A hierarchical control method and system for obstacle avoidance in AGV robots

Abstract

This application provides a hierarchical control method and system for obstacle avoidance of AGV robots, relating to the general field of control or regulation systems. The method includes: collecting AGV robot operating parameters as input to a risk assessment model and outputting risk level values; determining the first-triggered obstacle avoidance actuator based on the preset range of the risk level value, including a hardware control layer, a local computing layer, and a remote collaboration layer; when the local computing layer is triggered, dividing the obstacle avoidance area of ​​the AGV robot within a preset range into 16 sector spaces, determining the target sector space with the highest passage index as the avoidance direction, and selecting the corresponding target obstacle avoidance scheme based on the channel width parameter of the target sector space. Implementing this method can reduce obstacle avoidance response time, avoid the processing delay caused by the main control system's involvement in all obstacle avoidance decisions in traditional solutions, and reduce the collision risk caused by waiting for the main control system to process.

Description

Technical Field

[0001] This application relates to the general field of control or regulation systems, and more particularly to a hierarchical control method and system for obstacle avoidance of AGV robots. Background Technology

[0002] With the rapid development of industrial automation, AGV (Automated Guided Vehicle) robots, intelligent material handling robots, are increasingly widely used in factories, warehouses, and other scenarios. In complex industrial environments, AGV robots need to cope with various emergencies to ensure the safe and efficient completion of material handling tasks. Therefore, obstacle avoidance capability has become a key technology for ensuring the normal operation of AGV robots.

[0003] In related technologies, obstacle avoidance control for AGV robots is mainly achieved through centralized control. AGV robots collect environmental information using visual sensors (such as cameras) and LiDAR, transmitting this information to a central control system for analysis. Upon receiving an abnormal signal, the control system first verifies the authenticity of the anomaly, then generates obstacle avoidance commands based on preset scheduling rules, and finally issues the commands to the AGV robots to execute the corresponding avoidance actions. This obstacle avoidance method relies on the unified scheduling of the control system, facilitating the collaborative operation of multiple AGV robots.

[0004] However, in practical applications, the acquisition, transmission, confirmation, and command issuance of abnormal signals all require processing time. Furthermore, network transmission may introduce delays, often resulting in a significant time gap between the AGV robot detecting an obstacle and executing an avoidance maneuver. This processing delay is particularly problematic when the AGV robot is operating at high speed or encountering sudden obstacles; it can make it difficult for the AGV robot to react promptly, increasing the risk of collisions. Summary of the Invention

[0005] This application provides a hierarchical control method and system for obstacle avoidance of AGV robots, which addresses the problem of how to enable AGV robots to quickly perform avoidance actions after detecting obstacles, thereby reducing the risk of collisions caused by waiting for the main control system to process the data.

[0006] In a first aspect, this application provides a hierarchical control method for obstacle avoidance in AGV robots, applied to an AGV robot control system. The AGV robot control system includes three obstacle avoidance execution mechanisms: a hardware control layer, a local computing layer, and a remote collaboration layer. The method includes:

[0007] Collect AGV robot operating parameters, including load value, ground friction coefficient, obstacle distance value, historical accident data, and task time parameter;

[0008] Input the operating parameters into the risk assessment model and output the risk level value;

[0009] The obstacle avoidance mechanism that is triggered first is determined based on the preset range in which the risk level value is located. The preset range includes a first preset range corresponding to the hardware control layer, a second preset range corresponding to the local computing layer, and a third preset range corresponding to the remote collaboration layer. The hardware control layer is used to perform emergency braking on the AGV robot, and the remote collaboration layer is connected to the main control system.

[0010] When the local computing layer is triggered, the obstacle avoidance area of ​​the AGV robot within the preset range is divided into 16 sector spaces, and the passage index of each sector space is calculated according to the passage index calculation formula.

[0011] The target sector with the highest traffic index is determined as the avoidance direction, and the corresponding target obstacle avoidance scheme is selected according to the channel width parameter of the target sector.

[0012] Through the above embodiments, the AGV robot control system achieves hierarchical and rapid response of the AGV robot by establishing a three-level obstacle avoidance mechanism consisting of a hardware control layer, a local computing layer, and a remote collaboration layer. The system can automatically trigger the most appropriate obstacle avoidance level based on the risk level, without waiting for instructions from the main control system each time. When the risk is low, the local computing layer can independently complete obstacle avoidance decisions, quickly selecting the optimal obstacle avoidance path through 16-sector space analysis and passage index evaluation. This hierarchical control strategy reduces obstacle avoidance response time, avoids the processing delays caused by the main control system's involvement in all obstacle avoidance decisions in traditional solutions, and effectively reduces the collision risk caused by waiting for the main control system to process the requests.

[0013] In some embodiments, the step of inputting the operating parameters into the risk assessment model and outputting a risk level value specifically includes:

[0014] The hardware control layer performs a threshold judgment based on the load value, the ground friction coefficient, and the obstacle distance value to obtain a first risk index;

[0015] The local computing layer inputs the first risk index, the historical accident data, and the task time parameters into the fuzzy evaluator, and calculates the second risk index based on preset fuzzy rules.

[0016] The remote collaboration layer inputs the first risk index and the second risk index into a pre-trained deep learning model to obtain a third risk index;

[0017] The risk level value is determined based on the weighted average of the first risk index, the second risk index, and the third risk index.

[0018] Through the above embodiments, the hardware control layer can quickly determine whether emergency obstacle avoidance is needed based on basic parameters without waiting for confirmation from the main control system; the local computing layer performs rapid risk assessment through a fuzzy evaluator; and the remote collaboration layer provides more accurate risk assessment when conditions permit. This layered risk assessment mechanism ensures that the corresponding obstacle avoidance process can be initiated as quickly as possible under different risk levels, effectively overcoming the response delay problem caused by over-reliance on the main control system for risk assessment in traditional solutions.

[0019] In some embodiments, the step of dividing the obstacle avoidance area of ​​the AGV robot within a preset range into 16 sector-shaped spaces specifically includes:

[0020] The motion characteristic coefficients of the AGV robot are calculated based on its speed parameters, including current speed, turning angle, and acceleration.

[0021] Based on the motion characteristic coefficient, the obstacle avoidance area within the preset range is divided into a forward area, a lateral area, and a backward area, with the forward area, lateral area, and backward area corresponding to 8 sector spaces, 6 sector spaces, and 2 sector spaces, respectively.

[0022] The spatial weight coefficient is calculated based on the sector spatial distribution, and the spatial weight coefficient is used as the weighting parameter in the passage index calculation formula.

[0023] Through the above embodiments, the AGV robot control system divides the obstacle avoidance area into eight forward, six lateral, and two backward sector-shaped spaces, enabling the AGV robot to quickly assess the obstacle avoidance probability in each direction. This non-uniform spatial division scheme based on motion characteristics highlights the importance of the forward and lateral spaces, allowing the AGV robot to complete spatial analysis and select the optimal avoidance direction in the shortest possible time. This scheme eliminates the need for the main control system to perform complex path planning, significantly shortening the obstacle avoidance decision time and improving the AGV robot's rapid response capability when encountering sudden obstacles.

[0024] In some embodiments, the step of calculating the passage index of each sector space according to the passage index calculation formula specifically includes:

[0025] Real-time status parameters of each of the sector spaces are collected, and an instantaneous passage index is calculated based on the real-time status parameters, which include obstacle density, moving speed and relative distance.

[0026] Based on the changing trend of the real-time state parameters within the historical sampling period, the Kalman filter algorithm is used to predict the state evolution of each sector space within a preset time window to obtain the short-term traffic index.

[0027] A long-term evaluation model is established based on the historical passage records of each sector space, and the long-term passage index corresponding to each sector space is calculated.

[0028] The weighting coefficients of the instant passage index, the short-term passage index, and the long-term passage index are adjusted according to the complexity of the current scenario, and the passage index is obtained by weighted combination.

[0029] Through the above embodiments, the AGV robot control system calculates the instantaneous passage index in real time, allowing the AGV robot to respond immediately to changes in the surrounding environment; short-term predictions using the Kalman filter algorithm can predict obstacle movement trends; and long-term evaluations combined with historical data provide a more reliable basis for decision-making. This multi-dimensional local evaluation mechanism enables the AGV robot to independently complete most obstacle avoidance decisions, significantly reducing dependence on the main control system and avoiding safety hazards caused by communication delays.

[0030] In some embodiments, the step of selecting the corresponding target obstacle avoidance scheme based on the channel width parameter of the target sector space specifically includes:

[0031] The obstacle avoidance level is calculated based on the channel width parameter of the target fan-shaped space, and the initial obstacle avoidance scheme is determined based on the obstacle avoidance level. The obstacle avoidance levels are, in order from low to high, deceleration obstacle avoidance, lane change obstacle avoidance, and emergency obstacle avoidance.

[0032] After the initial obstacle avoidance scheme is started, if the obstacle avoidance effect parameters of the AGV robot are detected to be lower than the preset target threshold, the scheme upgrade judgment is triggered. The obstacle avoidance effect parameters include relative distance change rate, heading deviation and obstacle avoidance progress.

[0033] In the scheme upgrade judgment, when the obstacle avoidance margin index is detected to be greater than the preset upgrade threshold, the initial obstacle avoidance scheme is upgraded to the target obstacle avoidance scheme corresponding to a higher obstacle avoidance level. The obstacle avoidance margin index is determined based on the remaining decision time and maneuver space.

[0034] Through the above embodiments, the AGV robot control system directly selects the initial obstacle avoidance level based on the channel width and can automatically upgrade the plan based on the obstacle avoidance effect, without needing to consult the main control system each time. Introducing an obstacle avoidance margin index as an upgrade criterion ensures that the obstacle avoidance level can be immediately upgraded when the danger level increases. This adaptive local decision-making mechanism greatly improves the AGV robot's response speed to emergencies and effectively prevents missed obstacle avoidance opportunities due to waiting for the main control system's decision.

[0035] In some embodiments, after the step of determining the first obstacle avoidance actuator based on the preset range in which the risk level value falls, the method further includes:

[0036] When the hardware control layer is triggered, emergency braking is immediately executed;

[0037] After the emergency braking, the vehicle resumes driving according to the control commands issued by the main control system.

[0038] Through the above embodiments, when an emergency is detected, the AGV robot can immediately perform emergency braking without waiting for confirmation from the main control system, thus taking avoidance actions in the shortest possible time. This mechanism directly eliminates the response delay caused by requesting authorization from the main control system in traditional solutions, providing a final safety guarantee for the AGV robot. Simultaneously, by requiring authorization from the main control system to resume operation, safe recovery after emergency obstacle avoidance is ensured.

[0039] In some embodiments, after the step of determining the first obstacle avoidance actuator based on the preset range in which the risk level value falls, the method further includes:

[0040] When the remote collaboration layer is triggered, the AGV robot is controlled to wait for the control command issued by the main control system during normal operation.

[0041] The AGV robot is controlled to avoid obstacles according to the control instructions.

[0042] Through the above embodiments, the AGV robot control system has designed a special obstacle avoidance mechanism for the remote collaboration layer, enabling the AGV robot to maintain normal operation while waiting for instructions from the main control system. This mechanism avoids the chain reaction that may be caused by simply stopping and waiting, minimizing the impact of waiting for the main control system while ensuring safety. At the same time, by reserving a remote collaboration interface, the system can still make more optimized obstacle avoidance decisions from the global perspective of the main control system when necessary, achieving an organic combination of rapid response and global optimization.

[0043] Secondly, this application provides an AGV robot control system, which includes one or more processors and a memory;

[0044] The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the AGV robot control system can implement the hierarchical control AGV robot obstacle avoidance method provided in the above embodiments, which will not be described in detail here.

[0045] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on an AGV robot control system, enable the AGV robot control system to implement a hierarchical control AGV robot obstacle avoidance method provided in the above embodiments, which will not be elaborated here.

[0046] Fourthly, this application provides a computer program product that, when running on an AGV robot control system, enables the AGV robot control system to implement a hierarchical control AGV robot obstacle avoidance method provided in the above embodiments, which will not be elaborated here.

[0047] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0048] 1. By establishing a hierarchical control architecture comprising a hardware control layer, a local computing layer, and a remote collaboration layer, the AGV robot control system can automatically trigger the most suitable obstacle avoidance level based on the risk level. The hardware control layer can directly execute emergency braking, the local computing layer can independently complete general obstacle avoidance decisions, and the remote collaboration layer handles situations requiring global coordination. This hierarchical mechanism allows the AGV robot to autonomously select the fastest response path based on the risk level without waiting for instructions from the main control system, fundamentally solving the processing delay problem in traditional centralized control schemes.

[0049] 2. The area surrounding the AGV robot is divided into 16 sector-shaped spaces, with a non-uniform distribution of 8 forward, 6 lateral, and 2 backward sectors based on its motion characteristics. Combining a multi-dimensional accessibility index calculation method that integrates real-time status, short-term prediction, and long-term evaluation, the AGV robot can quickly complete its surrounding environment analysis and optimal obstacle avoidance path selection locally. This spatial division and evaluation system fully considers the AGV robot's motion characteristics, enabling it to independently make effective obstacle avoidance decisions in the shortest possible time.

[0050] 3. A hierarchical obstacle avoidance system has been established, encompassing deceleration obstacle avoidance, lane-changing obstacle avoidance, and emergency obstacle avoidance. The AGV robot control system can quickly select the initial obstacle avoidance level based on the channel width and automatically upgrade the scheme based on obstacle avoidance effectiveness and obstacle avoidance margin index. This adaptive mechanism allows the AGV robot to flexibly adjust its obstacle avoidance strategy according to actual conditions, ensuring both rapid response and effective obstacle avoidance. It effectively addresses the problem of untimely obstacle avoidance caused by the simplistic approach or complex decision-making process in traditional solutions. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating a hierarchical control method for obstacle avoidance in an AGV robot according to an embodiment of this application.

[0052] Figure 2 This is another flowchart illustrating a hierarchical control method for obstacle avoidance in AGV robots, as described in this application.

[0053] Figure 3 This is a schematic diagram of the physical device structure of an AGV robot control system in an embodiment of this application. Detailed Implementation

[0054] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0055] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0056] For ease of understanding, the method provided in this implementation is described in process below. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a hierarchical control method for obstacle avoidance in an AGV robot according to an embodiment of this application.

[0057] S101. Collect AGV robot operating parameters.

[0058] The AGV robot control system collects operational parameters in real time through a multi-sensor network, including load values, ground friction coefficient, obstacle distance values, historical accident data, and task time parameters. Load values ​​are collected by weight sensors on the AGV robot platform, allowing real-time monitoring of the current load status. The ground friction coefficient is measured by specialized ground detection sensors, adaptable to different road conditions. Obstacle distance values ​​are obtained by collecting multi-source data from devices such as LiDAR, ultrasonic sensors, and vision sensors mounted on the AGV robot, constructing a comprehensive obstacle detection network. Historical accident data is stored in a local database, containing information such as accident type, cause, and impact. Task time parameters include key indicators such as task urgency and remaining execution time; task urgency is inversely correlated with remaining execution time.

[0059] S102. Input the operating parameters into the risk assessment model and output the risk level value.

[0060] Specifically, the hardware control layer makes a threshold judgment based on the load value, ground friction coefficient and obstacle distance value to obtain the first risk index; the local computing layer inputs the first risk index, historical accident data and task time parameters into the fuzzy evaluator and calculates the second risk index based on the preset fuzzy rules; the remote collaboration layer inputs the first risk index and the second risk index into the pre-trained deep learning model to obtain the third risk index; finally, the risk level value is determined according to the weighted average of the first risk index, the second risk index and the third risk index.

[0061] Optionally, the hardware control layer uses a threshold-based algorithm to calculate the first risk index. This algorithm sets three safety thresholds: a load safety threshold (e.g., 80% of the maximum load capacity), a friction coefficient safety threshold (e.g., 0.4), and an obstacle distance safety threshold (e.g., 1.5 times the emergency braking distance). When any parameter exceeds a threshold, a hardware-level risk assessment is triggered. The first risk index... The calculation formula is:

[0062] Where W is the load percentage, μ is the friction coefficient, D is the distance to the obstacle, α1=0.3, α2=0.4, α3=0.3.

[0063] The local computation layer constructs a fuzzy evaluator, which includes three input variables ( (historical accident rate, mission urgency) and one output variable (second risk index) Fuzzy rules employ Mamdani-type reasoning, for example: if If the accident rate is high and the historical accident rate is high, and the mission urgency is high, then... The membership function of the input variables is distributed in a triangular pattern, and the centroid method is used to resolve fuzziness in the output variables.

[0064] The remote collaboration layer deploys a pre-trained LSTM neural network model, with the input being... and The output is the third risk index. This LSTM neural network model was trained on historical data and contains three hidden layers with 128 neurons each. The activation function is ReLU, and the loss function is mean squared error. The final risk level is calculated using a weighted average: R = 0.5. +0.3 + 0.2 .

[0065] Taking a specific industrial scenario as an example, hardware-layer computing... =0.7 (75% load, friction coefficient 0.5, obstacle distance 1.2 meters), the local layer is calculated based on the historical accident rate (0.5 times / hour) and mission urgency (high). =0.8, Long-range layer model prediction =0.85. The final risk level value R = 0.7*0.5 + 0.8*0.3 + 0.85*0.2 = 0.75, which belongs to the medium risk level.

[0066] It should be noted that the fuzzy logic controller contains 5... The system generates 5×3×3=45 fuzzy rules from 3 subsets of historical accident rates (low, medium, and high) and 3 subsets of task urgency (low, medium, and high).

[0067] S103. Determine the first obstacle avoidance mechanism to be triggered based on the preset range of the risk level value.

[0068] Optionally, the preset interval is divided into three levels: the first interval (R≥0.9) triggers the hardware control layer, the second interval (0.7≤R<0.9) triggers the local computing layer, and the third interval (R<0.7) triggers the remote collaboration layer. Optionally, the hardware control layer directly controls the AGV robot's motor braking system via PWM signals, with a response time of less than 50ms. The local computing layer activates the obstacle avoidance decision module, which includes sub-modules such as sector space division and passage index calculation. The remote collaboration layer connects to the main control system via a 5G communication module and uses the MQTT protocol for data exchange.

[0069] When the risk level value R=0.85, the local calculation layer is triggered. The AGV robot control system immediately stops transmitting data to the remote collaboration layer and initiates the local obstacle avoidance decision-making process.

[0070] S104. When the local computing layer is triggered, the obstacle avoidance area of ​​the AGV robot within the preset range is divided into 16 sector spaces, and the passage index of each sector space is calculated according to the passage index calculation formula.

[0071] Specifically, when the local computing layer is triggered, the AGV robot control system calculates the motion characteristic coefficient of the AGV robot based on speed parameters such as the current speed, turning angle, and acceleration of the AGV robot; then, based on the motion characteristic coefficient, the obstacle avoidance area within the preset range is divided into a forward area (including 8 sector spaces), a lateral area (including 6 sector spaces), and a backward area (including 2 sector spaces); finally, the spatial weight coefficient is calculated based on the sector space distribution, and the spatial weight coefficient is used as a weighting parameter in the passage index calculation formula.

[0072] The motion characteristic coefficient K is calculated using the following formula:

[0073] Where v is the current velocity, g is the gravitational acceleration, μ is the friction coefficient, and L is the wheelbase. The forward region is divided into 8 sectors, each 45°; the lateral region is divided into 6 sectors, each 60°; and the backward region is divided into 2 sectors, each 45°. Spatial weighting coefficients are set according to the importance of the regions: forward region sector weight 0.4, lateral region sector weight 0.3, and backward region sector weight 0.3.

[0074] Next, the AGV robot control system collects real-time status parameters such as obstacle density, moving speed, and relative distance for each sector space, and calculates the immediate passage index I based on the real-time status parameters. Then, based on the changing trend of the real-time status parameters within the historical sampling period, the Kalman filter algorithm is used to predict the state evolution of each sector space within a preset time window to obtain the short-term passage index S. Then, a long-term evaluation model is established based on the historical passage records of each sector space to calculate the long-term passage index L corresponding to each sector space. Finally, the weight coefficients corresponding to the immediate passage index, short-term passage index, and long-term passage index are adjusted according to the current scene complexity, and the passage index P is obtained by weighted combination.

[0075] Specifically, the traffic index calculation employs a multi-dimensional evaluation model. The formula for calculating the real-time traffic index I is as follows:

[0076] Where ρ is the obstacle density, Let d be the relative velocity and d be the relative distance.

[0077] The final traffic index is calculated using the following formula:

[0078] ,in, , and These represent the weight coefficients of I, S, and L, respectively, and the weight coefficients can be dynamically adjusted according to the complexity of the scenario, without any limitation here.

[0079] Taking a logistics center as an example, the AGV robot's current speed is 2 m / s, wheelbase is 0.8 m, friction coefficient is 0.6, and K=0.85 is calculated. The obstacle avoidance area is divided into 8 forward sectors (0-360°, one every 45°), 6 lateral sectors (3 on the left or right, one every 60°), and 2 backward sectors. The obstacle density in the third forward sector is detected in real time to be 0.3, with a relative speed of 0.5 m / s and a relative distance of 3 m, resulting in I=0.89. Kalman filtering predicts I=0.85 after 0.5 seconds. Historical data shows the average passage index for this area is 0.82. The current scene complexity is medium, and the weighting coefficients are... =0.5, =0.3, =0.2, and finally P=0.86.

[0080] S105. Determine the target sector space with the highest traffic index as the avoidance direction, and select the corresponding target obstacle avoidance scheme according to the channel width parameter of the target sector space.

[0081] After calculating the passage index for 16 sector-shaped spaces, the AGV robot control system first sorts the passage indexes of each sector in descending order and selects the sector with the highest passage index as the target direction. The system fits the obstacle boundary using LiDAR point cloud data to calculate the channel width of the target sector. The obstacle avoidance level classification standard is as follows: when the channel width is ≥ 2 times the vehicle width, deceleration obstacle avoidance is used; when it is 1.5-2 times the vehicle width, lane changing obstacle avoidance is used; and when it is < 1.5 times the vehicle width, emergency obstacle avoidance is triggered.

[0082] After the initial obstacle avoidance plan is determined, the system monitors obstacle avoidance performance parameters (relative distance change rate, heading deviation, and obstacle avoidance progress) in real time. When the obstacle avoidance performance is detected to be below a threshold, the plan upgrade mechanism is activated. The obstacle avoidance margin index is calculated using the remaining decision time and maneuver space. The obstacle avoidance margin index is calculated as follows:

[0083] ,in, where v is the remaining decision time, v is the current speed, and w is the channel width. This is the minimum safe width.

[0084] S106. When the hardware control layer is triggered, emergency braking is immediately executed, and after emergency braking, driving resumes according to the control instructions issued by the main control system.

[0085] When the risk level exceeds 0.9, triggering the hardware control layer, the system directly controls the motor braking system via PWM signals. The braking system employs a dual-loop redundant design, including both electromagnetic and mechanical brakes. During braking, the system monitors wheel speed signals in real time and uses an anti-lock braking system (ABS) to prevent wheel lock-up.

[0086] After emergency braking, the AGV robot enters a safe stop state, cutting off power output and maintaining braking. The system sends an emergency braking report to the main control system via the 5G communication module, including information such as location, time, and reason for braking. After assessing the situation, the main control system sends a recovery command. Upon receiving the command, the AGV robot releases the brakes and replans its path.

[0087] It should be noted that the main control system establishes a wireless connection with the remote collaboration layer of the AGV robot control system of each AGV robot, receives obstacle information uploaded by each AGV robot, and generates control commands for all AGV robots within a preset range based on the obstacle information and sends them to the corresponding AGV robot control system, thereby achieving obstacle avoidance control.

[0088] S107. When the remote collaboration layer is triggered, the AGV robot waits for the control command issued by the main control system during normal operation, and performs obstacle avoidance control on the AGV robot according to the control command.

[0089] When the risk level is below 0.7 and the remote collaboration layer is triggered, the AGV robot maintains its current path while simultaneously uploading environmental data to the main control system via the 5G communication module. The main control system generates optimal obstacle avoidance commands based on global path planning. While waiting for commands, the AGV robot uses predictive control algorithms to maintain a safe speed, ensuring sufficient maneuverability upon receiving a new command.

[0090] Upon receiving a master control command, the AGV robot executes lane-changing or deceleration operations via a fuzzy PID controller. The system transmits the execution status back in real time, and the master control system adjusts subsequent commands based on the feedback. If no command is received within a preset time, the AGV robot automatically degrades to local computing layer processing.

[0091] In the above embodiments, the AGV robot control system achieves hierarchical and rapid response by establishing a three-level obstacle avoidance mechanism consisting of a hardware control layer, a local computing layer, and a remote collaboration layer. The system can automatically trigger the most appropriate obstacle avoidance level based on the risk level, without waiting for instructions from the main control system each time. When the risk is low, the local computing layer can independently complete obstacle avoidance decisions, quickly selecting the optimal obstacle avoidance path through 16-sector space analysis and passage index evaluation. This hierarchical control strategy reduces obstacle avoidance response time and avoids the processing delays caused by the main control system's involvement in all obstacle avoidance decisions in traditional solutions, effectively reducing the collision risk caused by waiting for the main control system to process the data.

[0092] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating a hierarchical control method for obstacle avoidance in AGV robots, as described in this application.

[0093] S201. When determining the trigger for the local calculation layer based on the risk level value, calculate the motion characteristic coefficient of the AGV robot based on the speed parameters of the AGV robot.

[0094] After triggering the local computing layer, the AGV robot control system first evaluates the dynamic motion characteristics of the AGV robot to determine its obstacle avoidance capabilities in different directions. Specifically, the AGV robot collects its current speed (v), friction coefficient (μ), and acceleration (g) in real time through onboard sensors. Then, the collected current speed (v), friction coefficient (μ), and acceleration (g) are substituted into the motion characteristic coefficient K to calculate the motion characteristic coefficient of the AGV robot using the formula.

[0095] S202. Based on the motion characteristic coefficient, the obstacle avoidance area within the preset range is divided into a forward area, a lateral area, and a backward area.

[0096] Based on the motion characteristic coefficient K, the AGV robot control system divides the obstacle avoidance area within the preset range (usually 10 meters in front, 5 meters to the side, and 3 meters behind) into three areas: forward, side, and backward, and further subdivides them into 16 sector spaces.

[0097] The forward region, centered on the AGV robot's current direction of travel, extends ±180° and is divided into eight 45° sectors. This region contains the AGV robot's main travel path and requires the highest obstacle detection accuracy. The lateral region extends ±90° on each side, divided into three 60° sectors on each side, for a total of six. The sector angles in the lateral region are larger than those in the forward region to balance obstacle detection range and computational complexity. The backward region, centered on the AGV robot's tail, extends ±45° and is divided into two 45° sectors. The backward region has the fewest sectors because AGV robots rarely reverse.

[0098] Optionally, when the K value increases, the fan-shaped division angle of the forward region will decrease (e.g., from 45° to 30°) to improve the resolution of obstacles in front; when the K value decreases, the number of fans in the lateral region can be increased to 8 to enhance the perception of obstacles on the sides, which is not limited here.

[0099] S203. Calculate the spatial weight coefficient based on the sector spatial distribution, and use the spatial weight coefficient as the weighting parameter in the passage index calculation formula.

[0100] Specifically, since the forward direction is the main travel direction of the AGV robot, the obstacle threat level is the highest, so it is assigned the highest weight, with a weight coefficient of 0.4 for the forward area. Similarly, the weight coefficients for the lateral area and the backward area are both 0.3.

[0101] Optionally, when the AGV robot performs special tasks (such as carrying extra-long goods), the system can automatically increase the weight of the lateral area to 0.4 and decrease the weight of the forward area to 0.35 to prioritize the safety of goods transportation. In addition, if a moving obstacle (such as a person following behind) is detected in the rear direction, the weight of the rear direction can be temporarily increased to 0.4.

[0102] S204. Collect the real-time status parameters of each sector space and calculate the instantaneous passage index based on the real-time status parameters.

[0103] After completing the division of the sector space, the AGV robot control system collects parameters such as obstacle density, moving speed and relative distance of each sector space in real time through multi-source sensors, and calculates the instantaneous passage index based on these data.

[0104] The obstacle density (ρ) is determined by the ratio of the number of obstacle point clouds within a statistical sector to the volume of the space. For example, if the volume of a forward sector is 10 m³ and 5 obstacle points are detected, then ρ = 0.5. Relative velocity ( The obstacle's moving speed is calculated by the displacement difference between two consecutive frames of point cloud data, and the relative speed is obtained by combining this with the AGV robot's own speed. The relative distance (d) is taken as the closest distance to the obstacle measured by the lidar. The formula for calculating the real-time passage index is not elaborated here.

[0105] S205. Based on the changing trend of real-time state parameters within the historical sampling period, the Kalman filter algorithm is used to predict the state evolution of each sector within a preset time window to obtain the short-term traffic index.

[0106] The AGV robot control system uses real-time state parameters from historical sampling periods (usually 100ms) to predict the evolution of obstacle states within a future preset time window (such as 0.5 seconds) using a Kalman filter algorithm. The predicted obstacle density, relative speed, and distance are then substituted into the instantaneous passage index formula to obtain the short-term passage index S.

[0107] S206. Establish a long-term evaluation model based on the historical passage records of each sector space, and calculate the long-term passage index corresponding to each sector space.

[0108] The AGV robot control system analyzes historical passage records for each sector to establish a long-term evaluation model and calculate the long-term passage index (L). The local database records data such as the number of passages, average passage time, and number of accidents for each sector over the past 24 hours. The long-term evaluation model uses the Exponentially Weighted Moving Average (EWMA) algorithm to calculate the long-term passage index; the specific formula is as follows:

[0109] , where a is the forgetting factor (usually taken as 0.1). This represents the real-time accessibility index at the current moment. This represents the long-term traffic index at the previous moment.

[0110] Optionally, when a sector fails to pass through three times in a row, the system automatically multiplies the long-term passability index of that area by a penalty coefficient of 0.8; if it passes through five times in a row, it is multiplied by a reward coefficient of 1.2.

[0111] S207. Adjust the weight coefficients of the instant passage index, short-term passage index and long-term passage index according to the complexity of the current scenario, and obtain the passage index by weighted combination.

[0112] The AGV robot control system dynamically adjusts the weighting coefficients of the immediate passage index (I), short-term passage index (S), and long-term passage index (L) based on the current scene complexity, and obtains the final passage index (P) through a weighted combination. Scene complexity is determined by a combination of factors such as obstacle type (fixed / moving), density, and AGV robot speed. For example, when moving obstacles account for more than 50% and the AGV robot speed is >1.5m / s, it is considered a high-complexity scene.

[0113] Furthermore, the weighting coefficients can be dynamically adjusted, as follows: High complexity scenarios: =0.6 (Instantaneous Passage Index Coefficient). =0.3 (Short-term travel index coefficient). =0.1 (Long-Term Passivity Index); Medium Complexity Scenarios: =0.5, =0.3, =0.2; Low complexity scenario: =0.4, =0.2, =0.4.

[0114] S208. After confirming the avoidance direction, calculate the obstacle avoidance level based on the channel width parameter of the target fan-shaped space, and determine the initial obstacle avoidance plan based on the obstacle avoidance level.

[0115] After determining the target sector space, the AGV robot control system fits the obstacle boundary using LiDAR point cloud data, calculates the channel width, and selects an initial obstacle avoidance scheme based on preset standards.

[0116] Specifically, after cluster analysis, the effective channel width (w) of the target sector space is calculated using the minimum bounding rectangle method based on the lidar point cloud data. Then, an initial obstacle avoidance scheme is implemented according to the obstacle avoidance level classification. Specifically, for deceleration obstacle avoidance, w ≥ 2 times the vehicle width (e.g., if the vehicle width is 0.8 meters, w ≥ 1.6 meters); for lane change obstacle avoidance, 1.5 times the vehicle width ≤ w < 2 times the vehicle width; for emergency obstacle avoidance, w < 1.5 times the vehicle width.

[0117] Optionally, under deceleration obstacle avoidance, the system reduces the AGV robot speed to below 1m / s through a PID controller; under lane change obstacle avoidance, the system plans an S-shaped path with a maximum turning angle limited to 30°; under emergency obstacle avoidance, the system triggers the emergency braking system in the hardware control layer.

[0118] S209. After starting the initial obstacle avoidance scheme, if the obstacle avoidance effect parameter of the AGV robot is detected to be lower than the preset target threshold, the scheme upgrade judgment is triggered.

[0119] The AGV robot control system monitors obstacle avoidance performance parameters in real time when executing the initial obstacle avoidance plan. When the parameters fall below a preset threshold, a plan upgrade decision is triggered.

[0120] Specifically, the relative distance change rate must be ≥0.5m / s, the allowable heading deviation range is ±15°, and the obstacle avoidance progress must be updated every 100ms, with a requirement of ≥10% / cycle. When performing threshold comparisons, if the detected relative distance change rate is <0.5m / s and the allowable heading deviation range is >15°, or the obstacle avoidance progress is <5% / cycle, the system determines that the obstacle avoidance effect is poor and triggers the scheme upgrade process.

[0121] The AGV robot control system in this embodiment of the invention is an electronic device. Figure 3 A schematic diagram of the architecture of an electronic device suitable for implementing embodiments of the present invention is shown.

[0122] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0123] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in the embodiments of the present invention.

[0124] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.

[0125] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and can communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.

[0126] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.

[0127] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A hierarchical obstacle avoidance method for AGV robots, applied to an AGV robot control system, wherein the AGV robot control system comprises three obstacle avoidance execution mechanisms: a hardware control layer, a local computing layer, and a remote collaboration layer, characterized in that... The method includes: Collect AGV robot operating parameters, including load value, ground friction coefficient, obstacle distance value, historical accident data, and task time parameter; The operating parameters are input into a risk assessment model, and a risk level value is output. The step of inputting the operating parameters into the risk assessment model and outputting a risk level value includes: the hardware control layer performing a threshold judgment based on the load value, the ground friction coefficient, and the obstacle distance value to obtain a first risk index; the local computing layer inputting the first risk index, the historical accident data, and the task time parameters into a fuzzy evaluator, and calculating a second risk index based on preset fuzzy rules; the remote collaboration layer inputting the first risk index and the second risk index into a pre-trained deep learning model to obtain a third risk index; and determining the risk level value based on the weighted average of the first risk index, the second risk index, and the third risk index. The obstacle avoidance mechanism that is triggered first is determined based on the preset range in which the risk level value is located. The preset range includes a first preset range corresponding to the hardware control layer, a second preset range corresponding to the local computing layer, and a third preset range corresponding to the remote collaboration layer. The hardware control layer is used to perform emergency braking on the AGV robot, and the remote collaboration layer is connected to the main control system. When the local computing layer is triggered, the obstacle avoidance area of ​​the AGV robot within the preset range is divided into 16 sector spaces, and the passage index of each sector space is calculated according to the passage index calculation formula. The target sector with the highest traffic index is determined as the avoidance direction, and the corresponding target obstacle avoidance scheme is selected according to the channel width parameter of the target sector.

2. The method according to claim 1, characterized in that, The step of dividing the obstacle avoidance area of ​​the AGV robot within a preset range into 16 sector-shaped spaces specifically includes: The motion characteristic coefficients of the AGV robot are calculated based on its speed parameters, including current speed, turning angle, and acceleration. Based on the motion characteristic coefficient, the obstacle avoidance area within the preset range is divided into a forward area, a lateral area, and a backward area, with the forward area, lateral area, and backward area corresponding to 8 sector spaces, 6 sector spaces, and 2 sector spaces, respectively. The spatial weight coefficient is calculated based on the sector spatial distribution, and the spatial weight coefficient is used as the weighting parameter in the passage index calculation formula.

3. The method according to claim 1, characterized in that, The step of calculating the passage index of each sector space according to the passage index calculation formula specifically includes: Real-time status parameters of each of the sector spaces are collected, and an instantaneous passage index is calculated based on the real-time status parameters, which include obstacle density, moving speed and relative distance. Based on the changing trend of the real-time state parameters within the historical sampling period, the Kalman filter algorithm is used to predict the state evolution of each sector space within a preset time window to obtain the short-term traffic index. A long-term evaluation model is established based on the historical passage records of each sector space, and the long-term passage index corresponding to each sector space is calculated. The weighting coefficients of the instant passage index, the short-term passage index, and the long-term passage index are adjusted according to the complexity of the current scenario, and the passage index is obtained by weighted combination.

4. The method according to claim 1, characterized in that, The step of selecting the corresponding obstacle avoidance scheme based on the channel width parameter of the target sector space specifically includes: The obstacle avoidance level is calculated based on the channel width parameter of the target fan-shaped space, and the initial obstacle avoidance scheme is determined based on the obstacle avoidance level. The obstacle avoidance levels are, in order from low to high, deceleration obstacle avoidance, lane change obstacle avoidance, and emergency obstacle avoidance. After the initial obstacle avoidance scheme is started, if the obstacle avoidance effect parameters of the AGV robot are detected to be lower than the preset target threshold, the scheme upgrade judgment is triggered. The obstacle avoidance effect parameters include relative distance change rate, heading deviation and obstacle avoidance progress. In the scheme upgrade judgment, when the obstacle avoidance margin index is detected to be greater than the preset upgrade threshold, the initial obstacle avoidance scheme is upgraded to the target obstacle avoidance scheme corresponding to a higher obstacle avoidance level. The obstacle avoidance margin index is determined based on the remaining decision time and maneuver space.

5. The method according to claim 1, characterized in that, After the step of determining the first obstacle avoidance mechanism to be triggered based on the preset range of the risk level value, the method further includes: When the hardware control layer is triggered, emergency braking is immediately executed; After the emergency braking, the vehicle resumes driving according to the control commands issued by the main control system.

6. The method according to claim 1, characterized in that, After the step of determining the first obstacle avoidance mechanism to be triggered based on the preset range of the risk level value, the method further includes: When the remote collaboration layer is triggered, the AGV robot is controlled to wait for the control command issued by the main control system during normal operation. The AGV robot is controlled to avoid obstacles according to the control instructions.

7. An AGV robot control system, characterized in that, The AGV robot control system includes: one or more processors and a memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the AGV robot control system to perform the method as described in any one of claims 1-6.

8. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the AGV robot control system, the AGV robot control system performs the method as described in any one of claims 1-6.

9. A computer program product, characterized in that, When the computer program product is run on the AGV robot control system, it causes the AGV robot control system to perform the method as described in any one of claims 1-6.

Images