AGV motion deduction method based on motion model and event driving

Abstract

The invention relates to the technical field of moving track deduction, and provides a method comprising the following steps: step 1, a target robot generates a running route according to a three-dimensional topology network of a path in a park and a transportation task, and executes the running route; step 2, the target robot reads a current wireless position node, and generates a running route according to the wireless position node and the running route; for the high-risk robot, the target robot monitors the movement direction and the movement speed of the high-risk robot in real time, the movement habits of the high-risk robot are generated according to the movement direction and the movement speed of the high-risk robot, and the movement direction deduction of the high-risk robot is generated based on the movement habits of the high-risk robot and the movement track of the target robot. According to the method, through fusion of the motion model and the event-driven mechanism, the cooperative safety and the operation efficiency of the AGV in a complex multi-robot environment are remarkably improved. The problem of collision risk caused by information isolation between systems and sudden steering behaviors in the prior art is effectively solved.

Description

Technical Field

[0001] This application relates to the field of motion trajectory extrapolation technology, and more specifically, to an AGV motion extrapolation method based on motion models and event-driven methods. Background Technology

[0002] The content in this section provides only background information related to this application and may not constitute prior art.

[0003] AGVs (Automated Guided Vehicles) are a type of automated equipment widely used in industrial production, primarily for cargo transportation in ports, industrial parks, and factories. Currently, the management and control of AGVs mainly rely on visual algorithms and wireless positioning technology to achieve intelligent navigation and target point arrival.

[0004] However, existing transportation networks generally have low safety redundancy. Although AGVs can effectively identify stationary obstacles and accurately determine their relative positions through visual algorithms, thereby adjusting their own direction of travel, in large-scale work sites, there are often a large number of AGVs belonging to different management systems operating simultaneously. Due to the lack of coordination between systems, it is difficult for each AGV to obtain the next movement plan of other robots in real time.

[0005] When two AGVs traveling in opposite directions determine that there is no risk of collision based on visual perception, if one of them suddenly changes its path due to task requirements (such as emergency turning or U-turn), the other AGV may not be able to avoid the collision in time due to physical inertia or insufficient reaction time when it perceives this sudden situation, which may eventually lead to a collision, damage to goods, or even interruption of the continuous operation of the entire system. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide an AGV motion deduction method based on motion model and event-driven method. The AGV motion deduction method based on motion model and event-driven method disclosed in this application can solve the technical problems raised in the background art.

[0007] The objective of this application is achieved through the following technical solution: A motion simulation method for AGVs based on motion models and event-driven approaches includes the following steps: Step 1: The target robot generates a running route based on the three-dimensional topology network of paths within the park and the transportation task, and then executes the running route. Step 2: The target robot reads the current wireless location node and generates a running route based on the wireless location node and the running route; Step 3: The target robot acquires the environment in front of it using binoculars, generates a correction plan based on the environment, applies the correction plan to the running route, and generates an updated route. Step 4: The target robot uses the local area network and binoculars to obtain the location of surrounding obstacle robots and broadcasts updated routes on the local area network; Step 5: Continuously monitor the location of the surrounding obstacle robots using a local area network and binocular lenses, and classify the obstacle robots according to their positional changes; When an obstacle robot receives an updated route broadcast via the local area network, it is classified as a robot of the same type. When the obstacle robot corrects its route to avoid the target robot, the obstacle robot is classified as a robot with the same frequency. When the path of the obstacle robot overlaps with the corrected path of the target robot, the obstacle robot is classified as a high-risk robot. Step 6: For the target robot, for robots of the same type, generate a motion direction deduction for the target robot based on the received updated route; For a robot operating at the same frequency, the target robot generates a motion direction prediction for the robot operating at the same frequency based on the obstacle robot's path correction direction and its own updated path. For high-risk robots, the target robot monitors the movement direction and speed of the high-risk robot in real time. Based on the movement direction and speed of the high-risk robot, it generates the movement habits of the high-risk robot. Based on the movement habits of the high-risk robot and the movement trajectory of the target robot, it generates a movement direction prediction of the high-risk robot.

[0008] This method significantly improves the collaborative safety and operational efficiency of AGVs in complex multi-robot environments by integrating motion models and event-driven mechanisms. Specifically, the method can dynamically perceive and accurately classify surrounding obstacle robots (such as those of the same type, of the same frequency, and high-risk robots), and implement differentiated motion inference strategies for different types. This effectively solves the collision risk problem caused by information isolation between systems and sudden turning behaviors in existing technologies: for robots of the same type, collaborative prediction is achieved based on shared routes; for robots of the same frequency, intention perception inference is performed based on their avoidance intentions; for high-risk robots, high-precision trajectory prediction is performed by monitoring their motion status and habits in real time. Ultimately, this method greatly reduces the probability of collisions caused by sudden path changes, effectively avoids cargo damage, and ensures the efficient, continuous, and stable operation of the transportation system.

[0009] In cross-system AGV collaborative operation scenarios, due to the lack of a global path sharing mechanism, existing methods cannot effectively address three core challenges: the unpredictability of dynamic path adjustments of similar robots, the randomness of sudden avoidance behavior of robots with the same frequency, and the spatiotemporal conflict between the visual obstacle avoidance path of high-risk robots and the actively updated route of the target robot caused by the robot being disconnected from the communication network (only treating the target robot as a static obstacle). Furthermore, traditional algorithms cannot predict such collision points in advance, ultimately resulting in high system collision risk, delayed response, and poor robustness of path planning.

[0010] In some possible embodiments, for robots of the same type, the updated routes and original running routes of robots of the same type are acquired in real time to generate motion direction deductions; For robots operating at the same frequency, the robot's route correction direction is monitored in real time, and the robot's movement habits are generated based on the route correction direction; the robot's movement direction is deduced based on the position of obstacles around the robot and its movement habits. For high-risk robots, the starting point for the high-risk robot's change of position is calculated based on the updated route, the collision position of the high-risk robot, and its movement habits. Based on the current position of the target robot and the position of the high-risk robot, the motion trajectory of the high-risk robot is generated using a path algorithm, and the motion direction of the high-risk robot is deduced based on its motion trajectory.

[0011] This method achieves proactive safety control through a hierarchical deduction mechanism: for robots of the same type, dynamic collaborative deduction is completed based on real-time shared paths; for robots with the same frequency, an intention perception model is constructed by combining their avoidance intention with surrounding obstacles to predict obstacle avoidance trajectories; for high-risk robots, potential collision points are calculated in reverse to deduce the starting point of path change, and path algorithms are integrated to generate high-confidence trajectory predictions, thereby breaking through information barriers and upgrading the response to sudden path conflicts from passive avoidance to proactive deduction, significantly reducing collision risks and ensuring system continuity.

[0012] Existing AGV collaborative systems use a static, fixed classification mode for surrounding robots, which cannot respond to sudden path changes (such as emergency turns, task replanning, and other dynamic behaviors). This results in robots of the same type not being identified as co-current or high-risk objects in time when routes suddenly intersect. Co-current robots are not upgraded to high-risk objects after secondary path corrections, and high-risk robots are still continuously misjudged as threats even after they have successfully avoided them. Ultimately, this leads to a serious disconnect between the motion direction prediction results and real-time risks, increasing the probability of collisions.

[0013] In some possible embodiments, robots of the same type are transformed into robots of the same frequency or high-risk robots according to condition 1; According to condition 2, the same-frequency robot is transformed into a high-risk robot; High-risk robots are converted into risk-free robots based on condition 3; Condition 1: The update route sent by the robot of the same type and the update route of the target robot have spatiotemporal overlap; Condition 2: The movement direction of the robot with the same frequency is readjusted to intersect with the updated route; Condition 3: The movement direction of the high-risk robot is corrected to the principle update route.

[0014] This solution establishes a dynamic robot type conversion mechanism (same type → same frequency / high risk, same frequency → high risk, high risk → no risk), triggers type reclassification based on real-time path spatiotemporal relationships (intersection, intersection, separation), enabling the target robot to accurately perceive the immediate risk status of three types of obstacle objects, and then adaptively adjust the inference strategy.

[0015] Existing AGV systems perform indiscriminate motion simulations for all obstacle robots in dense environments, resulting in a large amount of computing resources being consumed by low-risk, long-distance or low-speed objects. This makes it impossible to prioritize the handling of high-threat, short-distance, high-speed obstacles, causing delays in critical avoidance decisions and significantly increasing the risk of collisions.

[0016] In some possible embodiments, the target robot generates priorities based on the distance information between the obstacle robots and the moving speed of each obstacle robot; the movement direction of the obstacle robot with higher priority is pre-calculated.

[0017] This solution uses a dynamic priority mechanism (based on the real-time distance and speed data of the obstacle robot) to enable the target robot to prioritize the prediction of the movement direction of high-threat targets, thereby achieving focused allocation of computing resources. This reduces redundant calculations for low-risk objects and ensures that the trajectories of close-range high-speed obstacles are predicted in advance, ultimately optimizing system performance and obstacle avoidance safety in complex scenarios.

[0018] Existing AGV systems rely on fixed-period obstacle detection mechanisms, which cannot trigger real-time path updates when obstacles change suddenly (such as when a mobile robot enters or an obstacle moves). Furthermore, there is a time delay between route replanning and the motion deduction of surrounding robots, resulting in delayed obstacle avoidance decisions and asynchronous system responses, which significantly increases the risk of collisions in dynamic environments.

[0019] In some possible embodiments, the target robot detects obstacle information in front of it in real time, and generates an obstacle handling event when the obstacle information in front of it changes. An updated route is generated based on obstacle handling events, and the robot's motion direction prediction is updated synchronously with surrounding obstacles. This solution uses an event-driven mechanism to convert obstacle changes into route update commands in real time and simultaneously link them with the obstacle robot's motion simulation, achieving a breakthrough in "responding to changes". Obstacle changes trigger path replanning with zero delay, and route updates and simulation calculations are executed synchronously to eliminate decision-making time lag.

[0020] Because the AGV obstacle monitoring system does not preload fixed obstacle information and relies on a unified detection strategy, the response mechanism for sudden static obstacles (such as fallen goods) and short-term moving objects (such as forklifts and pedestrians) on the running route is confused, resulting in a waste of computing power for repeated scanning of fixed obstacles.

[0021] In some possible embodiments, the monitoring of obstacle handling events includes the following steps: S1; The target robot records all fixed obstacles on its running route and generates obstacle information; S2: The target robot detects obstacles on its running path in real time. When an obstacle that is stationary relative to the ground appears on the running path, an obstacle handling event is triggered. An obstacle handling event is triggered when a moving object enters the running path within a short period of time.

[0022] This solution achieves static environment modeling by preloading fixed obstacle information and establishes a differentiated event triggering mechanism: static obstacles trigger immediate path replanning, short-term entry of moving objects triggers millisecond-level emergency response, and resource allocation is optimized simultaneously.

[0023] Furthermore, the method for deducing the direction of motion includes the following steps: Z1: Acquire movement habits, which include the distance to obstacles, the direction of deviation relative to obstacles, and the speed after deviation. Z2: For robots operating at the same frequency, calculate the closest position between the robot operating at the same frequency and the target robot. If the closest position is less than the processing distance of the obstacle, then load the motion habits onto the robot operating at the same frequency to obtain the motion direction deduction of the robot operating at the same frequency.

[0024] Z3: For high-risk robots, load the updated route of the target robot and the current route of the high-risk robot, calculate the point where the target robot is at a distance from the high-risk robot when it reaches the obstacle; take the current position of the high-risk robot as the starting point for changing the position of the high-risk robot. Z4: Based on the current location of the target robot and the location of the high-risk robot, generate multiple possible motion trajectories for the high-risk robot using a path algorithm; Among the multiple possible motion trajectories, at least the trajectory with the largest distance from the target robot and the trajectory with the smallest distance from the target robot are included; Z5: Generate a motion direction deduction result based on the motion trajectory that is furthest from the target robot and the motion trajectory that is furthest from the target robot; Z6: Normalize the predicted motion directions of the high-risk robot within the largest and smallest motion trajectories and generate a normal distribution probability map; where the mean of the normal distribution probability map is the predicted motion direction corresponding to the motion habit.

[0025] The obstacle handling distance refers to the distance between the obstacle and the point where the original route is changed to avoid a collision after an obstacle is detected.

[0026] The direction of deviation relative to the obstacle is the relative direction that the obstacle robot chooses to avoid colliding with the obstacle; The speed after deviating from the direction is the speed at which a collision is avoided. Attached Figure Description

[0027] Figure 1 This is a flowchart of an AGV motion simulation method based on motion models and event-driven approaches. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments. The same reference numerals in the accompanying drawings represent the same components. It should be noted that the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0029] Compared to the embodiments shown in the accompanying drawings, feasible embodiments within the scope of this application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or components with different connections, etc. Furthermore, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0030] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” and similar terms used in this specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “upper” and “lower” are used only to indicate relative positional relationships, and these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0031] Example 1: A motion deduction method for AGVs based on motion models and event-driven methods, comprising the following steps: Step 1: The target robot generates a running route based on the three-dimensional topology network of paths within the park and the transportation task, and then executes the running route.

[0032] Based on a pre-loaded high-precision 3D topology network map of the park (including floor aisle slopes, shelf coordinates, and restricted area coordinates) and received transportation task parameters, including the start and end points, the target robot generates a global running route through an optimal path planning algorithm. This route automatically avoids fixed obstacles marked in the 3D map (such as pillars and shelf dead corners) and meets the slope restrictions for cargo transportation, ultimately outputting a coherent and executable path from the start to the end point.

[0033] Step 2: The target robot reads the current wireless location node and generates a running route based on the wireless location node and the running route.

[0034] The target robot uses an onboard UWB ultra-wideband positioning module to read the wireless location nodes in its current area in real time (such as anchor point ID-07 deployed at the corner of the shelf). It then performs spatiotemporal matching of the acquired centimeter-level accurate coordinates (X: 35.2m, Y: 18.7m, Z: 1.2m) with the pre-generated running route. When the lateral deviation between the actual position and the planned path is detected to exceed the 0.5-meter threshold (such as due to slippery ground), a local path refitting algorithm based on B-spline curves is immediately triggered. This generates a smoothly transitioned corrected path segment (8.2 meters in length, radius of curvature ≥ 2 meters) while maintaining the original task objective, ensuring that the dynamic error between the robot's position and the running route is stable within ±0.1 meters.

[0035] Step 3: The target robot acquires the environment in front of it using binocular lenses, generates a correction scheme based on the environment, and applies the correction scheme to generate an updated path when the running route is updated.

[0036] The target robot uses a binocular stereo vision system (baseline distance 30cm, resolution 1280×720@30fps) to collect 3D point cloud data within a 15-meter range in front of it in real time. When it detects dynamic obstacles (such as mobile AGVs) or unloaded static obstacles (such as temporarily stacked cargo boxes) on the planned path, it generates a local correction scheme by fusing YOLOv7 target recognition and RRT obstacle avoidance algorithms. This scheme is then dynamically superimposed on the original running route to generate an updated route (such as modifying the original straight path into a circular arc path with a detour radius of 2.5 meters), achieving 10ms-level environmental response and real-time route optimization.

[0037] The target robot detects obstacle information in front of it in real time. When the obstacle information changes, it generates an obstacle handling event.

[0038] The monitoring methods for obstacle handling events include the following steps: S1; The target robot records all fixed obstacles on its running route and generates obstacle information.

[0039] Before executing its route, the target robot automatically extracts the coordinates of all fixed obstacles along the path based on a pre-loaded 3D topological network map of the park (such as the coordinate set of shelf bases [(X1,Y1,Z1),(X2,Y2,Z2)...], the set of column positions, and the polygonal outline of fixed equipment). The SLAM mapping module performs secondary verification on the point cloud data collected by the binoculars, generating a structured obstacle information table containing obstacle type (static / semi-fixed), 3D bounding box (length, width, and height accuracy ±5cm), and safety distance threshold (0.8 meters). This table is stored in the vehicle-mounted non-volatile memory as a benchmark database for real-time obstacle detection.

[0040] S2: The target robot detects obstacles on its running path in real time. When an obstacle that is stationary relative to the ground appears on the running path, an obstacle handling event is triggered. When a moving object enters the running path within a short period of time, an obstacle handling event is also triggered.

[0041] During its movement, the target robot scans a 10-meter sector area (60° opening angle) ahead of its path using a binocular vision system at a frequency of 30Hz. When it detects a stationary object (such as a fallen cargo box) that is not pre-loaded in the obstacle information table on the planned path and its lateral deviation from the centerline of the planned path is ≤0.4 meters, an obstacle handling event is immediately triggered and marked as "static anomaly". Simultaneously, if a moving object (such as a person or forklift) is detected by optical flow method and intrudes into the path safety boundary within 300ms (intrusion depth ≥0.5 meters), an obstacle handling event with the "dynamic emergency" label is triggered. Both types of events output the obstacle's three-dimensional coordinates, intrusion velocity vector, and safety time window (collision countdown) in real time, driving the subsequent path replanning module.

[0042] The robot generates updated routes based on obstacle handling events and simultaneously updates the motion simulation of the surrounding obstacles.

[0043] Step 4: The target robot uses the local area network and binoculars to obtain the location of surrounding obstacle robots and broadcasts updated routes on the local area network.

[0044] The target robot accesses the local area network via the vehicle-mounted 5G CPE module, and synchronously calls the binocular vision system to identify the three-dimensional coordinates of the robot in the obstacle within a radius of 20 meters. It then encapsulates its updated route (including the path point sequence [(x1,y1,t1),(x2,y2,t2)...] and velocity profile) into a JSON format data packet, which is broadcast through the local area network multicast address at a period of 200ms.

[0045] Step 5: Continuously monitor the location of the surrounding obstacle robots using a local area network and binocular lenses, and classify the obstacle robots according to their positional changes; When an obstacle robot receives an updated route broadcast via the local area network, it is classified as a robot of the same type.

[0046] When the target robot receives the updated route data packet (including path point sequence [(18.3,22.1,10:05:30.500), (18.5,22.3,10:05:31.200)] and velocity profile [1.0 m / s]) broadcast by the obstacle robot via local area network multicast, it is marked as a robot of the same type.

[0047] When the obstacle robot corrects its route to avoid the target robot, the obstacle robot is classified as a robot operating at the same frequency.

[0048] When the target robot detects that the obstacle robot's route correction direction carries clear avoidance characteristics (such as a sudden 15° deflection angle in the path point sequence at a distance of 3 meters from the target robot, and a velocity profile decreasing from 1.2 m / s to 0.6 m / s), and vector analysis confirms that its correction direction points to the outside of the path (with an angle >120° with the target robot's position vector), it is classified as a robot of the same frequency.

[0049] When the path of the obstacle robot overlaps with the corrected path of the target robot, the obstacle robot is classified as a high-risk robot.

[0050] When the updated route of the target robot (path point (28.6, 15.3, 10:07:05.800)) and the route of the obstacle robot (path point (28.5, 15.4, 10:07:05.500)) overlap in time and space (coordinate deviation < 0.2 meters and time window overlap 300 ms), and the collision probability calculated by the collision probability model is > 85%, it is immediately marked as a high-risk robot.

[0051] The target robot labels each obstacle robot it detects, thus accurately identifying each obstacle robot and its corresponding type.

[0052] Robots of the same type are transformed into robots of the same frequency or high-risk robots according to condition 1; robots of the same frequency are transformed into high-risk robots according to condition 2; high-risk robots are transformed into risk-free robots according to condition 3. Condition 1: The update route sent by the robot of the same type and the update route of the target robot have spatiotemporal overlap; Condition 2: The movement direction of the robot with the same frequency is readjusted to intersect with the updated route; Condition 3: The movement direction of the high-risk robot is corrected to the principle update route.

[0053] Specifically, when the target robot detects a new route broadcast by a robot of the same type (the path points intersect with its own updated route in a spatiotemporal manner (coordinate deviation of 0.12 meters / time window overlap of 200ms), it immediately starts calculating the collision probability model: if the angle between the velocity vector and the local trajectory is <30°, it is upgraded to a high-risk robot; if its velocity profile shows a 40% deceleration and a 20° outward deviation of the path (active avoidance feature), it is downgraded to a robot of the same frequency and the cooperative avoidance protocol is activated.

[0054] When the target robot detects that a robot of the same frequency (originally in the avoidance state) suddenly corrects its path, and through kinematic calculations it finds that the robot's movement direction has changed from outward deflection to inward tangency, and the spatiotemporal overlap between the corrected path point and the target robot's own path exceeds the threshold, the target robot is immediately reclassified as a high-risk robot.

[0055] When the trajectory prediction model verifies that the high-risk robot's movement direction shows a safe separation trend from the local updated route, and the final path point is at least 2.3 meters away from the local path, the target robot is downgraded to a risk-free robot.

[0056] Step 6: For the target robot, for the same type of robot, generate a motion direction deduction of the same type of robot based on the received updated route; for the same type of robot, obtain the updated route and the original running route of the same type of robot in real time, and generate a motion direction deduction.

[0057] The target robot analyzes the updated routes broadcast by similar robots and generates its future 5-second motion direction prediction by combining a linear extrapolation algorithm with a velocity profile: the predicted pose sequence is output with a step size of 0.5 seconds and an error radius of ≤0.15 meters (calibrated based on historical positioning data).

[0058] For robots operating at the same frequency, the robot's route correction direction is monitored in real time, and its movement habits are generated based on the route correction direction. The robot's movement direction is also predicted based on the location of obstacles around it and its movement habits. For the target robot, the robot's movement direction is predicted based on the obstacle robot's route correction direction and its own updated route.

[0059] For high-risk robots, the target robot monitors the movement direction and speed of the high-risk robot in real time. Based on the movement direction and speed of the high-risk robot, it generates the movement habits of the high-risk robot. Based on the movement habits of the high-risk robot and the movement trajectory of the target robot, it generates a movement direction prediction of the high-risk robot.

[0060] For high-risk robots, the starting point for the high-risk robot's change of position is calculated based on the updated route, the collision position of the high-risk robot, and its movement habits. Based on the current position of the target robot and the position of the high-risk robot, the motion trajectory of the high-risk robot is generated using a path algorithm, and the motion direction of the high-risk robot is deduced based on its motion trajectory.

[0061] The target robot generates priorities based on the distance between the obstacle robots and the movement speed of each obstacle robot; the movement direction of the obstacle robot with higher priority is pre-calculated and deduced.

[0062] Example 2: Based on Example 1, Example 2 provides a motion direction deduction method, which is mainly used for synchronous robots and high-risk robots.

[0063] The method for deducing the direction of motion includes the following steps: Z1: Acquire movement habits, which include the distance to obstacles, the direction of deviation relative to obstacles, and the speed after deviation.

[0064] The obstacle handling distance refers to the distance between the obstacle and the location where the original route is changed to avoid a collision after an obstacle is detected.

[0065] The direction of deviation relative to the obstacle is the relative direction that the obstacle robot chooses to avoid colliding with the obstacle; The speed after deviating from the direction is the speed at which a collision is avoided.

[0066] Movement habits can be detected by binocular cameras or obtained from the park's regulations. If the obstacle robot does not perform an avoidance action within the target robot's field of vision, the movement habit is set as the initial habit. If an avoidance action occurs, the movement habit is updated based on the extracted avoidance action.

[0067] Z2: For robots operating at the same frequency, calculate the closest position between the robot operating at the same frequency and the target robot. If the closest position is less than the processing distance of the obstacle, then load the motion habits onto the robot operating at the same frequency to obtain the motion direction deduction of the robot operating at the same frequency.

[0068] There is no direct collision path between the same-frequency robot and the target robot. The same-frequency robot will avoid the target robot only when the two robots are close to each other.

[0069] Z3: For high-risk robots, load the updated route of the target robot and the current route of the high-risk robot, calculate the point where the target robot is at a distance from the high-risk robot when it reaches the obstacle; take the current position of the high-risk robot as the starting point for changing the position of the high-risk robot. Z4: Based on the current location of the target robot and the location of the high-risk robot, generate multiple possible motion trajectories for the high-risk robot using a path algorithm; Among the multiple possible motion trajectories, at least the trajectory with the largest distance from the target robot and the trajectory with the smallest distance from the target robot are included; Z5: Generate a motion direction deduction result based on the motion trajectory that is furthest from the target robot and the motion trajectory that is furthest from the target robot; Z6: Normalize the predicted motion directions of the high-risk robot within the largest and smallest motion trajectories and generate a normal distribution probability map; where the mean of the normal distribution probability map is the predicted motion direction corresponding to the motion habit.

[0070] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A motion deduction method for AGVs based on motion models and event-driven approaches, characterized in that, The steps include: Step 1: The target robot generates a running route based on the three-dimensional topology network of paths within the park and the transportation task, and then executes the running route. Step 2: The target robot reads the current wireless location node and generates a running route based on the wireless location node and the running route; Step 3: The target robot acquires the environment in front of it using binoculars, generates a correction plan based on the environment, applies the correction plan to the running route, and generates an updated route. Step 4: The target robot uses the local area network and binoculars to obtain the location of surrounding obstacle robots and broadcasts updated routes on the local area network; Step 5: Continuously monitor the location of the surrounding obstacle robots using a local area network and binocular lenses, and classify the obstacle robots according to their positional changes; When an obstacle robot receives an updated route broadcast via the local area network, it is classified as a robot of the same type. When the obstacle robot corrects its route to avoid the target robot, the obstacle robot is classified as a robot with the same frequency. When the path of the obstacle robot overlaps with the corrected path of the target robot, the obstacle robot is classified as a high-risk robot. Step 6: For the target robot, for robots of the same type, generate a motion direction deduction for the target robot based on the received updated route; For a robot operating at the same frequency, the target robot generates a motion direction prediction for the robot operating at the same frequency based on the obstacle robot's path correction direction and its own updated path. For high-risk robots, the target robot monitors the movement direction and speed of the high-risk robot in real time. Based on the movement direction and speed of the high-risk robot, it generates the movement habits of the high-risk robot. Based on the movement habits of the high-risk robot and the movement trajectory of the target robot, it generates a movement direction prediction of the high-risk robot.

2. The AGV motion deduction method based on motion model and event-driven approach according to claim 1, characterized in that... For robots of the same type, the updated routes and original running routes of robots of the same type are obtained in real time, and motion direction inferences are generated. For robots operating at the same frequency, the direction of route correction for the robots is monitored in real time, and the movement habits of the robots are generated based on the direction of route correction. The robot's motion direction is predicted based on the location of obstacles around it and its movement habits. For high-risk robots, the starting point for the high-risk robot's change of position is calculated based on the updated route, the collision position of the high-risk robot, and its movement habits. Based on the current position of the target robot and the position of the high-risk robot, the motion trajectory of the high-risk robot is generated using a path algorithm, and the motion direction of the high-risk robot is deduced based on its motion trajectory.

3. The AGV motion deduction method based on motion model and event-driven approach according to claim 1, characterized in that, Robots of the same type are transformed into either high-frequency robots or high-risk robots based on condition 1. According to condition 2, the same-frequency robot is transformed into a high-risk robot; High-risk robots are converted into risk-free robots based on condition 3; Condition 1: The update route sent by the robot of the same type and the update route of the target robot have spatiotemporal overlap; Condition 2: The movement direction of the robot with the same frequency is readjusted to intersect with the updated route; Condition 3: The movement direction of the high-risk robot is corrected to the principle update route.

4. The AGV motion deduction method based on motion model and event-driven approach according to claim 3, characterized in that, The target robot generates priorities based on the distance between the obstacle robots and the movement speed of each obstacle robot; the movement direction of the obstacle robot with higher priority is pre-calculated and deduced.

5. The AGV motion deduction method based on motion model and event-driven approach according to claim 1, characterized in that... ； The target robot detects obstacle information in front of it in real time. When the obstacle information changes, it generates an obstacle handling event. The robot generates updated routes based on obstacle handling events and simultaneously updates the motion simulation of the surrounding obstacles.

6. The AGV motion deduction method based on motion model and event-driven approach according to claim 5, characterized in that, The monitoring methods for obstacle handling events include the following steps: S1; The target robot records all fixed obstacles on its running route and generates obstacle information; S2: The target robot detects obstacles on its running path in real time. When an obstacle that is stationary relative to the ground appears on the running path, an obstacle handling event is triggered. An obstacle handling event is triggered when a moving object enters the running path within a short period of time.

7. The AGV motion deduction method based on motion model and event-driven approach according to claim 6, characterized in that, The method for deducing the direction of motion includes the following steps: Z1: Acquire movement habits, which include the distance to obstacles, the direction of deviation relative to obstacles, and the speed after deviation. Z2: For robots operating at the same frequency, calculate the closest position between the robot operating at the same frequency and the target robot. If the closest position is less than the processing distance of the obstacle, then load the motion habits onto the robot operating at the same frequency to obtain the motion direction deduction of the robot operating at the same frequency.

8. The AGV motion deduction method based on motion model and event-driven approach according to claim 7, characterized in that, Methods for deducing the direction of motion also include: Z3: For high-risk robots, load the updated route of the target robot and the current route of the high-risk robot, calculate the point where the target robot is at a distance from the high-risk robot when it reaches the obstacle; take the current position of the high-risk robot as the starting point for changing the position of the high-risk robot. Z4: Based on the current location of the target robot and the location of the high-risk robot, generate multiple possible motion trajectories for the high-risk robot using a path algorithm; Among the multiple possible motion trajectories, at least the trajectory with the largest distance from the target robot and the trajectory with the smallest distance from the target robot are included; Z5: Generate a motion direction deduction result based on the motion trajectory that is furthest from the target robot and the motion trajectory that is furthest from the target robot; Z6: Normalize the predicted motion directions of the high-risk robot within the largest and smallest motion trajectories and generate a normal distribution probability map; where the mean of the normal distribution probability map is the predicted motion direction corresponding to the motion habit.

9. The AGV motion deduction method based on motion model and event-driven approach according to claim 7, characterized in that, The obstacle handling distance refers to the distance between the obstacle and the point where the original route is changed to avoid a collision after an obstacle is detected. The direction of deviation relative to the obstacle is the relative direction that the obstacle robot chooses to avoid colliding with the obstacle; The speed after deviating from the direction is the speed at which a collision is avoided.

Images