefore, the turning capability is adjusted using load information. The shared avoidance safety zone is a safe area established to avoid collisions when two AGVs are about to meet. By superimposing the turning radii of the two vehicles and dynamic environmental conditions, the safe meeting area for the two vehicles in a confined space is calculated. This safety zone is a geometric area that vehicles should avoid entering to ensure successful obstacle avoidance.

[0044] Using the shared safety zone as a yield boundary constraint, the first simulated yield route and the second simulated yield route are fitted and output.

[0045] Using the calculated shared obstacle avoidance safety zone as the boundary condition for yielding, in narrow passages or areas where intersections are imminent, the yielding boundary will restrict the travel trajectories of the two AGVs, generating two simulated yielding routes: The first simulated yielding route represents the travel trajectory of the first AGV under yielding conditions. This route is generated by constraining within the shared obstacle avoidance safety zone and using the shortest obstacle avoidance path fitting method, such as choosing a cubic spline curve to fit the route, ensuring that the path is smooth and does not conflict with obstacles; Similarly, the second AGV will also generate a second simulated yielding route, which works in conjunction with the first simulated yielding route to ensure that both AGVs can find a safe travel path within the shared obstacle avoidance safety zone.

[0046] Using the yielding direction as a constraint, the first yielding energy consumption and the second yielding energy consumption of the first simulated yielding route and the second simulated yielding route are predicted respectively.

[0047] Determine the yielding direction. Based on the calculation results of the previous steps, identify which AGV should yield. The yielding direction is determined based on a variety of factors, including the relative position, speed, and dynamic safety distance between the AGVs. If two AGVs are in a narrow passage where they meet, calculate which AGV should yield to avoid a collision.

[0048] Assuming the first AGV is determined to yield, energy consumption is predicted along its first simulated yielding route. This prediction is based on factors such as turning radius, gradient, speed, and road friction coefficient. For example, a larger turning radius means less steering effort and lower energy consumption, while a smaller turning radius may lead to higher energy consumption. Similarly, energy consumption is predicted along the second simulated yielding route for the second AGV. Like the prediction for the first AGV, the prediction factors also include turning radius, gradient, speed, and road friction coefficient.

[0049] By comparing the first yielding energy consumption and the second yielding energy consumption, the roles of the first AGV and the second AGV are assigned, and the real-time yielding route of the yielding vehicle and the original driving trajectory of the non-yielding vehicle are located.

[0050] Comparing the energy consumption of the first and second yielding actions, the path with lower energy consumption indicates greater efficiency, making the AGV more suitable for the yielding role. The AGV with lower energy consumption should be prioritized as the yielding vehicle to minimize unnecessary energy waste. Based on the comparison of yielding energy consumption, the AGV with lower energy consumption is assigned as the yielding vehicle because its travel path is more efficient and consumes less energy. The AGV with higher energy consumption is assigned as the non-yielding vehicle; this vehicle will continue to travel on its original trajectory, but its speed will be appropriately reduced to provide space for the yielding vehicle.

[0051] After determining the roles of yielding and non-yielding vehicles, real-time yielding routes and original driving trajectories are planned for them respectively. The real-time yielding route refers to the actual path traveled by the yielding vehicle during the yielding process, which is based on the simulated yielding route calculated earlier, but will be finely adjusted according to the real-time situation in actual control. The original driving trajectory refers to the original planned driving path of the non-yielding vehicle, with speed adjustments made, but the original path is basically maintained.

[0052] While controlling the real-time yielding vehicle's movement according to the real-time yielding route, the non-yielding vehicle is controlled to slow down along the original driving trajectory until the real-time yielding vehicle and the non-yielding vehicle leave the shared avoidance safety zone.

[0053] The system controls the yielding vehicle to travel along its real-time yielding route. During this process, the vehicle follows a predetermined path, and its speed and direction are adjusted by the vehicle's control system to ensure it does not deviate from the track and avoids collisions. Simultaneously, the non-yielding vehicle continues along its original driving trajectory. During this process, the non-yielding vehicle does not change its original path but instead slows down to avoid the yielding vehicle. Once both the yielding and non-yielding vehicles have left the shared yielding safety zone, meaning the two vehicles are no longer in a meeting state, the yielding process ends. At this point, the non-yielding vehicle can resume its original speed and continue its original trajectory, and the yielding vehicle can resume its normal driving mode. Through these steps, dynamic yielding and collaborative control of bidirectional AGVs in complex environments can be achieved, ensuring that two AGVs can automatically assign roles, adjust paths, and avoid collisions when traversing narrow passages, improving transportation efficiency and safety.

[0054] See Figure 2 One implementation also includes:

[0055] The first predetermined trajectory of the first AGV and the second predetermined trajectory of the second AGV are retrieved from the workshop logistics platform; the first real-time position of the first AGV and the second real-time position of the second AGV are obtained through the interactive vehicle UWB positioning system; the real-time conflict distance is calculated based on the first real-time position and the second real-time position, using the intersection segment of the first predetermined trajectory and the second predetermined trajectory as a constraint.

[0056] The workshop logistics platform is a centralized scheduling and management platform responsible for coordinating the scheduling tasks, path planning, and real-time monitoring of all AGVs. The platform retrieves the planned trajectories of the first and second AGVs, i.e., the planned paths of the two AGVs. The first planned trajectory refers to the path planned by the first AGV during task execution, based on the task scheduling plan. This trajectory is typically preset based on factors such as the workshop layout, task requirements, and the driving status of other vehicles. Similarly, the second planned trajectory refers to the path determined by the second AGV based on its task and scheduling rules.

[0057] Ultra-Wideband (UWB) is a high-precision positioning technology that provides accurate positioning information over a short range. UWB positioning systems can determine the real-time position of AGVs through signal transmission and reception devices, exhibiting a low error range and making them suitable for complex workshop environments. The UWB positioning system obtains the current precise position of a first AGV and a second AGV by tracking them in real time. The first real-time position indicates the accurate current position of the first AGV, and the second real-time position indicates the accurate current position of the second AGV.

[0058] Determine whether the predetermined trajectories of the first AGV and the second AGV intersect. The trajectory intersection segment refers to the part of the two predetermined paths that overlap in space. If the two predetermined trajectories overlap in the workshop, it means that the two AGVs will arrive at the same time in this overlapping area, so conflict detection and avoidance are required. If the two predetermined trajectories do not intersect, then there is no risk of conflict, and there is no need to further calculate the real-time conflict distance.

[0059] After determining the trajectory intersection segment, the first and second real-time positions are used to calculate the real-time collision distance between the two AGVs within the intersection segment. The real-time collision distance represents the actual distance between the two AGVs within the trajectory intersection segment. By calculating the distance between these two positions, it can be determined whether a collision will occur. If the calculated real-time collision distance is greater than the safe distance, then the two AGVs can continue to travel along their respective predetermined trajectories without needing to avoid each other.

[0060] See Figure 3 In one implementation, driving interference is identified based on the first real-time ground conditions obtained by the first AGV through onboard LiDAR scanning, and a first real-time safe speed is fitted and output, including:

[0061] The vehicle-mounted lidar of the first AGV scans the ground of the channel at a frequency of 10Hz to obtain the first reflection intensity point cloud data. Based on the reflectivity distribution characteristics, the first reflection intensity point cloud data is segmented using a dual threshold method to identify driving interference features, thereby obtaining the first oil stain distribution area and the first metal debris distribution area. After spatially superimposing the first oil stain distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safe speed is fitted and output.

[0062] Vehicle-mounted LiDAR is a high-precision sensor that generates point cloud data by emitting laser pulses and receiving reflected signals. Each reflected signal contains information such as the spatial location (X, Y, Z coordinates) and reflection intensity of the reflecting point. The vehicle-mounted LiDAR of the first AGV scans at a frequency of 10Hz, meaning it scans 10 times per second, generating corresponding point cloud data. The choice of scanning frequency depends on the AGV's speed and environmental requirements; 10Hz is sufficient to support high-precision ground detection, ensuring timely identification of ground obstacles and driving interference. With each scan, the vehicle-mounted LiDAR generates a series of point cloud data, where the data for each point includes the reflection intensity, i.e., the signal strength reflected back from the laser pulse. These reflection intensities provide information about the material of the ground and object surfaces; for example, metal objects and oily surfaces will produce reflection signals of different intensities.

[0063] Reflectivity distribution characteristics refer to the distribution of reflection intensity values ​​at each point in point cloud data. Different surface materials have different reflectivities; for example, metal surfaces typically have high reflectivity, while oil stains and other soft surfaces have low reflectivity. Therefore, by analyzing the distribution of reflection intensity, different ground interference areas can be identified. The dual-threshold segmentation method is an image processing and point cloud data processing method. In this method, two thresholds are set: an upper threshold and a lower threshold. Based on the different reflection intensities in the point cloud data, different types of ground interference are distinguished. The upper threshold is used to identify areas with high reflection intensity, usually associated with hard objects such as metal debris; the lower threshold is used to identify areas with low reflection intensity, usually associated with soft objects such as oil stains and dust. By setting the upper and lower thresholds, the first reflection intensity point cloud data is divided into multiple regions. The high reflection intensity region indicates the first metal debris distribution area, and the low reflection intensity region indicates the first oil stain distribution area. These two types of interference areas are important factors affecting the safe operation of AGVs.

[0064] The first oil spill distribution area and the first metal debris distribution area are spatially superimposed. Spatial superposition means merging the spatial positions of the two interference areas to obtain a comprehensive interference area, reflecting the overall impact of the two different interference areas on the ground. Adhesion refers to the frictional force between the AGV tires and the ground. In the oil spill and metal debris areas, the AGV's driving stability and safety are greatly affected due to the decrease in adhesion. By superimposing the oil spill and metal debris distribution areas and combining them with the adhesion model, the adhesion decay of the AGV in these areas is analyzed, and the driving safety under the current environment is evaluated. Based on the results of the adhesion safety analysis, a first real-time safe speed is fitted, which is the maximum speed at which the first AGV can maintain stable and safe driving under the current ground conditions. For example, if the influence of oil spill and metal debris is large, a lower safe speed is output to avoid the AGV slipping due to loss of adhesion.

[0065] In one implementation, after spatially superimposing the first oil spill distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safety velocity is fitted and output, including:

[0066] The system receives the first real-time transport load transmitted back by the load sensor of the first AGV; it receives the first real-time slope angle and the first real-time slope direction output by the IMU of the first AGV; it performs a safety adhesion attenuation analysis based on the first oil stain distribution area and the first metal debris distribution area to obtain the first oil stain attenuation coefficient and the first debris disturbance coefficient; it then constructs a dynamic adhesion model by spatially superimposing the first oil stain distribution area and the first metal debris distribution area, combined with the first real-time slope angle; based on the trajectory segment of the first predetermined trajectory at the first real-time position, it performs an adhesion safety analysis in the dynamic adhesion model and outputs the initial safety speed; it then compensates the initial safety speed with the first real-time slope direction and outputs the first real-time safety speed.

[0067] A load sensor is a sensor installed on an AGV that can monitor the AGV's load in real time. It obtains the AGV's current transport load by measuring the compression of the vehicle chassis or directly sensing the weight of the transported goods. The system receives real-time data from the load sensor of the first AGV to obtain the first AGV's current real-time transport load. This load data is dynamic and may change at any time depending on the transport task.

[0068] An IMU (Inertial Measurement Unit) is a sensor used to measure the acceleration, angular velocity, and attitude changes of an object. In an AGV (Automated Guided Vehicle), the main function of the IMU is to detect the vehicle's slope angle and slope direction in real time, that is, the vehicle's tilt angle and tilt direction on the ground. The first real-time slope angle represents the tilt angle of the vehicle relative to the horizontal plane. A larger slope angle means that the vehicle is on a steeper slope. The slope angle affects the AGV's traction and braking force, especially in environments with oil or metal debris, where changes in slope can affect the vehicle's stability. The first real-time slope direction refers to the tilt direction of the slope, that is, whether the slope is uphill or downhill, which affects the distribution of the vehicle's acceleration and braking force.

[0069] Safety adhesion degradation analysis is used to calculate the adhesion changes of AGVs under different ground conditions. Oil and metal debris both negatively affect the adhesion of AGVs, so the effects of these two disturbances need to be analyzed separately. Specifically, oil or other slippery substances reduce the friction between the tire and the ground, leading to adhesion degradation. The first oil degradation coefficient represents the degree of adhesion reduction of the first AGV within the oil distribution area, and this coefficient is related to factors such as the density and distribution range of the oil, as well as the load of the AGV. Metal debris affects tire stability and adhesion, especially during high-speed driving or turning, which may lead to loss of vehicle control. The first debris disturbance coefficient represents the degree of adhesion reduction of the first AGV within the metal debris distribution area, and it is related to the size, quantity, and distribution density of the debris.

[0070] The first identified oil stain distribution area and the first metal debris distribution area are spatially superimposed. Spatial superposition means merging the spatial data of the two areas to obtain a comprehensive ground interference area. This ground interference area includes both oil stains and metal debris, and this composite area can more accurately reflect the combined impact of the two interference factors on the vehicle. Next, a dynamic adhesion model is constructed by combining the first real-time slope angle. When driving on a slope, the slope angle has a significant impact on adhesion. When going uphill, the AGV's adhesion decreases because gravity causes more wheel slippage. When going downhill, the AGV's adhesion increases, but if the speed is too high, it may also cause slippage. The dynamic adhesion model combines the distribution of oil stains and metal debris, as well as the influence of the slope angle on adhesion, to dynamically evaluate the changes in vehicle adhesion during driving and to calculate the AGV's driving stability in this complex environment.

[0071] Based on the first predetermined trajectory, a trajectory segment is extracted at the first real-time position. This trajectory segment is part of the path of the first AGV near the current position. The trajectory segment can be represented by dividing the entire first predetermined trajectory into multiple small segments, which represent a small segment of the future travel path of the first AGV.

[0072] By combining a dynamic adhesion model, an adhesion safety analysis is performed on the trajectory segment. This involves assessing whether the first AGV can maintain sufficient adhesion to ensure safe operation on a specific trajectory segment. During this process, the impact of factors such as oil, metal debris, and slope on vehicle adhesion is evaluated. Areas with poor adhesion may cause the first AGV to slip or lose stability. Based on the adhesion decay information output by the dynamic adhesion model, the adhesion on the trajectory segment is calculated. If the adhesion falls below a certain threshold, the safe speed is reduced to prevent loss of vehicle control. Based on the adhesion safety analysis results, an initial safe speed is calculated. This initial safe speed is the maximum speed at which the first AGV can maintain stable operation on the current trajectory segment, based on real-time ground conditions and slope information. The initial safe speed is typically lower than the normal operating speed to ensure safety.

[0073] The initial safe speed is compensated based on the first real-time slope direction. The slope direction determines the inclination of the slope—whether it's uphill or downhill. If the first real-time slope direction is uphill, the vehicle needs more driving force to overcome gravity, potentially reducing traction. Therefore, increasing the initial safe speed ensures the vehicle can safely climb the slope. If the first real-time slope direction is downhill, the vehicle's gravity helps with acceleration, so the initial safe speed needs appropriate deceleration compensation to prevent slippage or loss of control due to excessive speed. Finally, after slope direction compensation, the first real-time safe speed is obtained. This is an adjusted speed value that maximizes the driving efficiency of the first AGV while ensuring traction.

[0074] In one implementation, based on the reflectivity distribution characteristics, a dual-threshold segmentation method is used to identify driving interference features in the first reflectivity point cloud data, resulting in a first oil stain distribution area and a first metal debris distribution area, including:

[0075] Background noise in the first reflection intensity point cloud data is filtered out by removing invalid points with a reflection intensity of <5%; discrete point cloud clusters with a reflection intensity >85% are extracted by traversing the first reflection intensity point cloud data; isolated specular points are removed by traversing the discrete point cloud clusters using a preset point cloud spatial density threshold, and the first metal debris distribution area is output; continuous point cloud clusters with a reflection intensity of 5% to 30% are extracted by traversing the first reflection intensity point cloud data; micro-region removal is performed by traversing the continuous point cloud clusters using a preset point cloud projection area threshold, and morphological closing operations are performed on the continuous point cloud clusters to fuse adjacent oil stains, thereby generating the first oil stain distribution area.

[0076] When acquiring the initial reflection intensity point cloud data, various environmental interferences, such as dust in the air or low-reflectivity objects far from the target, may lead to inaccurate measurement results. These irrelevant points are called background noise, which can affect subsequent analysis and therefore needs to be filtered out. By setting a reflection intensity threshold, such as 5%, invalid points with reflection intensities below this threshold are removed. These low-reflectivity points are mostly background noise, representing distant objects, airborne dust, or irrelevant surface reflections. After removing background noise, the remaining point cloud data contains valid reflection signals from ground obstacles or interfering areas, providing more accurate data for subsequent feature recognition and region segmentation.

[0077] After filtering out background noise, discrete point cloud clusters with a reflection intensity greater than 85% are extracted from the remaining first reflection intensity point cloud data. Discrete point cloud clusters refer to a group of points that are relatively concentrated in space. Point groups with a reflection intensity greater than 85% represent highly reflective objects, such as metal scraps, hard objects, and mechanical equipment. These objects strongly reflect the LiDAR signal and can therefore be distinguished from background noise.

[0078] Point cloud spatial density refers to the number of points contained within a unit volume within a certain spatial range. By setting a preset point cloud spatial density threshold, it is possible to determine which points belong to dense regions and which points are isolated highlights. Isolated highlights refer to single points in a point cloud cluster that have abnormally high reflection intensity and are relatively independent in location. These points may be caused by noise, reflection angle, or accidental reflection, and do not represent actual obstacles or interference. By traversing discrete point cloud clusters and calculating their point cloud spatial density, and based on the preset point cloud spatial density threshold, those isolated highlights with low density are identified. These isolated highlights are not actual metal debris, but rather accidental reflection anomalies. These isolated highlights are removed, and only point cloud clusters with high density, representing the actual distribution of metal debris, are retained. After removing isolated highlights, the remaining point cloud clusters are the first metal debris distribution areas, and these areas represent the actual distribution locations of metal debris on the ground.

[0079] The system iterates through the first reflection intensity point cloud data and extracts continuous point cloud clusters with reflection intensities between 5% and 30%. Objects such as oil stains, dust, or water surfaces typically have low reflection intensities, so their point cloud data often falls within this range. By selecting this range, these interfering objects can be identified. Continuous point cloud clusters refer to groups of points that are relatively continuous in space and whose reflection intensities conform to a set range. Extracting these continuous point cloud clusters from the first reflection intensity point cloud data represents oil stains or slippery areas.

[0080] Point cloud projection area refers to the projection area of ​​a continuous point cloud cluster on a two-dimensional plane. By setting a preset point cloud projection area threshold, it is possible to identify point cloud clusters with small areas that do not have actual interference significance. These clusters represent tiny spots that do not have actual interference. Removing these point cloud clusters can avoid unnecessary noise and allow for focused analysis of oil or slippery areas with large areas that may affect the operation of AGVs.

[0081] Morphological closing is a mathematical tool used in image processing to handle shape changes in images. Here, morphological closing is applied to continuous point cloud clusters. It involves two processes: dilation and erosion, which fill small gaps in the point cloud and smooth edges. In this step, morphological closing is used to merge adjacent oil stains, combining small, adjacent oil stains into a larger region. This allows for a more accurate description of the actual extent of the oil stain's influence. After morphological closing, a coherent first oil stain distribution region is generated, representing the location and extent of the oil stains on the ground.

[0082] In one implementation, the dynamic safety distance is calculated based on the first real-time safety velocity and the second real-time safety velocity, including:

[0083] Using the first real-time slope direction as a deceleration correction constraint, the first deceleration capacity is pre-calculated based on the first real-time transport load and the first real-time safe speed, and the second deceleration capacity of the second AGV is solved by analogy. A dominant safe speed is output by comparing the first real-time safe speed and the second real-time safe speed. A dominant deceleration is output by comparing the first deceleration capacity and the second deceleration capacity. A safe reaction distance is calculated based on a preset fixed response time and the dominant safe speed. A theoretical braking distance is calculated based on the dominant safe speed and the dominant deceleration. A first dynamic environmental margin is matched based on the first oil spill distribution area and the first metal debris distribution area. A second dynamic environmental margin is matched based on the second oil spill distribution area and the second metal debris distribution area. Based on the sum of the safe reaction distance and the theoretical braking distance, the maximum value of the first dynamic environmental margin and the second dynamic environmental margin is taken for safety distance compensation, and the dynamic safe distance is output.

[0084] During vehicle operation, the slope direction has a direct impact on the vehicle's braking performance. Especially when driving on a slope, the slope direction determines the way gravity acts. Therefore, using the first real-time slope direction as a correction constraint for deceleration can accurately adjust the AGV's braking capability to ensure safe driving.

[0085] Based on the first real-time transport load and the first real-time safe speed, combined with the first real-time slope direction, the first deceleration capacity of the first AGV is pre-calculated. The vehicle's load affects its braking capacity; a heavier load requires more braking force to stop the vehicle, thus the greater the load, the smaller the deceleration. The safe speed affects the vehicle's deceleration capacity; the higher the speed, the greater the braking force required, thus deceleration capacity and speed have an inverse relationship. The slope direction affects the effect of gravity on the vehicle; for example, when going uphill, the vehicle's gravity direction is opposite to the travel direction, resulting in better deceleration; when going downhill, the vehicle's gravity direction is the same as the travel direction, requiring stronger deceleration. Based on these factors, the first deceleration capacity of the first AGV is calculated. Subsequently, the second deceleration capacity of the second AGV is calculated in a similar manner. Deceleration capacity is the maximum deceleration value that a vehicle can achieve under given conditions.

[0086] The dominant safe speed refers to the highest safe speed between two AGVs. Its purpose is to ensure that the two AGVs can safely avoid collisions when they meet. The dominant safe speed is determined by comparing the first real-time safe speed and the second real-time safe speed, which are the maximum speeds at which the two AGVs can safely travel under the current ground conditions. The higher of the two safe speeds is used as the benchmark. This is done to ensure the most conservative safety margin, guaranteeing that either AGV can travel at a safe speed when they meet, thus avoiding a collision.

[0087] Comparing the first and second deceleration capabilities, the output dominant deceleration refers to the maximum deceleration capability between the two AGVs. In order to ensure that the two AGVs can stop safely in the shortest time, the higher deceleration capability is taken as the standard for dominant deceleration.

[0088] The safe reaction distance refers to the distance an AGV travels from sensing a potential collision to beginning to decelerate, including the distance the vehicle moves during the reaction process. A preset fixed response time is used to represent the time from when the AGV senses a potential hazard to when it begins to take action (such as decelerating). This time is usually an empirical or standardized value, representing the AGV's reaction speed when facing a hazard. The safe reaction distance is calculated by multiplying the dominant safe speed by the preset fixed response time; this safe reaction distance represents the distance the AGV travels within the reaction time when responding to an emergency.

[0089] The theoretical braking distance refers to the shortest distance required for an AGV to come to a complete stop from the start of deceleration. The higher the dominant safe speed, the longer the required theoretical braking distance. At the same time, the greater the dominant deceleration, the shorter the theoretical braking distance. The calculated theoretical braking distance reflects the distance the AGV travels from the start of deceleration to a complete stop.

[0090] The first dynamic environmental margin refers to the degree of influence of specific environmental conditions (such as oil stains, metal debris distribution, etc.) on the adhesion and braking performance of the first AGV. It reflects the factors in the environment that affect vehicle braking or adhesion, thus impacting safe vehicle operation. The first dynamic environmental margin is calculated and matched based on the first oil stain distribution area and the first metal debris distribution area. Oil stains reduce the adhesion between the tires and the ground, thereby reducing the AGV's braking ability. Metal debris can also lead to a decrease in adhesion, increasing the risk of vehicle slippage. Combining the characteristics of these two interference areas, the adhesion attenuation value of the environment is estimated, thus deriving the first dynamic environmental margin, which reflects the degree of weakening of the first AGV's adhesion under the influence of oil stains and metal debris. A larger dynamic environmental margin indicates a smaller impact of the environment on the AGV's adhesion, resulting in higher vehicle safety; conversely, a smaller dynamic environmental margin indicates a greater impact of the environment, resulting in lower AGV safety.

[0091] Similar to the first AGV, the driving performance of the second AGV in areas with oil and metal debris was evaluated, the degree of adhesion attenuation was calculated, and the second dynamic environmental margin was obtained through the same analysis process.

[0092] The safe reaction distance and the theoretical braking distance are summed. The sum represents the total safe distance required for the AGV to react and come to a complete stop from sensing danger. The total safe distance is compensated based on the maximum of the first and second dynamic environmental margins to ensure the most conservative safety considerations. A larger dynamic environmental margin indicates better adhesion and stronger braking performance; a smaller dynamic environmental margin indicates poorer adhesion and the need for more safety margin. The compensated dynamic safe distance ensures the AGV can operate safely in complex environments.

[0093] In one implementation, the asymmetric safe turning capability is calculated based on the first obstacle avoidance association data and the second obstacle avoidance association data, and a shared obstacle avoidance safety zone is located, including:

[0094] When the first AGV is an uphill AGV and the second AGV is a downhill AGV: Front wheel load attenuation compensation is performed based on the first real-time transport load, the first real-time slope angle, and the first real-time slope direction, outputting a first minimum turning radius; rear wheel adhesion attenuation compensation is performed based on the second real-time transport load, the second real-time slope angle, and the second real-time slope direction, outputting a second minimum turning radius; based on the uphill expansion rule, the first minimum turning radius is mapped in the first real-time position space to generate a first dynamic envelope; based on the downhill expansion rule, the second minimum turning radius is mapped in the second real-time position space to generate a second dynamic envelope; the first and second dynamic envelopes are solved by union to obtain the conflict core area; after interactively obtaining the physical boundary of the transport channel, geometric Boolean operations are performed on the physical boundary of the transport channel and the core conflict area through convex polygon Boolean subtraction operations to output the shared avoidance safety zone.

[0095] An uphill AGV refers to an AGV that is moving up a slope. When going uphill, the AGV's direction of movement is opposite to the direction of gravity, requiring more power to overcome gravity, and may require stronger deceleration capabilities when braking. A downhill AGV refers to an AGV that is moving down a slope. When going downhill, the AGV's direction of gravity is the same as the direction of movement, and the vehicle will be accelerated by gravity, requiring more control to prevent loss of control.

[0096] When the first AGV is traveling uphill, the weight on its front wheels increases because the vehicle's gravity direction is opposite to its travel direction. Front wheel load attenuation compensation is necessary to adjust for changes in front wheel adhesion caused by variations in slope angle and load. This compensation is essential because when going uphill, the first AGV needs to overcome gravity, which may reduce front wheel adhesion, affecting its turning ability. The initial real-time transport load affects the pressure on the front and rear wheels; a heavier load results in a greater load on both wheels, impacting the vehicle's turning radius and stability. The initial real-time slope angle and direction determine the additional load on the front wheels when going uphill. A steeper slope results in a greater weight on the front wheels, potentially leading to adhesion attenuation. Compensation needs to be based on real-time slope information. Based on these factors, a front wheel load attenuation compensation value is calculated, and the first minimum turning radius of the first AGV is adjusted accordingly. Larger slopes and heavier loads typically lead to a larger turning radius, resulting in a larger initial minimum turning radius.

[0097] When the second AGV is traveling downhill, its gravity direction is the same as its travel direction. The rear wheels experience acceleration due to gravity, which may reduce their traction. Rear wheel traction attenuation compensation is used to adjust for changes in rear wheel traction caused by gravity during downhill travel. Similar to the first AGV, the real-time transport load affects the rear wheel's load-bearing capacity; a heavier load results in greater pressure on the rear wheels. Downhill, the real-time slope angle and direction also affect rear wheel traction. A larger downhill angle further reduces rear wheel traction, affecting stability during turns. Based on these factors, a rear wheel traction attenuation compensation value is calculated, and the second minimum turning radius of the second AGV is calculated accordingly. A steeper slope and heavier load will increase the vehicle's turning radius, requiring adjustment to ensure no loss of control during downhill travel.

[0098] The uphill expansion rule is used to model the motion of the first AGV under uphill conditions. Specifically, when the first AGV is traveling on a slope, its turning radius changes. Gravity causes the front wheels to bear a greater load when the vehicle is going uphill, thus affecting the turning radius. Based on the first real-time position, i.e., the current position of the first AGV, the first minimum turning radius is mapped to an actual spatial region. This region represents the space the first AGV might occupy, i.e., the first dynamic envelope. The first dynamic envelope indicates the space the first AGV might occupy during turning, ensuring that subsequent path planning avoids these regions.

[0099] The downhill extension rule is used to handle the turning ability of the second AGV under downhill conditions. Similar to the uphill situation, the direction of gravity is the same as the direction of travel when going downhill, which may cause the rear wheels of the second AGV to bear greater traction, thus affecting its turning ability. Based on the second real-time position, i.e., the current position of the second AGV, the second minimum turning radius is mapped to an actual spatial region, forming a second dynamic envelope. The second dynamic envelope represents the space that the second AGV may occupy when turning downhill, providing a basis for obstacle avoidance calculations.

[0100] By performing a union solution on the first and second dynamic envelopes, the shared space that the two AGVs may occupy at the same time is identified, i.e., the area where they may collide. The core conflict area is the area where the dynamic envelopes of the two AGVs overlap. This means that if the two vehicles appear at the same time in this area, a collision may occur. The union solution merges the turning path spaces of the two vehicles to obtain a common conflict area, providing the space that needs to be avoided.

[0101] The physical boundary of the transport channel refers to the actual boundary of the area where the AGV can travel. It is usually determined by the physical structure of the workshop, such as walls and obstacles. Through interaction, the physical boundary information is obtained for accurate path planning. Boolean operations are geometric operations used to merge, intersect, and differ between multiple areas. Here, a convex polygon Boolean subtraction operation is used, that is, to remove the conflict core area from the physical boundary of the transport channel, resulting in a conflict-free area. By subtracting the conflict core area, a shared avoidance safety zone is obtained, ensuring that the AGV will not enter the potential collision area during the avoidance process.

[0102] In one implementation, the shared safety zone is used as a yield boundary constraint, and a first simulated yield route and a second simulated yield route are fitted and output, including:

[0103] The shared obstacle avoidance safety zone space is used to divide the first predetermined trajectory and the second predetermined trajectory to obtain the first segmented trajectory and the second segmented trajectory. In the scenario where the first segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a cubic spline curve that satisfies the second minimum turning radius and is used as the output of the second simulated yield route. In the scenario where the second segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a shortest obstacle avoidance Bezier curve that satisfies the first minimum turning radius and is used as the output of the first simulated yield route.

[0104] The shared avoidance safety zone refers to the area that two AGVs must avoid when they meet. It is formed by the overlapping part of the dynamic envelopes of the two AGVs. This area represents the space where a collision may occur. Based on the shared avoidance safety zone space, the possible overlapping parts in the first and second predetermined trajectories are spatially segmented. This process ensures that only the overlapping parts are analyzed, rather than the entire trajectory. The first segmented trajectory represents the part in the first predetermined trajectory where a collision may occur. The first segmented trajectory is based on the segmentation result of the shared avoidance safety zone and only focuses on these areas with potential conflicts. Similarly, the second segmented trajectory represents the part in the second predetermined trajectory where a collision may occur.

[0105] Using the first segmented trajectory as a fixed path, and employing a shortest path algorithm under the constraint of a shared obstacle avoidance safety space, a second segmented trajectory for the second AGV is fitted. This ensures the second AGV avoids the first AGV's path, with the goal of generating the shortest obstacle avoidance path while minimizing the second AGV's travel distance without violating the shared obstacle avoidance safety zone constraint. The second minimum turning radius is a limitation on the second AGV's turning capability. Considering the turning radius, the obstacle avoidance trajectory of the second AGV is ensured to meet its minimum turning radius requirement, meaning that no unsafe sharp turns will occur during actual obstacle avoidance. This constraint guarantees that the fitted trajectory satisfies both obstacle avoidance requirements and the second AGV's driving capability. To make the trajectory smooth and meet turning requirements, a cubic spline curve is used for fitting. Cubic spline curves can smoothly transition between control points on the path, ensuring the smoothness and continuity of the trajectory and avoiding sharp turns or discontinuous paths. Finally, a second simulated yielding route is generated, representing the shortest path that the second AGV should follow during the yielding process, ensuring it avoids the first AGV and passes smoothly.

[0106] Using the second segmented trajectory as a fixed path, and constrained by the shared obstacle avoidance safety space, the shortest obstacle avoidance trajectory for the first AGV is fitted. The aim is to calculate the path for the first AGV so that it can avoid the second AGV while ensuring the shortest possible travel path. The first minimum turning radius represents the turning capability of the first AGV. When fitting the path, it is essential to ensure that the first AGV can safely complete the turns on the path without exceeding its minimum turning radius limit. A Bézier curve is used to fit the obstacle avoidance trajectory of the first AGV. The Bézier curve has good flexibility and controllability, and can generate a smooth obstacle avoidance path while ensuring that the path does not contain inappropriate turns or discontinuous movements. Based on the fitting results, a first simulated yielding route is generated. The first simulated yielding route represents the shortest path that the first AGV should follow during the obstacle avoidance process, ensuring that it can smoothly avoid the second AGV and travel safely.

[0107] Example 2: Based on the same inventive concept as the AGV bidirectional obstacle avoidance trajectory generation method integrating road condition perception in the aforementioned examples, this application provides an AGV bidirectional obstacle avoidance trajectory generation system integrating road condition perception. See [link to example]. Figure 4 As shown, the system includes:

[0108] The first driving interference identification unit 1 is used to identify driving interference based on the first real-time ground conditions obtained by the first AGV through vehicle-mounted LiDAR scanning, and to fit and output a first real-time safe speed; the second driving interference identification unit 2 is used to identify driving interference based on the second real-time ground conditions obtained by the second AGV through vehicle-mounted LiDAR scanning, and to fit and output a second real-time safe speed; the dynamic safe distance solving unit 3 is used to solve the dynamic safe distance based on the first real-time safe speed and the second real-time safe speed; the obstacle avoidance association data acquisition unit 4 is used to trigger the first AGV and the second AGV to acquire the first obstacle avoidance association data and the second obstacle avoidance association data respectively if the real-time conflict distance is less than the dynamic safe distance; the obstacle avoidance safety zone positioning unit 5 is used to calculate the asymmetric safe turning capability based on the first obstacle avoidance association data and the second obstacle avoidance association data, and to position the obstacle avoidance safety zone. A shared yielding safety zone; a simulated yielding route output unit 6, used to use the shared yielding safety zone as a yielding boundary constraint, and fit and output a first simulated yielding route and a second simulated yielding route; a yielding energy consumption prediction unit 7, used to predict the first yielding energy consumption and the second yielding energy consumption of the first simulated yielding route and the second simulated yielding route, respectively, with the yielding direction as a constraint; a role allocation unit 8, used to allocate the roles of the first AGV and the second AGV by comparing the first yielding energy consumption and the second yielding energy consumption, and to locate the real-time yielding route of the real-time yielding vehicle and the original driving trajectory of the non-yielding vehicle; a vehicle control unit 9, used to control the non-yielding vehicle to slow down on the original driving trajectory while controlling the real-time yielding vehicle to drive according to the real-time yielding route, until the real-time yielding vehicle and the non-yielding vehicle leave the shared yielding safety zone.

[0109] In one implementation, the obstacle avoidance associated data acquisition unit 4 is used to perform the following operation steps:

[0110] The first predetermined trajectory of the first AGV and the second predetermined trajectory of the second AGV are retrieved from the workshop logistics platform; the first real-time position of the first AGV and the second real-time position of the second AGV are obtained through the interactive vehicle UWB positioning system; the real-time conflict distance is calculated based on the first real-time position and the second real-time position, using the intersection segment of the first predetermined trajectory and the second predetermined trajectory as a constraint.

[0111] In one implementation, the first driving interference identification unit 1 is used to perform the following operation steps:

[0112] The vehicle-mounted lidar of the first AGV scans the ground of the channel at a frequency of 10Hz to obtain the first reflection intensity point cloud data. Based on the reflectivity distribution characteristics, the first reflection intensity point cloud data is segmented using a dual threshold method to identify driving interference features, thereby obtaining the first oil stain distribution area and the first metal debris distribution area. After spatially superimposing the first oil stain distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safe speed is fitted and output.

[0113] In one implementation, the first driving interference identification unit 1 is used to perform the following operation steps:

[0114] The system receives the first real-time transport load transmitted back by the load sensor of the first AGV; it receives the first real-time slope angle and the first real-time slope direction output by the IMU of the first AGV; it performs a safety adhesion attenuation analysis based on the first oil stain distribution area and the first metal debris distribution area to obtain the first oil stain attenuation coefficient and the first debris disturbance coefficient; it then constructs a dynamic adhesion model by spatially superimposing the first oil stain distribution area and the first metal debris distribution area, combined with the first real-time slope angle; based on the trajectory segment of the first predetermined trajectory at the first real-time position, it performs an adhesion safety analysis in the dynamic adhesion model and outputs the initial safety speed; it then compensates the initial safety speed with the first real-time slope direction and outputs the first real-time safety speed.

[0115] In one implementation, the first driving interference identification unit 1 is used to perform the following operation steps:

[0116] Background noise in the first reflection intensity point cloud data is filtered out by removing invalid points with a reflection intensity of <5%; discrete point cloud clusters with a reflection intensity >85% are extracted by traversing the first reflection intensity point cloud data; isolated specular points are removed by traversing the discrete point cloud clusters using a preset point cloud spatial density threshold, and the first metal debris distribution area is output; continuous point cloud clusters with a reflection intensity of 5% to 30% are extracted by traversing the first reflection intensity point cloud data; micro-region removal is performed by traversing the continuous point cloud clusters using a preset point cloud projection area threshold, and morphological closing operations are performed on the continuous point cloud clusters to fuse adjacent oil stains, thereby generating the first oil stain distribution area.

[0117] In one implementation, the dynamic safety distance solving unit 3 is used to perform the following operation steps:

[0118] Using the first real-time slope direction as a deceleration correction constraint, the first deceleration capacity is pre-calculated based on the first real-time transport load and the first real-time safe speed, and the second deceleration capacity of the second AGV is solved by analogy. A dominant safe speed is output by comparing the first real-time safe speed and the second real-time safe speed. A dominant deceleration is output by comparing the first deceleration capacity and the second deceleration capacity. A safe reaction distance is calculated based on a preset fixed response time and the dominant safe speed. A theoretical braking distance is calculated based on the dominant safe speed and the dominant deceleration. A first dynamic environmental margin is matched based on the first oil spill distribution area and the first metal debris distribution area. A second dynamic environmental margin is matched based on the second oil spill distribution area and the second metal debris distribution area. Based on the sum of the safe reaction distance and the theoretical braking distance, the maximum value of the first dynamic environmental margin and the second dynamic environmental margin is taken for safety distance compensation, and the dynamic safe distance is output.

[0119] In one implementation, the avoidance safety zone positioning unit 5 is used to perform the following operation steps:

[0120] When the first AGV is an uphill AGV and the second AGV is a downhill AGV: Front wheel load attenuation compensation is performed based on the first real-time transport load, the first real-time slope angle, and the first real-time slope direction, outputting a first minimum turning radius; rear wheel adhesion attenuation compensation is performed based on the second real-time transport load, the second real-time slope angle, and the second real-time slope direction, outputting a second minimum turning radius; based on the uphill expansion rule, the first minimum turning radius is mapped in the first real-time position space to generate a first dynamic envelope; based on the downhill expansion rule, the second minimum turning radius is mapped in the second real-time position space to generate a second dynamic envelope; the first and second dynamic envelopes are solved by union to obtain the conflict core area; after interactively obtaining the physical boundary of the transport channel, geometric Boolean operations are performed on the physical boundary of the transport channel and the core conflict area through convex polygon Boolean subtraction operations to output the shared avoidance safety zone.

[0121] In one implementation, the simulated yielding route output unit 6 is used to perform the following operation steps:

[0122] The shared obstacle avoidance safety zone space is used to divide the first predetermined trajectory and the second predetermined trajectory to obtain the first segmented trajectory and the second segmented trajectory. In the scenario where the first segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a cubic spline curve that satisfies the second minimum turning radius and is used as the output of the second simulated yield route. In the scenario where the second segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint to generate a shortest obstacle avoidance Bezier curve that satisfies the first minimum turning radius and is used as the output of the first simulated yield route.

[0123] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for generating bidirectional obstacle avoidance trajectories for AGVs that integrates road condition perception, characterized in that, The method includes: Based on the first real-time ground conditions obtained by the first AGV through on-board LiDAR scanning, driving interference is identified, and the first real-time safe speed is fitted and output. Based on the second real-time ground conditions obtained by the second AGV through on-board LiDAR scanning, driving interference is identified, and a second real-time safe speed is fitted and output. The dynamic safety distance is calculated based on the first and second real-time safety speeds. If the real-time collision distance is less than the dynamic safety distance, the first AGV and the second AGV are triggered to collect the first obstacle avoidance association data and the second obstacle avoidance association data, respectively. Based on the first obstacle avoidance association data and the second obstacle avoidance association data, calculate the asymmetric safe turning capability and locate the shared obstacle avoidance safety zone; Using the shared safety zone as a yield boundary constraint, the first simulated yield route and the second simulated yield route are fitted and output; Using the yielding direction as a constraint, predict the first yielding energy consumption and the second yielding energy consumption of the first simulated yielding route and the second simulated yielding route, respectively; By comparing the first yielding energy consumption and the second yielding energy consumption, the roles of the first AGV and the second AGV are assigned, and the real-time yielding route of the yielding vehicle and the original driving trajectory of the non-yielding vehicle are located. While controlling the real-time yielding vehicle's movement according to the real-time yielding route, the non-yielding vehicle is controlled to slow down along the original driving trajectory until the real-time yielding vehicle and the non-yielding vehicle leave the shared avoidance safety zone.

2. The method for generating bidirectional obstacle avoidance trajectory for AGVs based on road condition perception as described in claim 1, characterized in that, Also includes: Retrieve the first predetermined trajectory of the first AGV and the second predetermined trajectory of the second AGV from the workshop logistics platform; The first real-time position of the first AGV and the second real-time position of the second AGV are obtained through the interactive vehicle UWB positioning system. Using the intersection segment of the first predetermined trajectory and the second predetermined trajectory as a constraint, the real-time conflict distance is calculated based on the first real-time position and the second real-time position.

3. The method for generating bidirectional obstacle avoidance trajectory for AGVs by integrating road condition perception as described in claim 2, characterized in that, Based on the first real-time ground conditions obtained by the first AGV through onboard LiDAR scanning, driving interference is identified, and a first real-time safe speed is fitted and output, including: The vehicle-mounted lidar of the first AGV scans the ground of the channel at a frequency of 10Hz to obtain the first reflection intensity point cloud data; Based on the reflectivity distribution characteristics, the dual threshold segmentation method is used to identify driving interference features in the first reflectivity point cloud data to obtain the first oil stain distribution area and the first metal debris distribution area. After spatially superimposing the first oil stain distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safety speed is fitted and output.

4. The method for generating bidirectional obstacle avoidance trajectory for AGVs by integrating road condition perception as described in claim 3, characterized in that, After spatially superimposing the first oil spill distribution area and the first metal debris distribution area, an adhesion safety analysis is performed, and the first real-time safety velocity is fitted and output, including: The first real-time transport load is transmitted back by the load sensor connected to the first AGV. Receive the first real-time slope angle and the first real-time slope direction output by the IMU of the first AGV; Based on the first oil stain distribution area and the first metal debris distribution area, a safety adhesion attenuation analysis was performed to obtain the first oil stain attenuation coefficient and the first debris disturbance coefficient. After spatially superimposing the first oil stain distribution area and the first metal debris distribution area, a dynamic adhesion model is constructed by combining the first real-time slope angle. Based on the trajectory segment of the first predetermined trajectory at the first real-time position, an attachment safety analysis is performed on the dynamic attachment model, and an initial safety speed is output. The initial safe speed is compensated using the first real-time slope direction, and the first real-time safe speed is output.

5. The AGV bidirectional obstacle avoidance trajectory generation method integrating road condition perception as described in claim 3, characterized in that, Based on the reflectivity distribution characteristics, a dual-threshold segmentation method is used to identify driving interference features in the first reflectivity point cloud data, resulting in the first oil stain distribution area and the first metal debris distribution area, including: Background noise in the first reflection intensity point cloud data is filtered out by removing invalid points with a reflection intensity of less than 5%. Traverse the first reflection intensity point cloud data to extract discrete point cloud clusters with a reflection intensity > 85%; The discrete point cloud clusters are traversed using a preset point cloud spatial density threshold to remove isolated highlight points, and the distribution area of ​​the first metal debris is output. Extract continuous point cloud clusters with a reflection intensity of 5% to 30% by traversing the first reflection intensity point cloud data; After traversing the continuous point cloud clusters using a preset point cloud projection area threshold to remove micro-regions, morphological closing operations are performed on the continuous point cloud clusters to merge adjacent oil stains and generate the first oil stain distribution area.

6. The method for generating bidirectional obstacle avoidance trajectory for AGVs by integrating road condition perception as described in claim 4, characterized in that, The dynamic safety distance is calculated based on the first and second real-time safety velocities, including: Using the first real-time slope direction as a deceleration correction constraint, the first deceleration capacity is pre-calculated based on the first real-time transport load and the first real-time safe speed, and the second deceleration capacity of the second AGV is solved by analogy. By comparing the first real-time security speed and the second real-time security speed, the dominant security speed is output. By comparing the first deceleration capability and the second deceleration capability, the dominant deceleration is output; The safe reaction distance is calculated based on a preset fixed response time and the dominant safety speed. The theoretical braking distance is calculated based on the dominant safe speed and dominant deceleration. A first dynamic environmental margin is matched based on the first oil spill distribution area and the first metal debris distribution area; The second dynamic environmental margin is matched based on the second oil spill distribution area and the second metal debris distribution area. Based on the sum of the safe reaction distance and the theoretical braking distance, the maximum value of the first dynamic environment margin and the second dynamic environment margin is taken for safety distance compensation, and the dynamic safe distance is output.

7. The method for generating bidirectional obstacle avoidance trajectory for AGVs by integrating road condition perception as described in claim 4, characterized in that, Based on the first obstacle avoidance association data and the second obstacle avoidance association data, the asymmetric safe turning capability is calculated, and the shared obstacle avoidance safety zone is located, including: When the first AGV is an uphill AGV and the second AGV is a downhill AGV: Based on the first real-time transport load, the first real-time slope angle, and the first real-time slope direction, front wheel load attenuation compensation is performed, and the first minimum turning radius is output. The second minimum turning radius is output by compensating for the rear wheel adhesion attenuation based on the second real-time transport load, the second real-time slope angle, and the second real-time slope direction. Based on the uphill expansion rule, the first minimum turning radius is mapped in the first real-time location space to generate the first dynamic envelope. Based on the downhill expansion rule, the second minimum turning radius is mapped in the second real-time location space to generate a second dynamic envelope. The conflict core region is obtained by solving the union of the first dynamic envelope and the second dynamic envelope. After obtaining the physical boundary of the transportation channel interactively, geometric Boolean operations are performed on the physical boundary of the transportation channel and the core conflict zone through convex polygon Boolean subtraction operations to output the shared avoidance safety zone.

8. The method for generating bidirectional obstacle avoidance trajectory for AGVs by integrating road condition perception as described in claim 7, characterized in that, Using the shared safe zone as a yield boundary constraint, the first simulated yield route and the second simulated yield route are fitted and output, including: The shared avoidance safety zone spatially divides the first predetermined trajectory and the second predetermined trajectory to obtain the first segmented trajectory and the second segmented trajectory. In the scenario where the first segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint, and a cubic spline curve that satisfies the second minimum turning radius is generated as the output of the second simulated yield route. In the scenario where the second segmented trajectory is fixed, the shortest obstacle avoidance trajectory is fitted with the shared obstacle avoidance safety space as a constraint, and the shortest obstacle avoidance Bezier curve that satisfies the first minimum turning radius is generated as the first simulated yield route output.

9. An AGV bidirectional obstacle avoidance trajectory generation system integrating road condition perception, characterized in that, The system is used to implement the AGV bidirectional obstacle avoidance trajectory generation method integrating road condition perception as described in any one of claims 1-8, the system comprising: The first driving interference identification unit is used to identify driving interference based on the first real-time ground conditions obtained by the first AGV through vehicle-mounted lidar scanning, and to fit and output the first real-time safe speed. The second driving interference identification unit is used to identify driving interference based on the second real-time ground conditions obtained by the second AGV through vehicle-mounted lidar scanning, and to fit and output the second real-time safe speed. The dynamic safety distance calculation unit is used to calculate the dynamic safety distance based on the first real-time safety speed and the second real-time safety speed. The obstacle avoidance association data acquisition unit is used to trigger the first AGV and the second AGV to acquire the first obstacle avoidance association data and the second obstacle avoidance association data respectively if the real-time conflict distance is less than the dynamic safety distance. The obstacle avoidance safety zone positioning unit is used to calculate the asymmetric safe turning capability based on the first obstacle avoidance association data and the second obstacle avoidance association data, and to locate the shared obstacle avoidance safety zone. The simulated yield route output unit is used to fit and output the first simulated yield route and the second simulated yield route by taking the shared yield safety zone as the yield boundary constraint. The yielding energy consumption prediction unit is used to predict the first yielding energy consumption and the second yielding energy consumption of the first simulated yielding route and the second simulated yielding route, respectively, with the yielding direction as a constraint. The role allocation unit is used to allocate roles to the first AGV and the second AGV by comparing the first yielding energy consumption and the second yielding energy consumption, and to locate the real-time yielding route of the yielding vehicle and the original driving trajectory of the non-yielding vehicle. The vehicle control unit is used to control the non-yielding vehicle to slow down on the original driving trajectory while controlling the real-time yielding vehicle to drive according to the real-time yielding route, until the real-time yielding vehicle and the non-yielding vehicle leave the shared avoidance safety zone.

Images