Safety decision method, device and equipment for automatic guided vehicle

Abstract

Embodiments of the present application provide a safety decision method, device and equipment for an automated guided vehicle (AVG). The method comprises: first acquiring multi-source running information of the AVG, then constructing a vehicle body equivalent envelope model according to vehicle type information and a cargo carrying state, and combining path planning information to calculate a safe driving region of interest parameter in a Frenet coordinate system; then pre-processing obstacle point cloud data and projecting it into the Frenet coordinate system to extract obstacle point clouds located in the safe driving region of interest, thereby obtaining candidate obstacle point clouds; further, the longitudinal nearest obstacle and the lateral nearest obstacle are respectively calculated according to the candidate obstacle point clouds, and a corresponding speed limit value is calculated based on the position of the nearest obstacle relative to the vehicle body; finally, the safe driving speed of the AVG is determined according to the speed limit value. Through the above method, the AVG can flexibly adapt to complex and variable industrial scenes, and the accuracy of safety decision is improved.

Description

Technical Field

[0001] This application relates to the field of automated guided vehicles (AGVs), and more particularly to a safety decision-making method, apparatus, and equipment for AGVs. Background Technology

[0002] In the fields of intelligent manufacturing and industrial automation, Automated Guided Vehicles (AGVs) are widely used in warehousing logistics, production line transportation, and internal material handling within factories. AGVs achieve unmanned transportation of goods through autonomous navigation and path planning, and their operating environment typically includes complex static and dynamic obstacles. In dynamic industrial scenarios, AGVs need to perceive their surroundings in real time and make safe decisions to avoid collisions while maintaining transportation efficiency. For example, in a warehousing environment, AGVs need to navigate between dense shelving units and cope with dynamic interference from forklifts and personnel; in a production line scenario, AGVs need to precisely align between workstations and adapt to dynamic changes such as pallet loading and unloading and equipment debugging. Existing AGV safety protection technologies generally suffer from insufficient environmental perception coverage, rigid safety zone presets, and delayed response to dynamic obstacles, leading to misjudgments or omissions in complex scenarios, seriously threatening personnel safety and equipment operating efficiency. Therefore, there is an urgent need for a safety decision-making system with high-precision obstacle recognition and graded speed limiting capabilities to meet the safe operation requirements of AGVs in complex industrial scenarios.

[0003] In existing technical solutions, a common approach is to use a pre-set safety zone of a fixed shape (such as a circle or rectangle) and rely on safety radar or photoelectric sensors to detect obstacles entering different zones, triggering graded deceleration or stopping. The core logic is to set multiple safety thresholds (such as emergency stop zones and deceleration zones) based on the distance between the obstacle and the AGV, and directly control the AGV's actions through hardware input / output signals.

[0004] However, existing solutions suffer from low efficiency and difficulty in adapting flexibly to complex and ever-changing environments. Summary of the Invention

[0005] This application provides a safety decision-making method, apparatus, and equipment for automated guided vehicles, which solves the problems of low efficiency and difficulty in adapting to complex and ever-changing environments in the prior art.

[0006] In a first aspect, embodiments of this application provide a safety decision-making method for automated guided vehicles, including:

[0007] Obtain multi-source operational information of the AVG, which includes at least vehicle model information, path planning information, cargo status, and obstacle point cloud data;

[0008] Based on the vehicle model information and cargo status, an equivalent envelope model of the vehicle body is constructed, and combined with path planning information, the parameters of the region of interest for safe driving are calculated in the Frenet coordinate system.

[0009] The obstacle point cloud data is preprocessed and projected onto the Frenet coordinate system to extract candidate obstacle point clouds located within the safe driving region of interest, thus obtaining candidate obstacle point clouds.

[0010] Based on the candidate obstacle point cloud, the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction are obtained respectively, and the corresponding speed limit value is calculated based on the position of the nearest obstacle relative to the vehicle body.

[0011] The safe driving speed of the AVG is determined based on the speed limit value.

[0012] In one possible implementation, the step of constructing an equivalent envelope model of the vehicle body based on the vehicle model information and cargo status, and calculating the region of interest parameters for safe driving in the Frenet coordinate system by combining path planning information and pose information, includes:

[0013] Based on the vehicle model information, a physical model of the vehicle body is created using one or more geometric envelopes;

[0014] When the AVG is in a loaded state, the cargo size information is obtained, and when the cargo size is larger than the vehicle body size, the geometric envelope is updated with the cargo size to form an equivalent envelope model of the vehicle body in a loaded state.

[0015] The planned path is downsampled based on the path planning information, and a Frenet coordinate system is created based on the downsampled planned path.

[0016] The minimum longitudinal safety distance, minimum lateral safety width, minimum safety height, and minimum safety radius in spin mode are calculated in the Frenet coordinate system to form the region of interest parameters for safe driving.

[0017] In one possible implementation, the preprocessing of the obstacle point cloud data and its projection onto the Frenet coordinate system to extract candidate obstacle point clouds located within the safe driving region of interest includes:

[0018] Motion compensation is performed on the obstacle point cloud data based on the timestamp of the obstacle point cloud data and the current vehicle speed and angular velocity of the AVG to obtain the compensated obstacle point cloud data.

[0019] The compensated obstacle point cloud data is filtered, and the filtered obstacle point cloud data is converted to the vehicle coordinate system.

[0020] The obstacle point cloud data in the vehicle coordinate system is projected onto the Frenet coordinate system to obtain the longitudinal and lateral coordinates of each obstacle point cloud in the Frenet coordinate system.

[0021] Based on the longitudinal and lateral coordinates, obstacle point clouds located within the safe driving area of ​​interest are selected as candidate obstacle point clouds.

[0022] In one possible implementation, the step of filtering obstacle point clouds located within the safe driving area of ​​interest based on the longitudinal and lateral coordinates as candidate obstacle point clouds includes:

[0023] Obtain the current operating mode of the AVG, wherein the operating mode includes at least path tracking mode and spin mode;

[0024] When the AVG is in path tracking mode, a planned path curve is generated based on the path planning information, and the planned path curve is laterally expanded in combination with the minimum lateral safety width to form a path envelope region, and point clouds located outside the path envelope region in the Frenet coordinate system are removed.

[0025] When the AVG is in spin mode, a circular safety area centered on the vehicle body center is constructed based on the minimum safety radius, and point clouds located outside the circular safety area in the Frenet coordinate system are removed.

[0026] The remaining point cloud after elimination is determined as the candidate obstacle point cloud.

[0027] In one possible implementation, after extracting the point cloud of candidate obstacles located within the region of interest for safe driving, the method further includes:

[0028] When the AVG is in charging mode, the point cloud of the corresponding area is removed according to the location of the charging pile.

[0029] When the AVG is a backpack structure and is in a loaded state, the corresponding point cloud is removed according to the position of the cargo support structure.

[0030] When the AVG is in the workstation entry / exit state, the corresponding point cloud is removed according to the geometric range of the workstation area;

[0031] When the AVG has a fork structure, the corresponding point cloud rejection strategy is executed according to the different operation stages of picking or placing goods.

[0032] In one possible implementation, a corresponding point cloud removal strategy is executed based on different stages of the picking or placing operation, including:

[0033] When the AGV is in the picking and loading stage, the point cloud in the area in front of the forks or the workstation area is removed according to the radar installation position.

[0034] When the AGV is in the stage of placing goods into the workstation, the point cloud of the workstation area that is below the preset height range is removed according to the installation height of the pallet detection radar.

[0035] When the AGV is in the stage of picking up or placing goods out of the workstation, the point cloud of the area below the forks is removed based on the spatial position of the forks.

[0036] In one possible implementation, the step of determining the longitudinal nearest obstacle and the lateral nearest obstacle based on the candidate obstacle point cloud, and calculating the corresponding speed limit value based on the position of each obstacle relative to the vehicle body, includes:

[0037] In the Frenet coordinate system, the vehicle body equivalent envelope model is placed point by point along the path planning information from near to far, and it is detected whether the vehicle body equivalent envelope model has a geometric collision with the candidate obstacle point cloud.

[0038] When a geometric collision is detected for the first time, the obstacle at the corresponding collision location is identified as the nearest obstacle in the longitudinal direction, and the longitudinal distance of the nearest obstacle in the longitudinal direction is obtained.

[0039] For candidate obstacle point clouds with a lateral distance greater than the minimum lateral safety width, the point with the smallest comprehensive distance is determined as the lateral nearest obstacle according to the pre-constructed comprehensive distance function, and the lateral distance of the lateral nearest obstacle is obtained.

[0040] In the Frenet coordinate system, the corresponding longitudinal speed limit and lateral speed limit are calculated based on the longitudinal distance to the nearest longitudinal obstacle and the lateral distance to the nearest lateral obstacle, respectively.

[0041] The speed limit value with the smaller value between the longitudinal speed limit value and the lateral speed limit value is determined as the speed limit value corresponding to the candidate obstacle point cloud.

[0042] In one possible implementation, determining the safe driving speed of the AVG based on the speed limit value includes:

[0043] In the Frenet coordinate system, the area in front of the vehicle is divided into an emergency stop zone, a quick stop zone, a normal stop zone, and a normal driving zone according to the longitudinal distance.

[0044] When the nearest longitudinal obstacle is located within the emergency stop zone, a zero-speed command is generated.

[0045] When the nearest longitudinal obstacle is located in the fast parking zone or normal parking zone, the target deceleration value is calculated based on the current vehicle speed, system response time and preset deceleration, and a deceleration control command is generated.

[0046] When the obstacle is located within the normal driving zone, a speed limit control command is generated to maintain or adjust the vehicle speed.

[0047] The speed limit value is compared with the current driving speed of the AVG, and a corresponding control command is generated based on the comparison result to determine the safe driving speed of the AVG.

[0048] Secondly, embodiments of this application provide a safety decision-making device for an automated guided vehicle, comprising:

[0049] The acquisition module is used to acquire multi-source operating information of the AVG, which includes at least vehicle model information, path planning information, cargo status and obstacle point cloud data.

[0050] The first processing module is used to construct an equivalent envelope model of the vehicle body based on the vehicle model information and cargo status, and to calculate the parameters of the region of interest for safe driving in the Frenet coordinate system in combination with path planning information.

[0051] The second processing module is used to preprocess the obstacle point cloud data, project it onto the Frenet coordinate system, extract candidate obstacle point clouds located in the safe driving region of interest, and obtain candidate obstacle point clouds.

[0052] The calculation module is used to obtain the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction based on the candidate obstacle point cloud, and to calculate the corresponding speed limit value based on the position of the nearest obstacle relative to the vehicle body.

[0053] The determining module is used to determine the safe driving speed of the AVG based on the speed limit value.

[0054] In one possible implementation, the first processing module is specifically used for:

[0055] Based on the vehicle model information, a physical model of the vehicle body is created using one or more geometric envelopes;

[0056] When the AVG is in a loaded state, the cargo size information is obtained, and when the cargo size is larger than the vehicle body size, the geometric envelope is updated with the cargo size to form an equivalent envelope model of the vehicle body in a loaded state.

[0057] The planned path is downsampled based on the path planning information, and a Frenet coordinate system is created based on the downsampled planned path.

[0058] The minimum longitudinal safety distance, minimum lateral safety width, minimum safety height, and minimum safety radius in spin mode are calculated in the Frenet coordinate system to form the region of interest parameters for safe driving.

[0059] In one possible implementation, the second processing module is specifically used for:

[0060] Motion compensation is performed on the obstacle point cloud data based on the timestamp of the obstacle point cloud data and the current vehicle speed and angular velocity of the AVG to obtain the compensated obstacle point cloud data.

[0061] The compensated obstacle point cloud data is filtered, and the filtered obstacle point cloud data is converted to the vehicle coordinate system.

[0062] The obstacle point cloud data in the vehicle coordinate system is projected onto the Frenet coordinate system to obtain the longitudinal and lateral coordinates of each obstacle point cloud in the Frenet coordinate system.

[0063] Based on the longitudinal and lateral coordinates, obstacle point clouds located within the safe driving area of ​​interest are selected as candidate obstacle point clouds.

[0064] In one possible implementation, the second processing module is specifically used for:

[0065] Obtain the current operating mode of the AVG, wherein the operating mode includes at least path tracking mode and spin mode;

[0066] When the AVG is in path tracking mode, a planned path curve is generated based on the path planning information, and the planned path curve is laterally expanded in combination with the minimum lateral safety width to form a path envelope region, and point clouds located outside the path envelope region in the Frenet coordinate system are removed.

[0067] When the AVG is in spin mode, a circular safety area centered on the vehicle body center is constructed based on the minimum safety radius, and point clouds located outside the circular safety area in the Frenet coordinate system are removed.

[0068] The remaining point cloud after elimination is determined as the candidate obstacle point cloud.

[0069] In one possible implementation, the device further includes a filtering module, which, after extracting the point cloud of candidate obstacles located within a safe driving region of interest, is used to:

[0070] When the AVG is in charging mode, the point cloud of the corresponding area is removed according to the location of the charging pile.

[0071] When the AVG is a backpack structure and is in a loaded state, the corresponding point cloud is removed according to the position of the cargo support structure.

[0072] When the AVG is in the workstation entry / exit state, the corresponding point cloud is removed according to the geometric range of the workstation area;

[0073] When the AVG has a fork structure, the corresponding point cloud rejection strategy is executed according to the different operation stages of picking or placing goods.

[0074] In one possible implementation, the filtering module is specifically used for:

[0075] When the AGV is in the picking and loading stage, the point cloud in the area in front of the forks or the workstation area is removed according to the radar installation position.

[0076] When the AGV is in the stage of placing goods into the workstation, the point cloud of the workstation area that is below the preset height range is removed according to the installation height of the pallet detection radar.

[0077] When the AGV is in the stage of picking up or placing goods out of the workstation, the point cloud of the area below the forks is removed based on the spatial position of the forks.

[0078] In one possible implementation, the computing module is specifically used for:

[0079] In the Frenet coordinate system, the vehicle body equivalent envelope model is placed point by point along the path planning information from near to far, and it is detected whether the vehicle body equivalent envelope model has a geometric collision with the candidate obstacle point cloud.

[0080] When a geometric collision is detected for the first time, the obstacle at the corresponding collision location is identified as the nearest obstacle in the longitudinal direction, and the longitudinal distance of the nearest obstacle in the longitudinal direction is obtained.

[0081] For candidate obstacle point clouds with a lateral distance greater than the minimum lateral safety width, the point with the smallest comprehensive distance is determined as the lateral nearest obstacle according to the pre-constructed comprehensive distance function, and the lateral distance of the lateral nearest obstacle is obtained.

[0082] In the Frenet coordinate system, the corresponding longitudinal speed limit and lateral speed limit are calculated based on the longitudinal distance to the nearest longitudinal obstacle and the lateral distance to the nearest lateral obstacle, respectively.

[0083] The speed limit value with the smaller value between the longitudinal speed limit value and the lateral speed limit value is determined as the speed limit value corresponding to the candidate obstacle point cloud.

[0084] In one possible implementation, the determining module is specifically used for:

[0085] In the Frenet coordinate system, the area in front of the vehicle is divided into an emergency stop zone, a quick stop zone, a normal stop zone, and a normal driving zone according to the longitudinal distance.

[0086] When the nearest longitudinal obstacle is located within the emergency stop zone, a zero-speed command is generated.

[0087] When the nearest longitudinal obstacle is located in the fast parking zone or normal parking zone, the target deceleration value is calculated based on the current vehicle speed, system response time and preset deceleration, and a deceleration control command is generated.

[0088] When the obstacle is located within the normal driving zone, a speed limit control command is generated to maintain or adjust the vehicle speed.

[0089] The speed limit value is compared with the current driving speed of the AVG, and a corresponding control command is generated based on the comparison result to determine the safe driving speed of the AVG.

[0090] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0091] The memory stores computer-executed instructions;

[0092] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0093] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0094] The safety decision-making method, apparatus, equipment, and medium for automated guided vehicles (AGVs) provided in this application first acquire multi-source operational information of the AGV, then construct an equivalent envelope region that accurately reflects the actual space occupied by the vehicle based on the operational information; simultaneously, a reference trajectory is established using path planning information, and the vehicle's real-time pose information is mapped to the Frenet coordinate system corresponding to the reference trajectory, thereby describing the relative positional relationship between the vehicle and its surrounding environment in the path direction and the lateral direction, respectively. Based on this, the region of interest parameters for safe driving of the vehicle in the current driving state are calculated according to the spatial relationship between the vehicle envelope model and the planned path, that is, the minimum drivable space range within which the vehicle can maintain safe passage in the longitudinal and lateral directions is determined, thus providing spatial constraints for subsequent obstacle screening. Subsequently, the environmental point cloud data acquired by the environmental perception system is preprocessed to improve the reliability and computational stability of the point cloud data, and the processed point cloud data is projected onto the Frenet coordinate system, so that environmental obstacle information can be expressed in terms of longitudinal distance and lateral offset relative to the planned path. Furthermore, spatial filtering of the point cloud data is performed in the Frenet coordinate system to extract candidate obstacle point clouds located within the region of interest for safe driving. This reduces the interference of irrelevant environmental information on decision-making calculations and improves the computational efficiency of subsequent obstacle recognition and speed decision-making. Then, the nearest longitudinal and lateral obstacles are calculated based on the candidate obstacle point clouds, and the corresponding safe speed limit is calculated based on the positional relationship of each obstacle relative to the vehicle body. Finally, the speed limit values ​​corresponding to each obstacle are comprehensively compared, and the minimum speed limit value is selected as the safe driving speed of the AGV. This allows the vehicle to adapt to changes in the distribution of environmental obstacles in real time while maintaining operational efficiency, achieving dynamic constraints and active control of driving safety risks. Attached Figure Description

[0095] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0096] Figure 1 A flowchart illustrating the safety decision-making method for automated guided vehicles provided in this application embodiment. Figure 1 ;

[0097] Figure 2 A flowchart illustrating the safety decision-making method for automated guided vehicles provided in this application embodiment. Figure 2 ;

[0098] Figure 3 A schematic diagram of the equivalent envelope model of the automated guided vehicle provided in the embodiments of this application;

[0099] Figure 4 A schematic diagram illustrating the safety area screening of an automated guided vehicle provided in this application embodiment;

[0100] Figure 5 A schematic diagram of point cloud data filtering for an automated guided vehicle in charging pile mode, provided in an embodiment of this application.

[0101] Figure 6 A schematic diagram of point cloud data filtering for an automated guided vehicle in working mode, provided in an embodiment of this application.

[0102] Figure 7 A schematic diagram of an automated guided vehicle provided in an embodiment of this application within different speed-limited zones;

[0103] Figure 8 A schematic diagram of the structure of the safety decision-making device for an automated guided vehicle provided in an embodiment of this application;

[0104] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0105] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0106] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0107] In the fields of intelligent manufacturing and industrial automation, Automated Guided Vehicles (AGVs) are widely used in warehousing logistics, production line transportation, and internal material handling within factories. AGVs achieve unmanned transportation of goods through autonomous navigation and path planning, and their operating environment typically includes complex static and dynamic obstacles. In dynamic industrial scenarios, AGVs need to perceive their surroundings in real time and make safe decisions to avoid collisions while maintaining transportation efficiency. For example, in a warehousing environment, AGVs need to navigate between dense shelving units and cope with dynamic interference from forklifts and personnel; in a production line scenario, AGVs need to precisely align between workstations and adapt to dynamic changes such as pallet loading and unloading and equipment debugging. Existing AGV safety protection technologies generally suffer from insufficient environmental perception coverage, rigid safety zone presets, and delayed response to dynamic obstacles, leading to misjudgments or omissions in complex scenarios, seriously threatening personnel safety and equipment operating efficiency. Therefore, there is an urgent need for a safety decision-making system with high-precision obstacle recognition and graded speed limiting capabilities to meet the safe operation requirements of AGVs in complex industrial scenarios.

[0108] In existing technical solutions, a common approach is to use a pre-set safety zone of a fixed shape (such as a circle or rectangle) and rely on safety radar or photoelectric sensors to detect obstacles entering different zones, triggering graded deceleration or stopping. The core logic is to set multiple safety thresholds (such as emergency stop zones and deceleration zones) based on the distance between the obstacle and the AGV, and directly control the AGV's actions through hardware input / output signals.

[0109] However, existing solutions cannot adapt to complex dynamic environments. For example, in densely packed shelving areas or workstation alignment scenarios, fixed areas are prone to coverage blind spots or overly conservative speed limits. Furthermore, new scenarios require reconfiguration of sensors and area parameters, resulting in high debugging costs and difficulty in dealing with sudden obstacles (such as temporarily placed equipment). In addition, dedicated safety radars need to be deployed, which are costly and complex to maintain, limiting the flexibility and scalability of the system.

[0110] Based on this, this application proposes a safety decision-making method for automated guided vehicles (AGVs). The inventors, based on vehicle kinematic constraints and path coordinate analysis principles, unify the coordinate representation of the vehicle's operating state and planned path, enabling obstacle positions to be described by longitudinal distance and lateral offset relative to the path. This achieves precise characterization of the vehicle's forward travel space and lateral safety margin. In the specific implementation process, firstly, multi-source operating information of the AGV is acquired, and an equivalent envelope model of the vehicle body is constructed based on vehicle parameters and cargo status to realistically reflect the actual space occupancy of the vehicle under different operating conditions. Subsequently, combined with path planning information, the safe driving region of interest corresponding to the vehicle's current state is calculated in the Frenet coordinate system. Environmental point cloud data is preprocessed and projected into this coordinate system to extract candidate obstacle point clouds located within the safe driving region of interest. Further, the longitudinal nearest obstacle and the lateral nearest obstacle are calculated based on the candidate obstacle point clouds, and the corresponding safe speed limit is calculated based on the positional relationship of the nearest obstacle relative to the vehicle body. Finally, the safe driving speed of the AGV is determined based on multiple speed limit values. The above methods enable safety decisions to no longer rely on fixed-shape safety zones or simple distance thresholds, but to be dynamically calculated based on vehicle structural characteristics, operating status, and the distribution of environmental obstacles. This allows for more refined and adaptive speed control, which not only improves the safety of AGVs in complex environments but also reduces unnecessary deceleration or stopping, thereby improving overall operating efficiency.

[0111] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0112] It should be noted that this application provides complete data support for the safety decision-making method through the collaborative work of multiple modules. The entire AGV safety decision-making system includes vehicle model information files, topology map files, a scheduling system, and on-vehicle perception, planning, control, positioning, and drive modules. Specifically, the vehicle model information file describes the structural dimensions of the AGV body, including parameters such as body length, width, height, and fork or superstructure dimensions, used to construct an equivalent envelope model of the vehicle body. The topology map file describes the path network information in the AGV's operating environment, containing multiple road segment information and point information. Road segment information includes at least a road segment ID, road segment length, start coordinates, and end coordinates. Point information includes at least a point ID, point attributes, and point coordinates. Point attributes include at least workstations, charging points, and parking points. The scheduling system is used to allocate tasks and schedule paths for the AGV according to production tasks, and sends target point IDs and related task information to the on-vehicle terminal. The perception module receives raw environmental point cloud data collected by sensors such as LiDAR and 3D cameras, and preprocesses the raw point cloud data, including noise point cloud removal, vehicle body point cloud removal, cargo point cloud removal, point cloud downsampling, and coordinate system transformation, to obtain environmental point cloud data in the vehicle coordinate system, which is then sent to the safety decision module. The planning module generates the vehicle's driving trajectory based on task information issued by the scheduling system, parses the target point ID and cargo size information in the task, and provides the planned trajectory information to the control and safety decision modules. The control module controls the vehicle's motion based on the trajectory generated by the planning module and publishes real-time vehicle control status information, including at least control status, control mode, and direction of travel; control modes include at least path tracking mode and spin mode. The positioning module outputs the vehicle's current pose information in real-time, including the vehicle's position and attitude in the map coordinate system. The drive module provides the vehicle's motion status information, including current speed, angular velocity, and superstructure status information, including at least lifting height, superstructure angle, and load status. Through the collaborative work of the above modules, the multi-source operational information in step S101 can be generated, providing basic data support for subsequent safe area calculation, point cloud screening, and safe driving speed calculation.

[0113] Figure 1 A flowchart illustrating the safety decision-making method for automated guided vehicles provided in this application embodiment. Figure 1 ;like Figure 1 As shown, the method includes:

[0114] S101. Obtain multi-source runtime information of AVG.

[0115] Among them, multi-source operational information includes at least vehicle type information, route planning information, cargo status, and obstacle point cloud data.

[0116] In this embodiment, the safe speed decision of the Automated Guided Vehicle (AGV) relies on the vehicle's own state information and environmental perception information. Therefore, it is necessary to first acquire multi-source operational information in real time through various sensors deployed on the AGV and onboard controllers. Specifically:

[0117] Vehicle type information describes the structural parameters of the AGV, including body length, width, height, wheelbase, track width, and turning radius; path planning information is the planned path output by the path planning module, which can be represented as a discrete path point sequence or a continuous curve; pose information indicates the real-time position and orientation of the AGV in the global coordinate system, generally provided by a positioning system (such as SLAM positioning, laser positioning, or vision positioning system); cargo status indicates whether the AGV is currently loaded and includes cargo size information; obstacle point cloud data is a three-dimensional obstacle point cloud collected by LiDAR or depth sensors.

[0118] Understandably, this embodiment achieves comprehensive perception of the AGV's own state and the external environment, laying the foundation for subsequent accurate modeling and decision-making. Through a redundant multi-sensor layout, even if a single sensor fails, basic information can still be acquired, improving the system's robustness.

[0119] It should also be noted that after acquiring multi-source operational information, it is necessary to accurately identify the specific operating scenario of the AGV, because the safety risk characteristics of different working conditions (such as picking up and placing goods at workstations, automatic charging, and driving on main roads) are significantly different. Therefore, this application adopts a dynamic judgment mechanism for working modes based on target point attributes and real-time poses to provide key contextual information for subsequent safety area calculation, point cloud removal, and speed decision-making. Specifically: if the target point ID of the current AGV is a workstation point, and the distance between the current AGV position and the target point ID is less than the set threshold for entering the workstation, and the pose angle between the AGV and the target point ID workstation is less than the set threshold for entering the workstation angle, then the AGV is considered to be in workstation mode at this time. Alternatively, if the previous target point ID of the current AGV is a workstation point, and the distance between the current AGV position and the previous target point ID is less than the set threshold for entering the workstation, and the pose angle between the AGV and the previous target point ID workstation is less than the set threshold for entering the workstation angle, then the AGV is also considered to be in workstation mode at this time. If the workstation mode is not met, the system determines whether it is in charging mode. Specifically: if the current AGV's target point ID is a charging point, the distance between the current AGV and the target point ID is less than the set threshold for entering the charging point, and the pose angle between the AGV and the target point ID is less than the set threshold for entering the charging point, then the AGV is considered to be in charging mode. Alternatively, if the current AGV's previous target point ID is a charging point, the distance between the current AGV and the previous target point ID is less than the set threshold for entering the charging point, and the pose angle between the AGV and the previous target point ID is less than the set threshold for entering the charging point, then the AGV is considered to be in charging mode. If neither of the above conditions is met, the AGV is considered to be in main road mode. If in workstation mode, the system calculates the distance between the current vehicle's pose and the target workstation ID in real time to determine whether the current state is entering or leaving the workstation. Similarly, if in charging mode, the system calculates the distance between the current vehicle's pose and the target charging pile ID in real time to determine whether the current state is entering or leaving the charging pile.

[0120] S102. Construct an equivalent envelope model of the vehicle body based on vehicle type information and cargo status, and calculate the parameters of the region of interest for safe driving in the Frenet coordinate system in combination with path planning information.

[0121] In one possible implementation, the vehicle body is first physically modeled using one or more geometric envelopes based on vehicle model information. Then, when the AVG is in a loaded state, the cargo size information is obtained, and when the cargo size is larger than the vehicle body size, the geometric envelope is updated with the cargo size to form an equivalent envelope model of the vehicle body in a loaded state. Next, the planned path is downsampled based on the path planning information, and a Frenet coordinate system is created based on the downsampled planned path. Finally, the minimum longitudinal safety distance, minimum lateral safety width, minimum safety height, and minimum safety radius in spin mode are calculated in the Frenet coordinate system to form the region of interest parameters for safe driving.

[0122] It should be understood that, in this embodiment, to accurately describe the space occupied by the AGV during its movement, an equivalent envelope model of the vehicle body needs to be constructed. First, the AGV body is geometrically modeled based on the vehicle model information. For example, a rectangular envelope, a cylindrical envelope, or a combination of multiple envelopes can be used to approximate the AGV body structure. When the AGV is carrying a load, the impact of the cargo on the vehicle's external dimensions needs to be considered. For example, when the cargo size is larger than the vehicle body size, the original geometric envelope is updated with the cargo size to form a new equivalent envelope model of the vehicle body.

[0123] Subsequently, a Frenet coordinate system is established based on the path planning information. The Frenet coordinate system is a curved coordinate system referenced to the planned path. In this system, the s-axis represents the longitudinal distance along the path, and the l-axis represents the lateral offset distance relative to the path. Combining the pose information, the current vehicle center is projected onto the Frenet coordinate system to obtain the current vehicle's longitudinal and lateral coordinates.

[0124] Based on this, the parameters of the region of interest for safe driving can be obtained. The specific calculation formula is as follows:

[0125]

[0126] It should be understood that, in the Frenet coordinate system, the region of interest for safe driving is not a simple geometric shape, but a dynamic space constrained by parameters in four dimensions: longitudinal, lateral, vertical, and rotational. Its calculation logic is as follows:

[0127] Minimum longitudinal safety distance Defined as the sum of the distance from the vehicle control center to the leading edge of the vehicle box in the forward direction and the set safety distance. The former depends on the geometric relationship between the vehicle center and the leading edge in the vehicle structure (such as the fork tips of a forklift, the front bumper), while the latter is dynamically calculated based on the current vehicle speed, system response time, and preset deceleration. This parameter ensures that the AGV has sufficient braking distance when it detects an obstacle in front, thus avoiding a collision.

[0128] Minimum lateral safety width This is half the width of the vehicle body box. In the Frenet coordinate system, the lateral safety boundary is defined as... That is, the width of the vehicle body is extended to both the left and right sides of the path.

[0129] Minimum safe height The calculation is performed dynamically using conditional expressions. When the AGV is unloaded, It equals the vehicle height; when loaded, it adds the vehicle height, pallet thickness, and cargo height to filter out point clouds above the vehicle and cargo in the vertical direction (e.g., to avoid misjudging high-bay rack beams as obstacles) or to detect the risk of overhead obstacles.

[0130] In spin mode, the minimum safe radius is defined as the radius of the smallest circumcircle of the vehicle body box, which is the distance from the center of the vehicle body to the farthest corner point. When the AGV is in spin mode, a circular safe area is constructed using this radius.

[0131] Understandably, by dynamically constructing an equivalent envelope model of the vehicle body, the system can accurately reflect the actual space occupied by the AGV under different loading states and different superstructure postures, avoiding decreased traffic efficiency due to an overly conservative model or collision risks due to an overly optimistic model. Furthermore, the Frenet coordinate system transforms complex two-dimensional or three-dimensional spatial problems into two-dimensional problems along the path and lateral directions, simplifying obstacle detection and safe zone calculation, thus achieving a balance between safety and traffic efficiency.

[0132] S103. Preprocess the obstacle point cloud data and project it onto the Frenet coordinate system to extract candidate obstacle point clouds located within the safe driving region of interest, thus obtaining candidate obstacle point clouds.

[0133] It should be noted that the specific implementation method of this step will be discussed later. Figure 2 The details are elaborated in the embodiments and will not be described in detail here.

[0134] Understandably, preprocessing environmental point cloud data can significantly reduce the computational load on the point cloud data, thereby quickly determining whether an obstacle is located near the vehicle's path and improving obstacle detection efficiency.

[0135] S104. Based on the candidate obstacle point cloud, find the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction, and calculate the corresponding speed limit value based on the position of the nearest obstacle relative to the vehicle body.

[0136] In one possible implementation, firstly, in the Frenet coordinate system, the vehicle body equivalent envelope model is placed point by point along the path planning information from near to far, and it is detected whether the vehicle body equivalent envelope model has a geometric collision with the candidate obstacle point cloud. Then, when a geometric collision is detected for the first time, the obstacle at the corresponding collision position is identified as the longitudinal nearest obstacle, and the longitudinal distance of the longitudinal nearest obstacle is obtained. Then, for the candidate obstacle point cloud with a lateral distance greater than the minimum lateral safety width, the point with the smallest comprehensive distance is identified as the lateral nearest obstacle according to the pre-constructed comprehensive distance function, and the lateral distance of the lateral nearest obstacle is obtained. Further, in the Frenet coordinate system, the corresponding longitudinal speed limit value and lateral speed limit value are calculated according to the longitudinal distance of the longitudinal nearest obstacle and the lateral distance of the lateral nearest obstacle, respectively. Finally, the speed limit value with the smaller value between the longitudinal speed limit value and the lateral speed limit value is determined as the speed limit value corresponding to the candidate obstacle point cloud.

[0137] It should be understood that, in the Frenet coordinate system, starting from the current vehicle position (s=0), the vehicle moves along the positive s-axis (the path ahead) with a preset step size. The vehicle body's equivalent envelope model is placed point by point. For each placement position, it is checked whether the envelope model at that position has a geometric collision with the candidate obstacle point cloud, i.e., whether the envelope contains the point cloud. When a collision is detected for the first time, the longitudinal distance corresponding to the collision point is recorded, and the obstacle at that position is identified as the longitudinally nearest obstacle. If there is no collision throughout the process, the longitudinally nearest obstacle does not exist.

[0138] Meanwhile, for all candidate obstacle point clouds, the lateral coordinate l is greater than the minimum lateral safe width. Obstacles located outside the lateral safety boundary but still potentially affecting safety (such as rearview mirrors, protruding cargo, etc.) are filtered using a pre-built comprehensive distance function. By calculating the comprehensive distance of each candidate point, the point with the smallest comprehensive distance is determined as the lateral nearest obstacle, and the lateral distance of the lateral nearest obstacle is recorded.

[0139] The specific calculation formula is as follows:

[0140]

[0141] Where k is the influence factor used to adjust the vertical distance, l is the horizontal distance, and s is the vertical distance.

[0142] Furthermore, after determining the nearest longitudinal and lateral obstacles, the corresponding safe speed limit needs to be calculated based on their positional relationship relative to the AGV body. In one implementation, in the Frenet coordinate system, the longitudinal distance to the nearest longitudinal obstacle and the lateral distance to the nearest lateral obstacle are first obtained and compared with the pre-calculated minimum longitudinal safe distance and minimum lateral safe width to reflect the relative relationship between the obstacle and the vehicle's safe driving area. In the longitudinal direction, a longitudinal safe distance model is established based on the vehicle's current speed, system response time, and maximum deceleration, and the longitudinal speed limit is calculated in conjunction with the longitudinal distance to the nearest longitudinal obstacle. That is, the closer the nearest longitudinal obstacle, the lower the vehicle's permissible speed needs to be to ensure sufficient braking distance. In the lateral direction, since the AGV may still experience some lateral offset or path correction during path tracking, the impact of obstacles on the vehicle's movement safety needs to be considered. The lateral speed limit can be calculated by obtaining the lateral distance to the nearest lateral obstacle and combining it with the minimum lateral safety width. After obtaining the longitudinal and lateral speed limits respectively, the smaller of the two values ​​is taken as the final speed limit to ensure the safety of the vehicle in both directions.

[0143] Understandably, collision detection by placing envelope models point by point can accurately identify obstacles that are actually likely to collide with the path ahead, avoiding false alarms that may occur from simply judging based on point cloud distance. The introduction of lateral speed limits compensates for the shortcomings of traditional methods that only consider longitudinal obstacles, effectively addressing the risk of side scrapes. The final speed limit value is the smaller of the two, achieving comprehensive safety protection.

[0144] S105. Determine the safe driving speed of the AVG based on the speed limit value.

[0145] In one possible implementation, firstly, in the Frenet coordinate system, the area in front of the vehicle is divided into an emergency stop zone, a quick stop zone, a normal stop zone, and a normal driving zone according to longitudinal distance. Then, when the nearest longitudinal obstacle is located in the emergency stop zone, a zero-speed command is generated. When the nearest longitudinal obstacle is located in the quick stop zone or the normal stop zone, a target deceleration value is calculated based on the current vehicle speed, system response time, and preset deceleration, and a deceleration control command is generated. When the obstacle is located in the normal driving zone, a speed limit control command is generated to maintain or adjust the vehicle speed. Finally, the speed limit value is compared with the current driving speed of the AVG, and a corresponding control command is generated based on the comparison result to determine the safe driving speed of the AVG.

[0146] It should be noted that, in order to adopt differentiated responses to risks at different distances and avoid the efficiency problems caused by the "one-size-fits-all" approach in the traditional model, this application divides the area in front of the vehicle into multiple zones according to the longitudinal distance. The emergency stop zone corresponds to extreme dangers that require immediate braking, the rapid stop zone requires urgent but controllable deceleration, the normal stop zone allows for gradual deceleration, and the normal driving zone maintains speed. This maximizes the smoothness of AGV passage and operational efficiency while ensuring safety.

[0147] It should be understood that if the nearest longitudinal obstacle is located in the emergency stop zone, a zero-speed command is generated to control the AGV to brake and stop immediately to ensure that no collision or other dangerous situations occur; if it is located in the fast stop zone or normal stop zone, the target deceleration value is calculated based on the current vehicle speed, system response time, and preset deceleration, and a deceleration control command is generated to make the AGV decelerate to the target deceleration value with appropriate acceleration; if it is located in the normal driving zone, a speed limit control command is generated to maintain the current speed for driving. Furthermore, the speed limit value calculated in S104 is compared with the current driving speed of the AGV.

[0148] If the speed limit is less than the current driving speed, a deceleration command is generated to control the AGV to decelerate to the speed limit at a preset deceleration rate; if the speed limit is greater than the current driving speed, a speed maintenance or acceleration command is generated to control the AGV to maintain the current speed or accelerate to the speed limit. At the same time, the aforementioned preliminary command is used as a supplementary constraint. For example, if the nearest longitudinal obstacle is located in the emergency stop zone, the AGV should stop immediately even if the speed limit is greater than 0.

[0149] Understandably, this embodiment converts the safe speed limit value into an actually executable speed control command, and combined with the division of longitudinal risk zones, realizes a multi-level safety response mechanism. The emergency stop zone corresponds to the highest risk and forces a stop; the fast / normal stop zone corresponds to medium to high risk and actively decelerates; the normal driving zone corresponds to low risk and allows normal driving. Combined with the speed limit value calculated in embodiment S104, it ensures that even in the absence of longitudinal obstacles, lateral obstacles can trigger reasonable deceleration.

[0150] The safety decision-making method for automated guided vehicles (AGVs) provided in this application first acquires multi-source operational information of the AGV, then constructs an equivalent envelope region that accurately reflects the actual space occupied by the vehicle based on the operational information. Simultaneously, a reference trajectory is established using path planning information, and the vehicle's real-time pose information is mapped to the Frenet coordinate system corresponding to the reference trajectory, thereby describing the relative positional relationship between the vehicle and its surrounding environment in the path direction and the lateral direction, respectively. Based on this, the region of interest parameters for safe driving in the current driving state are calculated according to the spatial relationship between the vehicle envelope model and the planned path, i.e., determining the drivable space range within which the vehicle can maintain safe passage in the longitudinal and lateral directions, thus providing spatial constraints for subsequent obstacle screening. Subsequently, the environmental point cloud data acquired by the environmental perception system is preprocessed to improve the reliability and computational stability of the point cloud data, and the processed point cloud data is projected onto the Frenet coordinate system, enabling environmental obstacle information to be expressed in terms of longitudinal distance and lateral offset relative to the planned path. Furthermore, spatial filtering of the point cloud data is performed in the Frenet coordinate system to extract candidate obstacle point clouds located within the region of interest for safe driving. This reduces the interference of irrelevant environmental information on decision-making calculations and improves the computational efficiency of subsequent obstacle recognition and speed decision-making. Then, the nearest longitudinal and lateral obstacles are calculated based on the candidate obstacle point clouds, and the corresponding safe speed limit is calculated based on the positional relationship of each obstacle relative to the vehicle body. Finally, the speed limit values ​​corresponding to each obstacle are comprehensively compared, and the minimum speed limit value is selected as the safe driving speed of the AGV. This allows the vehicle to adapt to changes in the distribution of environmental obstacles in real time while maintaining operational efficiency, achieving dynamic constraints and active control of driving safety risks.

[0151] Figure 2 A flowchart illustrating the safety decision-making method for automated guided vehicles provided in this application embodiment. Figure 2 ;like Figure 2 As shown, in this embodiment... Figure 1 Based on the embodiments, the process of calculating the parameters of the region of interest for safe driving is described in detail. The method includes:

[0152] S201. Perform motion compensation on the obstacle point cloud data based on the timestamp of the obstacle point cloud data and the current vehicle speed and angular velocity of the AVG to obtain the compensated obstacle point cloud data.

[0153] Since scanning by lidar or stereo cameras takes a certain amount of time (e.g., the scanning cycle of one frame of 3D radar is 100ms), the movement of the AGV itself during this period will cause point cloud distortion. That is, different points in the same frame of point cloud actually correspond to the AGV pose at different times.

[0154] In one possible approach, the timestamp t of each point in each frame of point cloud data can be obtained first. i and the reference time t of the point cloud in that frame. ref (Usually the frame start time or end time); then obtain the AGV in [t] ref , t i The vehicle speed v(t) and angular velocity ω(t) within the time interval; then for each point, based on its time offset Δt = t i -t ref Calculate the displacement increment Δp and rotation increment ΔR of the AGV during this time period; finally, convert the original coordinates p of the point. i Compensation to t using Δp and ΔR ref Under a unified coordinate system at any given time, the compensated point cloud coordinates p are obtained. i' =ΔR×p i +Δp.

[0155] Understandably, after motion compensation, the point cloud data is precisely aligned with the current pose of the AGV, which can significantly improve the quality of the point cloud, avoid false detections or missed detections caused by distortion, and lay the foundation for subsequent accurate obstacle detection.

[0156] S202. Filter the compensated obstacle point cloud data and convert the filtered obstacle point cloud data to the vehicle coordinate system.

[0157] In one feasible approach, the compensated point cloud data undergoes the following filtering operations: First, statistical filtering or radius filtering is used to remove obviously isolated and sparse noise points. For example, for each point, the number of points in its neighborhood is counted; if it is less than a preset threshold, it is identified as an outlier and removed. Then, voxel filtering (VF) is used to divide the space into a fixed-size cubic grid, retaining only one centroid point in each grid to reduce the point cloud density and subsequent computational load. Finally, based on a preset coarse range (e.g., within 30m around the AGV), distant irrelevant point clouds are removed to further reduce the data volume.

[0158] Understandably, filtering point cloud data can remove invalid data (noise, outliers) and redundant data (overly dense point clouds), improving data quality while preserving key environmental features and significantly reducing the computational burden of subsequent algorithms.

[0159] It should also be noted that when a forklift AGV performs a loaded transport task, the pallet and its supporting structure (i.e., the material legs) are key components for supporting the goods, and their spatial position changes in real time with the lifting height and tilt angle of the forks. If the entire point cloud containing the material legs is directly retained, it is very easy for the AGV to misjudge the material legs as obstacles during turning, approaching workstations, or obstacle avoidance, causing unnecessary sudden stops or decelerations, which seriously affects the smoothness of operations. However, the material legs, as part of the goods, are a permissible structure during transportation and should not trigger a safety response. Therefore, a dynamic material leg position update mechanism is introduced in the point cloud preprocessing stage. By acquiring pallet and goods size information and upper frame angle information, the position of the pallet material legs is calculated and updated. Specifically:

[0160]

[0161] In the formula, l and w are the length and width of the tray. This refers to the angle at which the upper structure rotates.

[0162] S203. Project the obstacle point cloud data in the vehicle coordinate system to the Frenet coordinate system to obtain the longitudinal and lateral coordinates of each obstacle point cloud in the Frenet coordinate system.

[0163] S204. Based on the longitudinal and lateral coordinates, select the obstacle point clouds located within the safe driving area of ​​interest as candidate obstacle point clouds.

[0164] In one possible implementation, the current operating mode of the AVG is first obtained; then, when the AVG is in path tracking mode, a planned path curve is generated based on the path planning information, and the planned path curve is laterally expanded in combination with the minimum lateral safety width to form a path envelope region, and point clouds located outside the path envelope region in the Frenet coordinate system are removed; then, when the AVG is in spin mode, a circular safety region centered on the vehicle body center is constructed based on the minimum safety radius, and point clouds located outside the circular safety region in the Frenet coordinate system are removed; finally, the remaining point clouds after removal are determined as candidate obstacle point clouds.

[0165] The operating modes include at least path tracking mode and spin mode.

[0166] It should be understood that, firstly, the controller can obtain the current operating mode of the AVG. If it is in path tracking mode, then the endpoint and starting point of the planned path curve are generated based on the path planning information, and combined with... Figure 1 In the embodiment, the minimum lateral safety width obtained is used to laterally expand the planned path curve, forming a path envelope region, and point clouds outside this path envelope region in the Frenet coordinate system are removed. If the AVG is in spin mode, then according to... Figure 1The minimum safe radius obtained in the embodiment is used to construct a circular safe area with the center of the vehicle body (corresponding to s=0, l=0 in the Frenet coordinate system) as the center. Point clouds located outside the circular safe area in the Frenet coordinate system are removed, and the remaining point clouds after removal are determined as candidate obstacle point clouds.

[0167] Understandably, the motion characteristics and hazardous areas of AGVs differ under different operating modes. In path-tracking mode, the AGV travels along a path, and the main risk comes from the area near the path; therefore, the path envelope region is used for screening. In spin mode, the AGV rotates around its own center, and the main risk comes from the surrounding nearby area; therefore, a circular safety area is used for screening. Through mode-adaptive screening, the number of candidate obstacle point clouds is further reduced, and the focus is more on the key areas of the current operating mode, improving the efficiency and accuracy of subsequent processing.

[0168] In one possible implementation, after extracting the point cloud of candidate obstacles located within the region of interest for safe driving, the method further includes:

[0169] When the AVG is in charging mode, the point cloud of the corresponding area is removed based on the location of the charging station.

[0170] When the AVG is a backpack structure and is in a loaded state, the corresponding point cloud is removed according to the position of the cargo support structure.

[0171] When the AVG is in the workstation entry / exit state, the corresponding point cloud is removed according to the geometric range of the workstation area.

[0172] When the AVG is a forklift structure, the corresponding point cloud rejection strategy is executed according to the different operation stages of picking or placing goods.

[0173] Specifically: when the AGV is in the picking-up-to-workstation stage, the point cloud of the area in front of the forks or the workstation area is eliminated according to the radar installation position; when the AGV is in the unloading-to-workstation stage, the point cloud of the workstation area below the preset height is eliminated according to the installation height of the pallet detection radar; when the AGV is in the picking-up or unloading-to-workstation stage, the point cloud of the area below the forks is eliminated according to the spatial position of the forks.

[0174] It should be understood that after extracting the point cloud of candidate obstacles within the region of interest for safe driving, the point cloud of specific areas is further eliminated based on the current specific operating conditions of the AGV:

[0175] When the AGV is in automatic charging mode, an envelope of the charging pile area is constructed based on the known charging pile locations, and point clouds located within the envelope are removed (to avoid misjudging the charging pile as an obstacle).

[0176] If the AGV is a backpack structure (such as a lurking AGV) and is in a loaded state, construct the support structure envelope according to the position and size of the cargo support structure (such as the shelf leg) and remove the corresponding point cloud;

[0177] When the AGV is in the process of entering or leaving a workstation (such as under a shelf or inside a device), the workstation envelope is constructed based on the geometric range of the workstation area (such as the space under the shelf), and the point cloud within that range is removed (to avoid misjudging the workstation structure as an obstacle). If the AGV is a fork structure and is in the process of picking up or putting down goods (such as the forks extending into a pallet), the envelope of the fork tip area or the area below the forks is constructed based on the width, height and current spatial position of the forks, and the corresponding point cloud is removed (to avoid misjudging the pallet that the fork tip is about to contact as an obstacle, or to avoid interference from ground reflection point clouds).

[0178] Understandably, under certain operating conditions, some point clouds, while representing real obstacles, may be expected or permissible objects (such as charging piles, shelf legs, and pallets), and should not trigger safety deceleration or stopping. By dynamically removing point clouds from these areas based on operating conditions, unnecessary false alarms can be avoided, improving the operational efficiency and smoothness of the AGV.

[0179] Figure 3 A schematic diagram of the equivalent envelope model of the automated guided vehicle provided in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the safety area screening of an automated guided vehicle (AGV) provided in an embodiment of this application; as shown below. Figure 4 As shown in the diagram, this schematic illustrates how to perform region filtering on environmental point clouds under different control operation modes.

[0180] In one implementation, when the Automated Guided Vehicle (AGV) is in path-following operation mode, such as Figure 4 As shown on the left, the AGV moves along the planned path. The system determines the start and end points of the path based on the planned path output by the path planning module, and constructs a path envelope region centered on this path. When constructing the path envelope region, the minimum lateral safety width on both sides of the trajectory is considered, and the path envelope region is then expanded by a preset distance to form a safe passage area for obstacle detection. Subsequently, the environmental point cloud data is filtered, removing point cloud data located outside the path envelope region, and retaining only the point cloud data located within the safe passage area as candidate obstacle point clouds.

[0181] In another embodiment, when the automated guided vehicle is in spin operation mode, such as Figure 4As shown on the right, the AGV rotates in place around its own center. A spin safety zone is then constructed with the AGV's center as the center, and the minimum safe spin radius is determined based on the vehicle's structural dimensions and operational safety requirements. This safe radius is then expanded by a preset distance to form a spin safety detection zone. Subsequently, the environmental point cloud data is filtered, removing point cloud data that are farther from the AGV's center than the radius of the spin safety detection zone, retaining only the point cloud data located within the safety detection zone as candidate obstacle point clouds.

[0182] By using the above methods, safety detection zones are dynamically constructed according to different operating modes of the automated guided vehicle (AGV), and environmental point cloud data is filtered by region. This effectively reduces irrelevant point cloud data, improves obstacle detection efficiency, reduces computational complexity, and enhances the real-time performance and reliability of AGV safety decisions.

[0183] Figure 5 This is a schematic diagram of point cloud data filtering for an automated guided vehicle in charging pile mode, as provided in an embodiment of this application; Figure 5 As shown, to avoid misjudging charging piles as obstacles to be avoided during environmental perception and obstacle analysis, this application introduces a point cloud removal method based on the charging port location in the point cloud data processing stage. Specifically, firstly, the fixed location area of ​​the charging port on the vehicle body is determined based on the AGV's vehicle model information. Then, the charging pile point cloud portion corresponding to this charging port area is identified in the environmental point cloud data and removed from the candidate obstacle point cloud, such as... Figure 5 The left side of the image shows the area corresponding to the charging pile that has been marked and removed, while the right side shows the point cloud removal results when the AGV approaches the charging pile. This method ensures that the AGV will not misjudge the charging pile as an obstacle to be avoided when it is approaching it for charging. This guarantees that the AGV's path planning and safety decisions are both accurate and efficient, while avoiding unnecessary deceleration or stopping, thus improving the safety and reliability of the charging operation.

[0184] Figure 6 This is a schematic diagram of point cloud data filtering for an automated guided vehicle (AGV) in its operating mode, as provided in an embodiment of this application. Figure 6As shown in Figure a, when a forklift AGV performs a picking or placing operation, the forks need to extend under the pallet or near the edge of the workstation. At this point, the distance between the forks and the pallet is extremely close, and the fork-tip radar typically detects a dense point cloud. To avoid misjudging the pallet as an obstacle and causing unnecessary emergency stops, the point cloud scanned by the fork-tip radar needs to be filtered. Specifically, point clouds within a certain range of the fork tips are removed, retaining only fork-tip related point clouds used for environmental perception. Figure b shows that when the AGV is in an unloaded state and entering a designated workstation, it can be considered that the AGV is performing a picking operation. At this time, the entire workstation area (including shelves, pallet support structures, etc.) is considered an expected contact object and should not trigger a safety alarm. Therefore, point clouds not collected by fork-tip sensors within the workstation area are removed to eliminate the impact of static facilities or other interference on navigation or path planning. Figure c shows that when the AGV is in a loaded state and entering a workstation, it can be considered that the AGV is performing a placing operation. At this point, the pallet is positioned on the forks, and there may be ground point clouds or workstation edge point clouds in the area below the forks. To avoid misjudging the ground or workstation structure as obstacles, the point cloud in the workstation area below the height of the servo radar used for pallet detection should be removed. That is, only the point cloud above the servo radar height should be retained for obstacle detection, thus allowing the forks to smoothly carry the pallet into the workstation. Figure d shows that when the AGV completes the picking or placing operation and begins to leave the workstation, whether empty or loaded, the point cloud in the workstation area below the forks must be removed. This is because there are no work targets in the area below the forks at this time, and there may be ground reflections or residual point clouds at the workstation edge. Removing these points can prevent accidental emergency stops and ensure the AGV smoothly leaves the workstation.

[0185] Figure 7 This is a schematic diagram of an automated guided vehicle (AGV) provided in an embodiment of this application within different speed-limited zones; as shown. Figure 7 As shown, starting from the current position of the vehicle body, the area is divided into different regions along the planned path (positive direction of the longitudinal coordinate s-axis). Specifically:

[0186] The emergency stop zone is defined as an area extending longitudinally less than S1 meters and laterally less than L1 meters from the foremost point of the AGV's longitudinal profile. When an object is detected within the emergency stop zone, the AGV immediately issues a speed command to zero, triggering an emergency stop and halting the AGV's operation. S1 is the safety distance set according to the safety requirements of the AGV's operating environment, and L1 is calculated as follows:

[0187] L1=l min +0.1 meters

[0188] Among them, l min This is the minimum safe width on both sides.

[0189] Starting from the foremost edge of the AGV's body, the longitudinal distance extending within this area is less than S2, and the lateral distance is still less than L1 meters. This area is defined as the rapid stopping zone. When an obstacle is located within the rapid stopping zone, the AGV will apply a significant deceleration to brake until the vehicle comes to a complete stop, thus achieving the rapid stopping function. Here, S2 = 2.2 × max(0.5, current speed) + S1 meters.

[0190] Starting from the foremost edge of the AGV's body, the longitudinal distance is less than S3, and the lateral distance is still less than L1 meters; this area is defined as the normal parking zone. When an obstacle is located within the normal parking zone, the AGV will decelerate and brake at a normal rate until the vehicle comes to a stop, achieving a smooth stop. Here, S3 = 4 × max(0.5, current speed) + S1 meters.

[0191] Areas with a longitudinal distance greater than S3 meters or a lateral distance greater than L1 meters are defined as normal driving zones. When obstacles exist within the buffer zone, the AGV will maintain a low driving speed to reduce the potential threat of obstacles to driving safety and ensure smooth passage.

[0192] As can be seen, by dividing the area and corresponding control strategies, the AGV can achieve adaptive speed adjustment under different obstacle threat levels, taking into account both driving efficiency and safety, and improving the operational reliability and safety of the AGV in complex working scenarios.

[0193] Figure 8 This is a schematic diagram of the structure of the safety decision-making device for an automated guided vehicle provided in an embodiment of this application; as shown below. Figure 8 As shown, the device includes:

[0194] The acquisition module 801 is used to acquire multi-source operating information of the AVG. The multi-source operating information includes at least vehicle type information, path planning information, cargo status and obstacle point cloud data.

[0195] The first processing module 802 is used to construct an equivalent envelope model of the vehicle body based on vehicle type information and cargo status, and to calculate the parameters of the region of interest for safe driving in the Frenet coordinate system in combination with path planning information.

[0196] The second processing module 803 is used to preprocess the obstacle point cloud data, project it onto the Frenet coordinate system, extract the candidate obstacle point cloud located in the safe driving region of interest, and obtain the candidate obstacle point cloud.

[0197] The calculation module 804 is used to obtain the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction based on the candidate obstacle point cloud, and to calculate the corresponding speed limit value based on the position of the nearest obstacle relative to the vehicle body.

[0198] The determination module 805 is used to determine the safe driving speed of the AVG based on the speed limit value.

[0199] In one possible implementation, the first processing module 802 is specifically used for:

[0200] Based on the vehicle model information, a physical model of the vehicle body is created using one or more geometric envelopes;

[0201] When the AVG is in a loaded state, the cargo size information is obtained, and when the cargo size is larger than the vehicle body size, the geometric envelope is updated with the cargo size to form the vehicle body equivalent envelope model in the loaded state.

[0202] The planned path is downsampled based on the path planning information, and a Frenet coordinate system is created based on the downsampled planned path.

[0203] The minimum longitudinal safety distance, minimum lateral safety width, minimum safety height, and minimum safety radius in spin mode are calculated in the Frenet coordinate system to form the region of interest parameters for safe driving.

[0204] In one possible implementation, the second processing module 803 is specifically used for:

[0205] Motion compensation is performed on the obstacle point cloud data based on the timestamp of the obstacle point cloud data and the current vehicle speed and angular velocity of the AVG to obtain the compensated obstacle point cloud data.

[0206] The compensated obstacle point cloud data is filtered and then transformed into the vehicle coordinate system.

[0207] The obstacle point cloud data in the vehicle coordinate system is projected to the Frenet coordinate system to obtain the longitudinal and lateral coordinates of each obstacle point cloud in the Frenet coordinate system.

[0208] Based on the longitudinal and lateral coordinates, obstacle point clouds located within the safe driving area of ​​interest are selected as candidate obstacle point clouds.

[0209] In one possible implementation, the second processing module 803 is specifically used for:

[0210] Obtain the current running mode of the AVG, which includes at least path tracking mode and spin mode;

[0211] When the AVG is in path tracking mode, it generates a planned path curve based on the path planning information, and expands the planned path curve laterally in combination with the minimum lateral safety width to form a path envelope region, and removes the point cloud outside the path envelope region in the Frenet coordinate system.

[0212] When the AVG is in spin mode, a circular safe area with the center of the vehicle body as the center is constructed according to the minimum safe radius, and the point cloud outside the circular safe area in the Frenet coordinate system is removed.

[0213] The remaining point cloud after elimination is identified as the candidate obstacle point cloud.

[0214] In one possible implementation, the apparatus further includes a screening module, which, after extracting the point cloud of candidate obstacles located within the region of interest for safe driving, is used to:

[0215] When the AVG is in charging mode, the point cloud of the corresponding area is removed based on the location of the charging station.

[0216] When the AVG is a backpack structure and is in a loaded state, the corresponding point cloud is removed according to the position of the cargo support structure.

[0217] When the AVG is in the workstation entry / exit state, the corresponding point cloud is removed according to the geometric range of the workstation area.

[0218] When the AVG is a forklift structure, the corresponding point cloud rejection strategy is executed according to the different operation stages of picking or placing goods.

[0219] In one possible implementation, the filtering module is specifically used for:

[0220] When the AGV is in the picking and entering workstation stage, the point cloud in the area in front of the forks or the workstation area is removed according to the radar installation position.

[0221] When the AGV is in the stage of placing goods into the workstation, the point cloud of the workstation area that is below the preset height range is removed according to the installation height of the pallet detection radar.

[0222] When the AGV is in the stage of picking up or placing goods out of the workstation, the point cloud of the area below the forks is removed based on the spatial position of the forks.

[0223] In one possible implementation, the computing module 804 is specifically used for:

[0224] In the Frenet coordinate system, the vehicle body equivalent envelope model is placed point by point along the path planning information from near to far, and it is detected whether the vehicle body equivalent envelope model has geometric collision with the candidate obstacle point cloud.

[0225] When a geometric collision is detected for the first time, the obstacle at the collision location is identified as the nearest obstacle in the longitudinal direction, and the longitudinal distance of the nearest obstacle in the longitudinal direction is obtained.

[0226] For candidate obstacle point clouds with a lateral distance greater than the minimum lateral safety width, the point with the smallest lateral distance is determined as the lateral nearest obstacle based on the pre-built comprehensive distance function, and the lateral distance of the lateral nearest obstacle is obtained.

[0227] In the Frenet coordinate system, the corresponding longitudinal speed limit and lateral speed limit are calculated based on the longitudinal distance to the nearest longitudinal obstacle and the lateral distance to the nearest lateral obstacle, respectively.

[0228] The speed limit with the smaller value between the longitudinal and lateral speed limits is determined as the speed limit corresponding to the candidate obstacle point cloud.

[0229] In one possible implementation, the determining module 805 is specifically used for:

[0230] In the Frenet coordinate system, the area in front of the vehicle is divided into emergency stopping zone, fast stopping zone, normal stopping zone and normal driving zone according to the longitudinal distance;

[0231] When the nearest longitudinal obstacle is located within the emergency stop zone, a zero-speed command is generated.

[0232] When the nearest longitudinal obstacle is located in the fast parking zone or normal parking zone, the target deceleration value is calculated based on the current vehicle speed, system response time and preset deceleration, and a deceleration control command is generated.

[0233] When the obstacle is located within the normal driving zone, a speed limit control command is generated to maintain or adjust the vehicle speed.

[0234] The speed limit is compared with the current speed of the AVG, and corresponding control commands are generated based on the comparison results to determine the safe driving speed of the AVG.

[0235] The safety decision-making device for automated guided vehicles provided in this application embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0236] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 9 As shown, the electronic device 90 provided in this embodiment includes at least one processor 901 and a memory 902. Optionally, the device 90 further includes a communication component 903. The processor 901, memory 902, and communication component 903 are connected via a bus 904.

[0237] In a specific implementation, at least one processor 901 executes computer execution instructions stored in memory 902, causing at least one processor 901 to perform the above-described method.

[0238] The specific implementation process of processor 901 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0239] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0240] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0241] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0242] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0243] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0244] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0245] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0246] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0247] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0248] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0249] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0250] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A safety decision-making method for automated guided vehicles, characterized in that, include: Acquire multi-source operational information of the automated guided vehicle (AVG), wherein the multi-source operational information includes at least vehicle type information, path planning information, cargo status, and obstacle point cloud data; Based on the vehicle model information and cargo status, an equivalent envelope model of the vehicle body is constructed, and combined with path planning information, the parameters of the region of interest for safe driving are calculated in the Frenet coordinate system. The obstacle point cloud data is preprocessed and projected onto the Frenet coordinate system to extract the obstacle point cloud located within the safe driving region of interest, thus obtaining candidate obstacle point clouds. Based on the candidate obstacle point cloud, the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction are obtained respectively, and the corresponding speed limit value is calculated based on the position of the nearest obstacle relative to the vehicle body. The safe driving speed of the AVG is determined based on the speed limit value.

2. The method according to claim 1, characterized in that, The step of constructing an equivalent envelope model of the vehicle body based on the vehicle model information and cargo status, and calculating the parameters of the region of interest for safe driving in the Frenet coordinate system in conjunction with path planning information, includes: Based on the vehicle model information, a physical model of the vehicle body is created using one or more geometric envelopes; When the AVG is in a loaded state, the cargo size information is obtained, and when the cargo size is larger than the vehicle body size, the geometric envelope is updated with the cargo size to form an equivalent envelope model of the vehicle body in a loaded state. The planned path is downsampled based on the path planning information, and a Frenet coordinate system is created based on the downsampled planned path. The minimum longitudinal safety distance, minimum lateral safety width, minimum safety height, and minimum safety radius in spin mode are calculated in the Frenet coordinate system to form the region of interest parameters for safe driving.

3. The method according to claim 1 or 2, characterized in that, The step of preprocessing the obstacle point cloud data and projecting it onto the Frenet coordinate system to extract candidate obstacle point clouds located within the safe driving region of interest includes: Motion compensation is performed on the obstacle point cloud data based on the timestamp of the obstacle point cloud data and the current vehicle speed and angular velocity of the AVG to obtain the compensated obstacle point cloud data. The compensated obstacle point cloud data is filtered, and the filtered obstacle point cloud data is converted to the vehicle coordinate system. The obstacle point cloud data in the vehicle coordinate system is projected onto the Frenet coordinate system to obtain the longitudinal and lateral coordinates of each obstacle point cloud in the Frenet coordinate system. Based on the longitudinal and lateral coordinates, obstacle point clouds located within the safe driving area of ​​interest are selected as candidate obstacle point clouds.

4. The method according to claim 2, characterized in that, The step of filtering obstacle point clouds located within the safe driving area of ​​interest based on the longitudinal and lateral coordinates as candidate obstacle point clouds includes: Obtain the current operating mode of the AVG, wherein the operating mode includes at least path tracking mode and spin mode; When the AVG is in path tracking mode, a planned path curve is generated based on the path planning information, and the planned path curve is laterally expanded in combination with the minimum lateral safety width to form a path envelope region, and point clouds located outside the path envelope region in the Frenet coordinate system are removed. When the AVG is in spin mode, a circular safe area with the center of the vehicle body as the center is constructed according to the minimum safe radius, and the point cloud outside the circular safe area in the Frenet coordinate system is removed. The remaining point cloud after elimination is determined as the candidate obstacle point cloud.

5. The method according to claim 3, characterized in that, After extracting the point cloud of candidate obstacles located within the region of interest for safe driving, the method further includes: When the AVG is in charging mode, the point cloud of the corresponding area is removed according to the location of the charging pile. When the AVG is a backpack structure and is in a loaded state, the corresponding point cloud is removed according to the position of the cargo support structure. When the AVG is in the workstation entry / exit state, the corresponding point cloud is removed according to the geometric range of the workstation area; When the AVG has a fork structure, the corresponding point cloud rejection strategy is executed according to the different operation stages of picking or placing goods.

6. The method according to claim 5, characterized in that, Execute corresponding point cloud removal strategies based on different stages of the picking or placing operation, including: When the AGV is in the picking and entering workstation stage, the point cloud in the area in front of the forks or the workstation area is removed according to the radar installation position. When the AGV is in the stage of placing goods into the workstation, the point cloud of the workstation area that is below the preset height range is removed according to the installation height of the pallet detection radar. When the AGV is in the stage of picking up or placing goods out of the workstation, the point cloud of the area below the forks is removed based on the spatial position of the forks.

7. The method according to claim 1 or 2, characterized in that, The step of determining the nearest longitudinal obstacle and the nearest lateral obstacle based on the candidate obstacle point cloud, and calculating the corresponding speed limit value based on the position of each obstacle relative to the vehicle body, includes: In the Frenet coordinate system, the vehicle body equivalent envelope model is placed point by point along the path planning information from near to far, and it is detected whether the vehicle body equivalent envelope model has a geometric collision with the candidate obstacle point cloud. When a geometric collision is detected for the first time, the obstacle at the corresponding collision location is identified as the nearest obstacle in the longitudinal direction, and the longitudinal distance of the nearest obstacle in the longitudinal direction is obtained. For candidate obstacle point clouds with a lateral distance greater than the minimum lateral safety width, the point with the smallest comprehensive distance is determined as the lateral nearest obstacle according to the pre-constructed comprehensive distance function, and the lateral distance of the lateral nearest obstacle is obtained. In the Frenet coordinate system, the corresponding longitudinal speed limit and lateral speed limit are calculated based on the longitudinal distance to the nearest longitudinal obstacle and the lateral distance to the nearest lateral obstacle, respectively. The speed limit value with the smaller value between the longitudinal speed limit value and the lateral speed limit value is determined as the speed limit value corresponding to the candidate obstacle point cloud.

8. The method according to claim 7, characterized in that, Determining the safe driving speed of the AVG based on the speed limit value includes: In the Frenet coordinate system, the area in front of the vehicle is divided into an emergency stop zone, a quick stop zone, a normal stop zone, and a normal driving zone according to the longitudinal distance. When the nearest longitudinal obstacle is located within the emergency stop zone, a zero-speed command is generated. When the nearest longitudinal obstacle is located in the fast parking zone or normal parking zone, the target deceleration value is calculated based on the current vehicle speed, system response time and preset deceleration, and a deceleration control command is generated. When the obstacle is located within the normal driving zone, a speed limit control command is generated to maintain or adjust the vehicle speed. The speed limit value is compared with the current driving speed of the AVG, and a corresponding control command is generated based on the comparison result to determine the safe driving speed of the AVG.

9. A safety decision-making device for an automated guided vehicle, characterized in that, include: The acquisition module is used to acquire multi-source operating information of the automated guided vehicle (AVG), which includes at least vehicle model information, path planning information, pose information, cargo status, and environmental point cloud data. The first processing module is used to construct an equivalent envelope model of the vehicle body based on the vehicle model information and cargo status, and to calculate the parameters of the region of interest for safe driving in the Frenet coordinate system by combining path planning information and pose information. The second processing module is used to preprocess the environmental point cloud data and project it onto the Frenet coordinate system to extract candidate obstacle point clouds located in the safe driving region of interest. The calculation module is used to obtain the nearest obstacle in the longitudinal direction and the nearest obstacle in the lateral direction based on the candidate obstacle point cloud, and to calculate the corresponding speed limit value based on the position of each obstacle relative to the vehicle body. The determining module is used to determine the safe driving speed of the AVG based on the speed limit value.

10. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

Images