Task remaining duration prediction method and device, electronic equipment and storage medium

Abstract

The application provides a task remaining duration prediction method and device, electronic equipment and storage medium, belonging to the technical field of automatic logistics, comprising: acquiring a planning path and a road network congestion point set of a target automatic guided transport vehicle; calculating the path distance between the target congestion point and the current position in the path; calculating the turning time consumption based on the node distribution characteristics; determining the penalty weight that decays with the increase of the path distance, combining the basic congestion penalty to determine the congestion additional time consumption; and obtaining the task remaining duration based on the basic time, the turning time consumption and the congestion additional time consumption. The application introduces the turning time consumption and constructs a congestion time consumption quantification mechanism based on distance decay, accurately evaluates the dynamic influence of spatial road network details and real-time congestion on vehicle operation, overcomes the distortion defects of traditional uniform speed prediction, and significantly improves the accuracy of multi-vehicle task prediction and the overall collaborative efficiency of the scheduling system in complex limited scenarios.

Description

Technical Field

[0001] This invention relates to the field of automated logistics technology, and in particular to a method, apparatus, electronic device, and storage medium for predicting the remaining time of a task. Background Technology

[0002] In modern production workshops, a large number of Automated Guided Vehicles (AGVs) are usually distributed to coordinate material transfer. The scheduling system needs to allocate new tasks reasonably based on the precise time required for each AGV to perform its current task. Therefore, accurately predicting the remaining time of each AGV's task is a core technical requirement for optimizing scheduling decisions and ensuring the efficient operation of the entire logistics flow system.

[0003] To meet the above requirements, existing task remaining time prediction schemes usually rely on historical statistical data and theoretically set indicators to estimate time. These schemes mainly obtain the total length of the AGV's planned path in advance and combine it with the extracted average driving speed of the AGV to abstract the AGV's driving process into an ideal uniform motion process. Then, based on the simple mathematical relationship of distance divided by speed, the total time required for each AGV to complete the task is estimated.

[0004] Existing prediction schemes simply abstract AGV movement as uniform motion, which fails to reflect the inherent physical losses caused by the complex road topology within the actual confined space on the traffic rhythm. Furthermore, when complex road networks exhibit dynamically fluctuating traffic congestion, current technologies cannot effectively quantify the differentiated impacts of traffic anomalies in different spatial dimensions on the current travel process. This results in significant distortion between the estimated time period and the actual task duration in application scenarios with frequently changing road network conditions, easily leading to scheduling system allocation misalignment or deadlock. Summary of the Invention

[0005] This invention provides a method, device, electronic device, and storage medium for predicting the remaining time of a task. It addresses the shortcomings of existing technologies that idealize the AGV's driving process as uniform motion while ignoring the topological loss of the restricted road network space and lack the ability to quantify the impact of dynamic congestion spatial differences, resulting in prediction distortion. It achieves a refined characterization of the spatial details of obstacles on the driving path and the impact of real-time road condition evolution, thereby dynamically assessing the actual time required for the AGV to perform the task with high fidelity, and empowering the efficient collaboration and optimal task scheduling of multi-AGV logistics systems.

[0006] This invention provides a method for predicting the remaining time of a task, applied to a multi-autonomous vehicle logistics system, comprising the following steps: Obtain the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location. Calculate the path distance between each target congestion point and the current location, where the target congestion point is a node in the serialized nodes that overlaps with the set of congestion points; The turning time is obtained based on the number of turns on the planned path. The number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes. Based on the path distance between each target congestion point and the current location, a penalty weight corresponding to each target congestion point is determined, and the penalty weight decreases as the path distance increases; Based on the penalty weights of each target congestion point and the preset basic congestion penalty, the additional congestion time is determined. The remaining mission duration of the target automated guided vehicle is obtained based on the basic travel time of the planned route, the turning time, and the additional time spent due to congestion.

[0007] According to a method for predicting the remaining duration of a task provided by the present invention, determining the penalty weight corresponding to each of the target congestion points based on the path distance between each target congestion point and the current location includes: Obtain a preset congestion forgetting factor, which is used to characterize the spatial decay rate of the impact of congestion on vehicle driving in the road network. Multiply the congestion forgetting factor by the path distance from the target congestion point to the current location to obtain the impact attenuation degree of the target congestion point; Based on the impact attenuation degree, the penalty weight corresponding to the target congestion point is determined so that the penalty weight decreases exponentially with the increase of the path distance.

[0008] According to the method for predicting the remaining duration of a task provided by the present invention, the step of determining the additional congestion time based on the penalty weight of each target congestion point and a preset basic congestion penalty includes: By aggregating the penalty weights of all the target congestion points on the planned path, the overall congestion impact coefficient of the planned path is obtained. Obtain a preset base congestion penalty, which represents the baseline delay time under single-point congestion conditions; The overall congestion impact coefficient is weighted and calculated with the basic congestion penalty to determine the additional congestion time corresponding to the planned path.

[0009] According to the method for predicting the remaining duration of a task provided by the present invention, the step of obtaining the turning time based on the number of path turns on the planned path includes: In the serialization nodes of the planned path, extract at least three consecutive path nodes in sequence; Based on the first two and last two path nodes of the three path nodes, the entry direction vector and exit direction vector of the target automated guided vehicle are determined respectively. Obtain the steering angle between the inbound direction vector and the outbound direction vector; If the steering angle is within a preset turning determination range, the middle path node among the three path nodes is marked as the turning position, so as to count the number of path turns. The number of turns on the path is correlated with the preset baseline time for a single turn to obtain the turning time corresponding to the planned path.

[0010] According to the method for predicting the remaining duration of a task provided by the present invention, after determining the additional congestion time based on the penalty weights of each target congestion point and a preset basic congestion penalty, the method further includes: Determine whether the current location is a target congestion point in the set of congestion points; If the current location is the target congestion point, the additional congestion time is updated to a preset constant penalty value, which is greater than the threshold of additional congestion time accumulated along the planned path. Based on the updated additional congestion time, the scheduling system of the multi-autonomous vehicle logistics system is triggered to implement a task avoidance strategy for the target automated guided vehicle.

[0011] According to the method for predicting the remaining time of a task provided by the present invention, before obtaining the planned path of the target automated guided vehicle and the set of congestion points in the current road network, the method further includes: Obtain the number of vehicles on each road segment element in the current road network, and the smooth driving speed of the vehicles on each road segment element; If the number of vehicles corresponding to the road segment element is greater than or equal to a preset first threshold, and the average smooth driving speed is less than a preset second threshold, then the location of the vehicle that meets the preset low speed condition on the road segment element is determined as a multi-vehicle congestion point. If it is detected that the dwell time of a single vehicle at a non-permitted long-term dwell node exceeds a preset third threshold, the location of the single vehicle will be determined as an abnormal congestion point for the single vehicle. The congestion points are aggregated from the multi-vehicle congestion points and the single-vehicle abnormal congestion points to generate the congestion point set.

[0012] According to a method for predicting the remaining duration of a task provided by the present invention, obtaining the smooth driving speed of vehicles on each road segment primitive in the current road network includes: Extract multiple frames of historical state data for each vehicle within a preset sliding time window, wherein the historical state data includes timestamps and location coordinates; Based on the timestamp difference and position coordinate difference between adjacent frames in the multi-frame historical state data, the single-step time difference and the single-step displacement difference based on Manhattan distance are calculated for each vehicle. The cumulative time is obtained by summing the time differences of each step, and the cumulative displacement is obtained by summing the displacement differences of each step. Based on the cumulative displacement and the cumulative time, the smooth driving speed of each vehicle is determined.

[0013] According to the method for predicting the remaining duration of a task provided by the present invention, before obtaining the number of vehicles on each road segment primitive in the current road network, the method further includes: Obtain the map topology data of the current road network and construct an undirected graph with path nodes as vertices and node connections as the connection relationships; Count the number of connecting lines between each path node in the undirected graph to determine the degree of each path node; Identify path nodes in the undirected graph whose degree is not equal to 2, and mark the identified path nodes as branch nodes; Along the path node link with a node degree of 2, path tracing is performed between two adjacent branch nodes, and the resulting continuous non-branching path segments are determined as each of the path segment primitives.

[0014] According to a method for predicting the remaining mission duration provided by the present invention, after obtaining the remaining mission duration of the target automated guided vehicle, the method further includes: Get the waiting time of the currently pending tasks and the preset latest assignment time threshold; Determine whether the waiting time has reached the latest allocation time threshold, and determine whether the additional congestion time exceeds the preset time limit threshold; If the waiting time reaches the latest allocation time threshold and the additional congestion time exceeds the time limit threshold, then the additional congestion time is removed from the remaining task time. Based on the remaining time of the task after removing the additional time spent due to congestion, the scheduling system of the multi-autonomous vehicle logistics system is triggered to forcibly assign the currently unassigned task to the target automated guided vehicle.

[0015] According to a method for predicting the remaining duration of a task provided by the present invention, the basic travel time is determined based on the following method: Determine the total path length of the planned path; The current load status of the target automated guided vehicle is obtained, including both empty and fully loaded states. Based on the current load status, determine the corresponding target driving speed from the preset unloaded driving speed and the preset fully loaded driving speed; The base travel time of the planned route is obtained by calculating the ratio of the total path length to the target travel speed.

[0016] The present invention also provides a task remaining time prediction device, applied to a multi-autonomous vehicle logistics system, comprising: The data acquisition module is used to acquire the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location. The distance calculation module is used to calculate the path distance between each target congestion point and the current location, wherein the target congestion point is a node in the serialized node that overlaps with the set of congestion points; The turning time determination module is used to obtain the turning time based on the number of turns on the planned path. The number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes. The weight determination module is used to determine the penalty weight corresponding to each of the target congestion points based on the path distance between each target congestion point and the current position, wherein the penalty weight decreases as the path distance increases; The additional time determination module is used to determine the additional congestion time based on the penalty weight of each target congestion point and the preset basic congestion penalty; The duration prediction module is used to obtain the remaining mission duration of the target automated guided vehicle based on the basic travel time of the planned route, the turning time, and the additional time spent due to congestion.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the task remaining time prediction method as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the task remaining duration prediction method as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the task remaining time prediction method as described above.

[0020] The method, apparatus, electronic device, and storage medium for predicting remaining task duration provided by this invention accurately assess the dynamic impact of spatial road network details and real-time congestion on vehicle operation by introducing turning time and constructing a congestion time quantification mechanism based on distance decay. This overcomes the distortion defects of traditional uniform speed prediction and significantly improves the accuracy of multi-vehicle task prediction and the overall collaborative efficiency of the scheduling system in complex and constrained scenarios. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the task remaining time prediction method provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the process for determining penalty weights provided by the present invention.

[0024] Figure 3 This is a schematic diagram of the penalty weight calculation formula provided by the present invention.

[0025] Figure 4 This is a schematic diagram of the process for determining the additional time spent on congestion provided by the present invention.

[0026] Figure 5 This is a schematic diagram of the process for obtaining the turning time provided by the present invention.

[0027] Figure 6 This is a schematic diagram of vector calculation for turn recognition provided by the present invention.

[0028] Figure 7 This is a schematic diagram of the congestion detection process provided by the present invention.

[0029] Figure 8 This is a schematic diagram of the smooth driving speed calculation provided by the present invention.

[0030] Figure 9 This is a schematic diagram of the road segment determination primitive provided by the present invention.

[0031] Figure 10 This is a schematic diagram of the task remaining time prediction device provided by the present invention.

[0032] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0035] The terms "first," "second," etc., used in this invention are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more.

[0036] In the existing field of traffic flow prediction and congestion detection, conventional outdoor traffic condition prediction solutions are typically designed for wide, multi-lane open road networks with high network complexity and random vehicle trajectories. Congestion is mostly caused by factors such as excessive traffic volume. However, when directly applied to indoor logistics scenarios in industrial plants, they exhibit significant inapplicability. Factory logistics scenarios, especially intensive discrete manufacturing plants such as home appliance assembly, 3C electronics, and automotive parts manufacturing, have unique physical constraints and business logic. First, the internal road network of factories is mostly narrow passages, one-way streets, or strictly restricted spaces with two-way single lanes, making it impossible for AGVs to overtake on most sections. Second, congestion in factory scenarios is often caused by production line anomalies and scheduling deadlocks, making the prolonged stay of a single vehicle at non-special nodes (such as non-workstation nodes) a unique congestion phenomenon in narrow factory passages.

[0037] To overcome the failure of existing general transportation planning technologies in factory spaces, this invention proposes a method and apparatus specifically adapted for predicting the remaining time of multi-AGV tasks in industrial scenarios. This method relies on a topological network map with node attribute semantics (clearly distinguishing industrial semantic nodes such as regular driving, workstations, rest areas, and charging stations). By combining the unique spatial constraints of a factory with the planned paths of AGVs under unified scheduling by the logistics system, it can accurately identify hidden congestion caused by production line anomalies and scheduling deadlocks, thereby providing highly reliable task prediction and scheduling support for industrial logistics systems.

[0038] The following is combined with Figures 1-11 The present invention provides a detailed description of the method, apparatus, electronic device, and storage medium for predicting the remaining time of a task.

[0039] First, it should be noted that the task remaining time prediction method provided in this embodiment of the invention can be executed by hardware devices such as central control equipment, scheduling servers, or processors of edge computing nodes deployed in a multi-AGV logistics system. This multi-AGV logistics system can be widely used in intensive discrete manufacturing plants such as home appliance assembly, 3C electronics manufacturing, and automotive parts processing, as well as modern warehousing centers, for the coordinated management of a large number of AGVs for efficient material handling. For the sake of brevity, the executing entities of the actions will be omitted in the following process descriptions of the embodiments.

[0040] Figure 1 This is a flowchart illustrating the task remaining time prediction method provided by the present invention, as shown below. Figure 1 As shown, including but not limited to the following steps: Step 11: Obtain the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location.

[0041] In one alternative embodiment, the specific vehicle requiring duration prediction, i.e., the target AGV, is first identified. The task instructions currently being executed or about to be executed by the target AGV can be read from the upper-level scheduling task queue to obtain its planned path from its current location to the target location, i.e., the destination workstation or shelf.

[0042] The planned path is represented in the data structure as a series of sequential nodes arranged in the order of travel. In the industrial field, these sequential nodes can be QR code nodes affixed to the ground, radio frequency identification (RFID) electronic tag nodes, or coordinate waypoints in a hybrid topological geometry map built based on LiDAR.

[0043] Meanwhile, this embodiment obtains the current set of congestion points in the road network in real time through internal interfaces or external communication. This set of congestion points contains global abnormal location information of traffic delays and obstructions across the entire logistics road network at the current moment. It should be noted that this embodiment can either dynamically generate this set of congestion points using its own monitoring algorithm or retrieve it directly from a third-party traffic monitoring system.

[0044] Step 12: Calculate the path distance between each target congestion point and the current location. The target congestion point is the node in the serialized node that overlaps with the set of congestion points.

[0045] In order to accurately assess the substantial obstacles that road network congestion poses to the target AGV, the scheduling system will traverse the acquired planned paths.

[0046] Specifically, the spatial coordinates or node identifiers of each serialized node in the planned path can be cross-compared with the global set of congestion points to filter out the overlapping nodes. These overlapping nodes represent the traffic congestion locations that the target AGV will inevitably pass through on its future travel route, i.e., the target congestion points.

[0047] After identifying all target congestion points, the edge lengths between nodes can be accumulated segment by segment along the connection direction of the serialized nodes, or a pre-built road network distance index table can be directly called to calculate the path distance of each target congestion point relative to the current position of the target AGV.

[0048] Step 13: Obtain the turning time based on the number of turns on the planned path. The number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes.

[0049] Due to the severely limited space within factories, which are mostly narrow passages, one-way streets, or two-way single lanes, the target AGV needs to frequently accelerate, decelerate, and turn during its operation. To quantify this physical loss, this embodiment extracts the distribution characteristics of the serialized nodes in the planned path. These distribution characteristics can be the spatial geometric trend or azimuth change information formed by the lines connecting adjacent nodes.

[0050] Furthermore, based on these distribution characteristics, the turning points in the planned path where the target AGV needs to change its direction can be identified, and the number of turns in the entire planned path can be counted.

[0051] In an alternative embodiment, the statistically obtained number of path turns can be combined with a pre-calibrated theoretical time parameter for a single turn, for example, by multiplication or lookup table mapping, to obtain the turning time required for the target AGV to complete all turning actions.

[0052] Step 14: Determine the penalty weight corresponding to each target congestion point based on the path distance between each target congestion point and the current location. The penalty weight decreases as the path distance increases.

[0053] Research has revealed that in real-world logistics dynamics, congestion points closer to the target AGV cause more precise and significant direct delays. Conversely, congestion points farther from the AGV are more likely to have dissipated or been alleviated by the time the target AGV arrives after a long journey. Based on this finding, this embodiment assigns a penalty weight to each target congestion point according to the calculated path distance. The determination of this penalty weight follows the core logic of decreasing as the path distance increases.

[0054] As an optional implementation, a segmented, step-by-step decreasing logic can be used to assign constant weights from high to low to the target congestion points in the near-distance, medium-distance, and long-distance intervals, respectively. Alternatively, a linear decay mathematical model can be used to continuously reduce the penalty weight value proportionally based on the path distance. This embodiment does not impose specific limitations on this approach.

[0055] Step 15: Determine the additional congestion time based on the penalty weights of each target congestion point and the preset basic congestion penalty.

[0056] In this embodiment, the preset basic congestion penalty is mainly used to characterize the average additional delay benchmark value generated when a standard congestion point is completely stopped or travels at extremely low speed. The magnitude of this average additional delay benchmark value can be calculated from the factory's early statistical data.

[0057] After obtaining the penalty weights corresponding to each target congestion point, these weights can be mathematically calculated and combined with a preset base congestion penalty. For example, by applying each penalty weight separately to the base congestion penalty and then summing the results, the additional time spent due to current traffic congestion along the entire planned route can be determined. In this way, the complex and variable discrete congestion situation can be quantified into a compensation value that can be directly used in the time dimension.

[0058] Step 16: Based on the basic travel time of the planned route, the turning time, and the additional time spent due to congestion, obtain the remaining mission duration of the target automated guided vehicle.

[0059] Finally, the scheduling system can integrate the analysis results from all the above dimensions, summarizing the basic travel time required for the target AGV to traverse the planned path under ideal, uniform speed conditions without any obstructions, the turning time caused by its specific spatial structure, and the additional time due to congestion caused by dynamic road conditions. This can be done by methods such as direct addition or weighted summation, ultimately accurately predicting the actual time required for the target AGV to complete its current task, i.e., the remaining task duration. The obtained remaining task duration can then be output to guide the dispatch of subsequent orders or route replanning.

[0060] The task remaining time prediction method provided in this embodiment conducts in-depth analysis of the distribution characteristics of serialized nodes in the planned path, incorporates the turning time of any target AGV into the remaining time prediction, and constructs a penalty weight mechanism based on path distance decay to quantify the impact of target congestion points. It not only accurately considers the inherent physical time loss caused by frequent turning in narrow channel environments, but also conforms to the fact that long-distance traffic congestion tends to dissipate dynamically over time. It overcomes the prediction distortion caused by the traditional method's simple idealization of AGV movement as uniform motion, providing a high-fidelity time decision basis for multi-AGV logistics systems and significantly improving the overall task scheduling rationality and collaborative operation efficiency.

[0061] Figure 2 This is a schematic diagram of the process for determining penalty weights provided by the present invention. Figure 3 This is a schematic diagram of the penalty weight calculation formula provided by the present invention, combined with... Figure 2 and Figure 3 As shown, determining the penalty weight corresponding to each target congestion point based on the path distance between each target congestion point and the current location mainly includes, but is not limited to, the following steps: Step 21: Obtain a preset congestion forgetting factor, which is used to characterize the spatial decay rate of the impact of congestion on vehicle driving in the road network.

[0062] In an alternative embodiment, in order to accurately model the dissipation characteristics of congestion effects mathematically, a key congestion forgetting factor is pre-introduced and obtained. Congestion-induced forgetting factor It is mainly used to characterize the spatial decay rate of the impact of congestion on the driving of a target AGV in a road network. Its core logic is to define how quickly the impact of congestion on the current vehicle weakens as the distance increases.

[0063] As an alternative implementation method, the congestion-forgetting factor The preferred value range is between 0.01 and 0.1, and its specific value can be calibrated and dynamically adjusted through on-site multi-AGV operation testing. If congestion forgetting factor The larger the value, the faster the spatial decay rate, meaning that only congestion points very close to the AGV will have a significant impact on the remaining time; if the congestion forgetting factor... The smaller the value, the slower the decay, meaning that traffic congestion at a greater distance will still retain a larger weight in the prediction.

[0064] Step 22: Multiply the congestion forgetting factor by the path distance from the target congestion point to the current location to obtain the influence attenuation degree of the target congestion point.

[0065] After obtaining the congestion forgetting factor Then, for each identified target congestion point, this embodiment will extract the path distance from that point to the current location, which was calculated earlier. .

[0066] As an optional implementation, the path distance will be... With congestion and forgetting factor Perform a direct product operation. The result of multiplying the two is... This is the influence attenuation degree. Therefore, the influence attenuation degree is an intermediate variable that integrates static physical distance and dynamic traffic dissipation rate. In essence, it prepares the scaled distance parameters for subsequent nonlinear mapping.

[0067] Step 23: Determine the penalty weight corresponding to the target congestion point based on the impact attenuation degree, so that the penalty weight decreases exponentially with the increase of the path distance.

[0068] After obtaining the impact attenuation degree, in order to closely fit the nonlinear characteristics of traffic condition evolution in real industrial scenarios, the scheduling system will further determine the penalty weight corresponding to the target congestion point based on the impact attenuation degree.

[0069] Specifically, combined Figure 3The calculation logic shown can be performed using an exponential decay model, so that the penalty weight decreases exponentially with the increase of path distance. The specific calculation formula can be: ; in, For the first i The penalty weight for each target congestion point; It is a natural constant; From current position to number i The path distance to each target congestion point. The above calculation method based on the exponential decay model ensures that nearby congestion is given extremely high attention (weight close to 1), while the weight of distant congestion quickly converges to close to 0.

[0070] The task remaining time prediction method provided in this invention introduces a congestion forgetting factor that characterizes the spatial decay rate and constructs a penalty weight calculation by combining the exponential decay model of path distance. This transforms the topological proximity in the distance dimension into a quantitative assessment of traffic congestion uncertainty, which conforms to the actual driving experience that the closer the congestion, the more accurate its impact, and the farther the congestion, the easier it is to dissipate. This achieves a differentiated and nonlinear assessment of the impact of congestion, thereby significantly improving the accuracy and reliability of the prediction results under complex dynamic road networks.

[0071] Figure 4 This is a schematic diagram of the process for determining additional congestion time provided by the present invention. For example... Figure 4 As shown, based on the above embodiments, the determination of additional congestion time based on the penalty weights of each target congestion point and the preset basic congestion penalty mainly includes, but is not limited to, the following steps: Step 41: Aggregate the penalty weights of all the target congestion points on the planned path to obtain the overall congestion impact coefficient of the planned path.

[0072] In one alternative embodiment, the planned path of the AGV often has more than one congestion node. In order to comprehensively assess the overall traffic resistance faced by the entire path, the penalty weight of each target congestion point calculated above is extracted, and these penalty weights scattered in different spatial locations are aggregated.

[0073] As an optional implementation, the above aggregation process can be carried out by direct summation, that is, by accumulating the penalty weights of all the target congestion points predicted to be traversed on the planned path. Alternatively, other optional mathematical fusion methods, such as weighted averaging, can be used to calculate the overall congestion impact coefficient of the planned path. Generally speaking, the obtained overall congestion impact coefficient is a dimensionless comprehensive index that can objectively reflect the degree of cumulative congestion deterioration of the current path after considering the spatial distance attenuation effect.

[0074] Step 42: Obtain the preset basic congestion penalty, which represents the baseline delay time under single-point congestion conditions.

[0075] When converting dimensionless coefficients into specific physical time losses, this embodiment will continue to introduce a time-dimensional reference constant as a preset basic congestion penalty.

[0076] The baseline congestion penalty is mainly used to characterize the baseline delay time (usually in seconds) generated when any target AGV encounters a standard single-point congestion state.

[0077] As an optional implementation, the baseline delay time can be calculated based on the factory's early historical operation logs and statistical data, or it can be dynamically and adaptively assigned based on the overall traffic density in the current factory area. This embodiment does not specifically limit this.

[0078] Step 43: Perform a weighted calculation of the overall congestion impact coefficient and the basic congestion penalty to determine the additional congestion time corresponding to the planned path.

[0079] After obtaining the overall congestion impact coefficient and the basic congestion penalty with time dimensions, the scheduling system can further perform a weighted calculation to obtain the final time compensation value. As an optional calculation method, a product operation can be used to determine the additional congestion time; the specific calculation formula could be: ; in, The calculated additional time spent in congestion; Penalties based on congestion; The target number of congested points; For the first i The penalty weight for each target congestion point; This represents the overall congestion impact coefficient after aggregation.

[0080] Assuming there are two predicted congestion points ahead of the planned path, with penalty weights of 0.8 and 0.2 respectively after distance decay calculation, the aggregated overall congestion impact coefficient, calculated using the aforementioned formula for calculating additional congestion time, is 1.0. Further assuming a preset base congestion penalty of 30 seconds, the additional time due to congestion on the planned path is ultimately determined to be 30 seconds through multiplicative weighted calculation.

[0081] The implementation example first aggregates the penalty weights of all congested nodes on the path to generate an overall impact coefficient, and then performs a weighted calculation with the basic congestion penalty with a time dimension. This decouples and integrates the spatial impact in the distance dimension with the delay benchmark in the time dimension, enabling discrete traffic congestion events of varying severity to be accurately converted into continuous time loss metrics. This ensures that the scheduling system can still output dynamic time prediction compensation with low latency and high accuracy when facing massive AGV concurrent scheduling, significantly enhancing the robustness and engineering application value of the prediction model.

[0082] Figure 5 This is a schematic diagram of the process for obtaining the turning time provided by the present invention, as shown in the figure. Figure 5 As shown, the process of obtaining the turning time based on the number of turns on the planned path mainly includes, but is not limited to, the following steps: Step 51: Extract at least three consecutive path nodes from the serialized nodes of the planned path.

[0083] In one alternative embodiment, to identify the vehicle's steering actions during driving, it is necessary to obtain a sequence of nodes with spatial context association. Specifically, a sliding window approach can be used to extract three or more consecutive path nodes sequentially along the planned path's serialized nodes.

[0084] Figure 6 This is a schematic diagram of vector calculation for turn recognition provided by the present invention, as shown below. Figure 6 As shown, as an optional example, suppose the three consecutive path nodes extracted are respectively , and .

[0085] Step 52: Based on the first two and last two path nodes of the three path nodes, determine the entry direction vector and exit direction vector of the target automated guided vehicle.

[0086] Combination Figure 6 The spatial geometric relationships shown, after obtaining continuous path nodes, utilize a spatial coordinate system to construct a geometric vector that characterizes the target AGV's driving orientation. Specifically, this can be achieved using... Figure 6 The first two path nodes and Construct an entry direction vector representing the vehicle's entry into the area. Similarly, the next two path nodes and Construct a vector representing the direction of exit of a vehicle entering or leaving the area. .

[0087] The two vectors mentioned above mathematically accurately represent the target AGV at intermediate nodes along its path. The change in the physical trajectory of motion over time.

[0088] Step 53: Obtain the steering angle between the inbound direction vector and the outbound direction vector.

[0089] After constructing the two direction vectors mentioned above, we can further... Figure 6 As shown, geometric operations are used to solve for the deflection angle between the two.

[0090] As an alternative implementation, the turning angle between the two vectors can be determined by calculating the relationship between the dot product of the two vectors and their magnitudes. Its specific calculation formula can be expressed as: ; .

[0091] The above steering angle It can objectively reflect the heading deflection range that the target AGV needs to make in this local section of the road.

[0092] Step 54: If the steering angle is within the preset turning determination range, then mark the middle path node among the three path nodes as the turning position, so as to count the number of path turns.

[0093] In industrial plants, such as home appliance assembly plants and 3C electronics manufacturing plants, logistics channels are often distributed in a cross-shaped orthogonal grid pattern, and target AGVs typically make right-angle turns at intersections. Considering the slight coordinate offsets that may occur during map construction, this embodiment will preset a turning judgment range with a certain degree of tolerance.

[0094] As an optional implementation, the turning range can be set between 75° and 105°. The aforementioned calculated turning angle... Compare with the judgment range, if 75° At 105°, logically it is confirmed that the target AGV has performed a valid turning action at this point, and the middle path node among the three nodes will be... Mark it as a valid turning position.

[0095] By traversing the entire planned path and summing up all the turning positions that meet the criteria, the total number of turns on the entire planned path can be calculated. .

[0096] Step 55: The number of turns on the path is correlated with the preset single turn baseline time to calculate the turn time corresponding to the planned path.

[0097] When a target AGV performs a turning maneuver, it typically involves a series of complex kinematic processes, including deceleration, stationary rotation or differential turning, and re-acceleration, which results in additional time loss. This embodiment calculates the number of turns along the path. Then, the preset baseline time for a single turn will be obtained. The baseline turning time is usually based on the vehicle's mechanical performance parameters, such as maximum angular velocity and acceleration / deceleration, and is calibrated to obtain the average turning time.

[0098] Furthermore, the number of turns on the path can be correlated with the baseline time for each turn. For example, the following multiplication operation can be used to finally obtain the turning time corresponding to the planned path. : .

[0099] This invention, by introducing vector construction and dot product geometric operations of continuous nodes in a two-dimensional topology map, can identify effective turning actions that conform to the physical constraints of the factory and quantitatively convert the number of turns into additional time loss. This overcomes the blind spot of existing technologies that only calculate Euclidean or Manhattan distances while ignoring the inherent losses of frequent acceleration, deceleration and steering of AGVs. It deeply restores the real kinematic performance under narrow and restricted channels, making the predicted task travel time more consistent with the underlying physical control logic of the vehicle, and significantly improving the accuracy and matching degree of time estimation.

[0100] Based on the above embodiments, as an optional embodiment, after determining the additional congestion time based on the penalty weights of each target congestion point and the preset basic congestion penalty, the following steps are also included, but are not limited to: Determine whether the current location is a target congestion point in the set of congestion points.

[0101] In one optional embodiment, after calculating the distance decay of congestion ahead of the planned path, the system will also retrospectively examine the current state of the target AGV itself. Specifically, the current physical coordinates or node identifier of the target AGV can be extracted and compared with the obtained global congestion point set to determine whether the target AGV is currently at a known traffic congestion node.

[0102] If the current location is the target congestion point, the additional congestion time is updated to a preset constant penalty value, which is greater than the threshold of additional congestion time accumulated along the planned path.

[0103] Specifically, if the comparison results indicate that the target AGV's current location is itself the target congestion point, this means that the target AGV is already deeply stuck in traffic, and there is great uncertainty as to when it will be able to get out of the predicament. In this case, the time compensation calculated using conventional cumulative methods can no longer accurately reflect its actual waiting cost. Therefore, this embodiment will logically perform a forced overwrite, directly updating the originally calculated additional congestion time to a very large preset constant penalty value. .

[0104] As an optional implementation, this constant penalty value The value of is set to be much greater than the additional congestion time threshold that may accumulate along the conventional planning path; for example, it is set to 10 times or even higher than the aforementioned basic congestion penalty. By imposing such an extremely high time penalty, the predicted remaining task time of the target AGV can be artificially and drastically increased.

[0105] Furthermore, based on the updated additional congestion time, the scheduling system of the multi-autonomous vehicle logistics system is triggered to implement a task avoidance strategy for the target automated guided vehicle.

[0106] In this embodiment, because the target AGV is given a very large constant penalty value, its final output task remaining time will far exceed that of other AGVs in normal driving state. Therefore, under the overall order dispatch logic, based on the updated extra time spent due to excessive congestion, a task avoidance strategy targeting the target AGV can be naturally triggered.

[0107] Optionally, the task avoidance strategy may specifically manifest as follows: when bidding for or allocating new tasks, actively excluding them from the candidate queue, or significantly reducing their order acceptance priority, and instead prioritizing other AGVs that are not in a congested state to perform new material handling tasks.

[0108] The technical solution provided in this embodiment introduces a self-check of the congestion status of the target AGV at its current position and assigns a very large constant penalty value when it gets congested. This enables deep linkage between the underlying time prediction logic and the upper-level deadlock prevention scheduling strategy, thereby ensuring that the target AGV at the traffic paralysis node will no longer be blindly stuffed with new orders. This effectively prevents the further deterioration and cascading spread of local congestion, and establishes an optimal dynamic balance between prediction accuracy and the rationality of global task distribution, effectively improving the overall operating efficiency of multi-AGV systems in complex and confined spaces.

[0109] Figure 7 This is a schematic diagram of the congestion detection process provided by the present invention, as follows: Figure 7As shown, before obtaining the planned route of the target automated guided vehicle and the set of congestion points in the current road network, the dynamic generation process of the set of congestion points mainly includes, but is not limited to, the following steps: Obtain the number of vehicles on each road segment primitive in the current road network, and the smooth driving speed of the vehicles on each road segment primitive.

[0110] In one optional embodiment, the continuous road network of the entire logistics plant area is pre-divided into multiple road segment primitives with independent topological semantics. During the operation of the scheduling system, the positioning information of each AGV is acquired in real time and mapped and matched to the corresponding road segment primitives, thereby counting the total number of AGVs located on each road segment primitive at the current time.

[0111] Meanwhile, in order to prevent sudden speed changes caused by single-frame positioning errors or brief stops, this embodiment will obtain the smooth driving speed of each AGV within a past sliding time window to more realistically reflect its recent movement trend.

[0112] If the number of vehicles corresponding to the road segment element is greater than or equal to a preset first threshold, and the average smooth driving speed is less than a preset second threshold, then the location of the vehicle that meets the preset low-speed condition on the road segment element is determined as a multi-vehicle congestion point.

[0113] To accurately identify multi-vehicle congestion caused by traffic overload, this embodiment sets up dual judgment conditions. Specifically, for each road segment element, the average smooth travel speed of all AGVs within it is calculated.

[0114] As an optional implementation, the preset first threshold can be set to 2 units by default, and the preset second threshold can be set to 0.1 m / s by default. These two thresholds can also be dynamically and adaptively adjusted according to the production rhythm of the factory at different times. Thus, when the number of AGVs gathered on a certain road segment is greater than or equal to the first threshold, and the average speed of that road segment is less than the second threshold, the physical appearance will highly match the state of multiple vehicles continuously lingering at low speed.

[0115] After identifying the congested road segment, this embodiment performs a secondary screening of individual AGVs to prevent those that are still moving normally on the edge of the road segment from being mistakenly identified as congestion sources. Only when an AGV within the road segment's basic element has a smooth driving speed that meets preset low-speed conditions, such as its own speed being below 0.1 m / s, is its precise location or node officially identified as a multi-vehicle congestion point.

[0116] If it is detected that the dwell time of a single vehicle at a non-permitted long-term dwell node exceeds a preset third threshold, then the location of the single vehicle is determined as an abnormal congestion point for the single vehicle.

[0117] In addition to the aforementioned congestion caused by multiple vehicles queuing, factory scenarios also present the problem of single vehicles being stuck for extended periods due to production line anomalies or scheduling deadlocks. To address this, the scheduling system incorporates a parallel branch for determining abnormal vehicle dwell times.

[0118] refer to Figure 7 As shown, nodes not permitted to remain for extended periods refer to regular travel lane nodes excluding charging pile nodes, rest area nodes, and fixed workstation nodes, which have special functions. If a single AGV is detected to have remained stationary at these regular nodes for a period exceeding a preset third threshold, such as the default 60 seconds or a duration dynamically set based on task urgency, then regardless of the total number of AGVs on that road segment, an abnormal congestion is directly determined, and the location of that single AGV is identified as a single-vehicle abnormal congestion point.

[0119] Finally, the congestion points of multiple vehicles and the abnormal congestion points of a single vehicle are summarized to generate the congestion point set.

[0120] The scheduling system merges and deduplicates all multi-vehicle congestion points and single-vehicle abnormal congestion points identified by the two parallel detection branches mentioned above, and uniformly summarizes them to generate a congestion point set with a global perspective. This congestion point set is then output and applied to predict the remaining task time of the target AGV in the aforementioned embodiment.

[0121] This embodiment effectively filters out instantaneous noise in the underlying positioning by using smooth driving speed, and adopts a dual-link perception logic that combines multi-vehicle low-speed queuing based on the average speed of primitives with single-vehicle abnormal stop based on node dwell time. This is in line with the unique business constraints of discrete manufacturing plants, which include both dense traffic queuing and narrow channel deadlocks or production line failures causing hidden congestion. It completely solves the problem of missed detections that are easy to occur when relying solely on traffic flow to determine congestion in traditional methods, and significantly enhances the capture rate of complex abnormal road conditions and the robustness of system perception.

[0122] Figure 8 This is a schematic diagram of the smooth driving speed calculation provided by the present invention, as shown below. Figure 8 As shown, obtaining the smooth driving speed of vehicles on each road segment primitive in the current road network includes: Extract multiple frames of historical state data for each vehicle within a preset sliding time window, wherein the historical state data includes timestamps and location coordinates; Based on the timestamp difference and position coordinate difference between adjacent frames in the multi-frame historical state data, the single-step time difference and the single-step displacement difference based on Manhattan distance are calculated for each vehicle. The cumulative time is obtained by summing the time differences of each step, and the cumulative displacement is obtained by summing the displacement differences of each step. Based on the cumulative displacement and the cumulative time, the smooth driving speed of each vehicle is determined.

[0123] In an alternative embodiment, to avoid speed measurement errors caused by single sensor scan delay or positioning drift, this embodiment further introduces a time-dimensional smoothing filtering mechanism.

[0124] In terms of specific operation, for each AGV in operation, a first-in-first-out status queue is maintained, and a sliding time window is constructed by backtracking a preset time period based on the current time.

[0125] As an optional implementation, the preset duration can be set to 5 to 15 seconds (e.g., preferably 5.0 seconds). From this sliding time window, multiple frames of historical status data continuously reported by the AGV within that time period are extracted. Each frame of historical status data contains at least the exact position coordinates of the AGV at the moment of acquisition. x , y (and the corresponding timestamp information.)

[0126] After extracting the historical state data from multiple frames within the window, these data frames are arranged in chronological order according to their timestamps. Then, adjacent frames are iterated through, and the timestamp difference between them is calculated to determine the single-step time difference for each state update. At the same time, the positional coordinate differences between these two frames of data are calculated.

[0127] Considering that the driving channels inside industrial plants are mostly orthogonal grid structures with horizontal and vertical lines, the actual driving trajectory of AGVs is closer to a right-angled broken line than a straight line between two points. Therefore, combined with the attached... Figure 8 The calculation principle shown in this embodiment uses Manhattan distance to solve for the single-step displacement difference. Its calculation formula can be expressed as: ; in, and These represent the absolute values ​​of the horizontal and vertical coordinate differences between two adjacent frames in a two-dimensional coordinate system.

[0128] Furthermore, after calculating the single-step time difference and single-step displacement difference of all adjacent frames within the sliding time window, this embodiment performs macroscopic integration of these discrete micro data.

[0129] Specifically, all the individual time differences are summed together to obtain the total time spent within the window, i.e., the accumulated time. : ; Similarly, all single-step displacement differences based on Manhattan distance... Perform a continuous summation operation to obtain the total travel distance within the window, i.e., the cumulative displacement. : .

[0130] After obtaining the two cumulative values ​​representing physical displacement and physical time respectively, the actual average driving efficiency of the AGV within the time window can be calculated by the ratio of the two values. ; in, This refers to the determined smooth driving speed.

[0131] As another optional fault-tolerant implementation, if the scheduling system detects that there are fewer than two valid historical data frames extracted within the sliding time window, such as when the vehicle has just been turned on or communication has just been restored, the calculation of the smoothing speed can be degraded to the instantaneous speed calculated based on the latest two adjacent frames of data, so as to ensure the continuous availability of the algorithm process.

[0132] The embodiments of the present invention construct a sliding time window mechanism and use Manhattan distance to calculate the cumulative displacement. This not only fits the actual physical geometric constraints of the indoor gridded channel, but also effectively filters out the noise caused by instantaneous speed change caused by single-frame positioning errors or extremely short-term point fine-tuning of the vehicle. This makes the calculated smooth driving speed more stable and realistically reflect the passage status of the AGV in a recent period of time.

[0133] Figure 9 This is a flowchart illustrating the process of determining road segment primitives provided by the present invention, as shown below. Figure 9 As shown, the left image illustrates the original road network topology. Path nodes in the left image are distinguished by different colors: black nodes represent regular path nodes, including AGV passage paths and idle workstations; red nodes represent functional nodes, encompassing line-side loading / unloading points, material storage nodes, and AGV charging nodes. The connections between nodes represent path relationships, collectively forming the complete road network spatial topology. Figure 9 The right figure shows the result of road segment primitive division based on node degree analysis: blue solid dots mark the position of each bifurcation node, and thick gray line segments represent continuous non-branching path segments between adjacent bifurcation nodes. Each segment corresponds to an independent road segment primitive unit.

[0134] Specifically, combined Figure 9 As shown, before obtaining the number of vehicles on each road segment primitive in the current road network, this embodiment also includes, but is not limited to, the following execution steps: Obtain the map topology data of the current road network and construct an undirected graph with path nodes as vertices and node connections as the connection relationships; Count the number of connecting lines between each path node in the undirected graph to determine the degree of each path node; Identify path nodes in the undirected graph whose degree is not equal to 2, and mark the identified path nodes as branch nodes; Along the path node link with a node degree of 2, path tracing is performed between two adjacent branch nodes, and the resulting continuous non-branching path segments are determined as each of the path segment primitives.

[0135] In one optional embodiment, map topology data of the factory's internal underlying navigation system is first acquired. This map topology data mainly includes a set of path nodes consisting of QR codes affixed to the ground, magnetic strip nodes, or virtual waypoints constructed based on laser SLAM, as well as driving network segments that connect the nodes.

[0136] To facilitate spatial dimensionality reduction analysis of complex factory passages using graph theory algorithms, these isolated path nodes are abstracted as vertices of the graph at the logical level, and the connecting routes between adjacent nodes that allow AGVs to travel are abstracted as edges of the connection relationship, thereby constructing an undirected graph that can accurately reflect the connectivity status of the physical passages in the real factory area.

[0137] After constructing the undirected graph, all vertices in the graph are traversed and statistically analyzed. Specifically, the number of connected edges directly linking each path node is counted, and this number represents the degree of that path node. The degree of a node directly and uniquely reflects its topological position within the entire transportation network, that is, how many directional branches the AGV can choose from at that location.

[0138] Furthermore, this embodiment will perform logical classification and key node identification based on the degree of each node obtained from statistics.

[0139] As an optional implementation, this involves prioritizing the selection of path nodes with a degree of 2 or less and marking them as bifurcation nodes. Specifically, a node with a degree of 1 typically corresponds to a lane endpoint, dead end, or isolated workstation end in the physical road network; nodes with a degree of 3 or greater correspond to crossroads, T-junctions, or multi-directional branching points. A node with a degree of exactly 2 represents a node with only one inbound and one outbound direction, i.e., a typical linear continuous traffic node. Therefore, this embodiment extracts all special nodes with a degree of 2 and marks them as bifurcation nodes, effectively providing start and end anchor points for subsequent road network segmentation.

[0140] Finally, this embodiment uses each identified branch node as a starting point and executes a topology traversal algorithm such as depth-first search or breadth-first search to trace forward along the links composed of ordinary path nodes with a degree of 2. This forward traversal process continues until the next branch node or endpoint node is encountered. At this point, the continuous sequence of ordinary nodes with a degree of 2 extracted between these two adjacent branch nodes constitutes a continuous, non-branching path segment. The scheduling system will formally identify this independently extracted continuous, non-branching path segment as an independent road segment primitive.

[0141] As an optional extended implementation, after all road segment primitives are determined, the attribute information of the original road network can be used to supplement the judgment and assign the traffic direction attribute to each road segment primitive, such as one-way traffic, two-way traffic, or two-way conflict intersection, so as to enrich the business semantics of the road segment primitives.

[0142] The task remaining time prediction method provided in this embodiment automatically and modularly segments the complex continuous road network of a factory area into multiple continuous path segments without branching, using node degree statistics and undirected graph depth tracking algorithms from graph theory. This refined road segment primitive partitioning mechanism ensures a high degree of logical consistency between the detection carrier of congestion events and the physical scenario of congestion occurring in continuous restricted channels in reality. Compared to the traditional coarse-grained approach that uses entire large areas or extremely small single points as congestion monitoring units, this invention establishes the optimal spatial granularity of congestion perception that balances local details and computational overhead, thereby significantly improving the spatial positioning accuracy of subsequent multi-vehicle congestion identification.

[0143] Based on the above embodiments, considering the strict production cycle time constraints often present in industrial production sites, individual high-priority tasks cannot be indefinitely delayed due to global road network congestion. Therefore, this embodiment further defines how to establish a dynamic balance between prediction accuracy and task urgency. Specifically, after obtaining the remaining task time of the target automated guided vehicle, the following steps are included, but are not limited to: Get the waiting time of the currently pending tasks and the preset latest assignment time threshold; Determine whether the waiting time has reached the latest allocation time threshold, and determine whether the additional congestion time exceeds the preset time limit threshold; If the waiting time reaches the latest allocation time threshold and the additional congestion time exceeds the time limit threshold, then the additional congestion time is removed from the remaining task time. Based on the remaining time of the task after removing the additional time spent due to congestion, the scheduling system of the multi-autonomous vehicle logistics system is triggered to forcibly assign the currently unassigned task to the target automated guided vehicle.

[0144] In an optional embodiment, for currently pending tasks that have not yet been dispatched in the upper-level task queue, such as a call order issued by a workstation that urgently needs material replenishment, the scheduling system will continuously monitor its life cycle and obtain the waiting time of the currently pending task from the date of its generation to the current moment.

[0145] Simultaneously, based on the business priority or production line cycle time requirements of the currently pending task, a preset latest allocation time threshold is obtained. This latest allocation time threshold represents the maximum tolerable idle limit for the currently pending task; exceeding this limit may lead to serious consequences such as production line shutdown or capacity loss.

[0146] After obtaining the aforementioned time parameters, this embodiment further performs parallel dual-condition determination. On the one hand, it determines whether the waiting time of the currently assigned task has reached or exceeded the latest assignment time threshold to assess the absolute urgency of the task itself; on the other hand, it reviews the additional congestion time calculated for the target AGV in the aforementioned steps to determine whether the time exceeds the preset upper limit threshold to assess whether the congestion obstacles currently faced by the target AGV are too severe.

[0147] If both of the above judgments are true, it means that the scheduling system is in a contradictory state: the task to be assigned is about to time out or even default, and a vehicle needs to be dispatched urgently; however, the target AGV, as a candidate, has serious traffic congestion on its route to the destination, resulting in a very high predicted value for the total remaining task time, including the additional time spent due to congestion.

[0148] To break this deadlock caused by mutual avoidance of tasks, this embodiment will logically perform a downgrade intervention, that is, forcibly deduct or completely remove the additional time spent due to congestion from the already calculated remaining task time, so that the time prediction value of the target AGV returns to a lower level under the ideal traffic conditions.

[0149] Since the extra time spent due to congestion is eliminated, the target AGV's scheduling bidding score or time evaluation ranking within the scheduling system will be significantly improved. Based on this updated prediction value after removing the congestion penalty, the upper-level logic is triggered to forcibly issue and assign the currently pending task to the target AGV, requiring it to ignore the congestion obstacles ahead and immediately proceed to execute the order.

[0150] This embodiment introduces dual monitoring based on task waiting time and congestion severity, and decisively removes extra time spent due to congestion to trigger forced allocation in critical task situations. It establishes a circuit breaker logic that prioritizes congestion avoidance under normal conditions and ensures production rhythm under extreme conditions. This not only retains the value of high-precision time prediction for overall traffic management under normal operating conditions, but also eliminates scheduling blind spots caused by large-scale road network paralysis that lead to the indefinite postponement of high-priority tasks. It maximizes the overall robustness of the multi-AGV logistics system in maintaining the continuous operation of the core production line under extremely severe road conditions.

[0151] Based on the above embodiments, as an optional embodiment, to ensure that the baseline value of the final prediction result has extremely high accuracy, this embodiment provides a method for calculating the base driving time in the aforementioned steps. Specifically, the base driving time is determined based on the following methods, mainly including but not limited to the following steps: Determine the total path length of the planned path; The current load status of the target automated guided vehicle is obtained, including both empty and fully loaded states. Based on the current load status, determine the corresponding target driving speed from the preset unloaded driving speed and the preset fully loaded driving speed; The base travel time of the planned route is obtained by calculating the ratio of the total path length to the target travel speed.

[0152] In one optional embodiment, the base travel time represents the ideal constant-speed travel time of the target AGV under conditions of no traffic obstruction and without considering the effects of frequent turning deceleration. Therefore, this embodiment first calculates the absolute spatial span of the target AGV from its current position to its target position. Specifically, this can be achieved by progressively accumulating the adjacent distances between each serialized node along the planned path, thereby accurately determining the total path length of the planned path. L The unit is usually meters.

[0153] In industrial automated logistics scenarios, the physical load of the target AGV directly determines the output power limit of its chassis motor and its safety braking strategy. The scheduling system obtains the current load status of the target AGV by reading the sensor data reported in real time by the underlying controller (such as the vehicle-mounted PLC) or by querying the task assignment log of the upper-level scheduling system.

[0154] As an alternative implementation, in order to simplify the model and take into account practicality, the load state is clearly divided into an empty state without carrying any materials and a fully loaded state carrying or towing materials on the shelf.

[0155] Due to dynamic constraints and on-site safety braking distance requirements, the cruising speed limits of the target AGV under different load conditions are physically distinct. The scheduling system pre-calibrates the unloaded and fully loaded speeds to prevent safety accidents caused by excessive inertia when fully loaded. The fully loaded speed is usually significantly lower than the unloaded speed. The scheduling system performs a logical mapping based on the current load status: if the current state is fully loaded, the preset fully loaded speed is selected as the target speed for this task; if the current state is unloaded, the preset unloaded speed is selected as the target speed. .

[0156] After obtaining the total travel distance in the spatial dimension and the theoretical speed limit in the physical attribute dimension, a division operation can be performed based on the classical uniform kinematics formula. Specifically, the total path length can be... L The selected target driving speed will be used as the dividend. Used as a divisor in ratio calculations Finally, the basic travel time of the planned route is calculated. The calculated baseline driving time, as a pure, ideal benchmark, will be combined and added together with the previously calculated turning time and additional time spent in congestion.

[0157] The technical solution provided in this embodiment completely eliminates the drawback of using a uniform fixed average speed for all dispatched vehicles in the traditional time prediction by introducing the real-time physical load status of the target AGV itself into the decision logic for the value of the driving speed, thus making the calculated basic driving time have a high physical fidelity.

[0158] It should be noted that the specific algorithms and data structures listed in the above specific embodiments for data acquisition, speed calculation, congestion determination, and model prediction are mainly preferred examples provided to clearly illustrate the core technical concept of the present invention. Those skilled in the art should understand that these specific technical means can be flexibly modified and extended without departing from the core principles of the present invention.

[0159] As an optional embodiment, regarding the acquisition and application mechanism of map topology data, the data source of the map topology data is not limited to a statically preset set of QR code nodes. In more complex dynamic factory environments, the underlying map data acquired by the scheduling system can also be extended to an occupancy grid map, a topology-geometry hybrid map, or a high-precision lane-level navigation map built in real time based on LiDAR. As long as its underlying logic still performs spatial division of road segment primitives based on these map structures, it is applicable to the framework of the present invention.

[0160] As another optional embodiment, in addition to using the ratio of displacement to accumulated time within the aforementioned sliding time window for calculating the smooth driving speed of vehicles, the scheduling system can also introduce higher-order temporal processing mechanisms. For example, signal processing algorithms such as Kalman filtering or exponential moving average (EMA) can be used to smooth the speed across multiple frames, or an instantaneous speed prediction network based on sequence models such as RNN / LSTM can be directly deployed to further improve the smoothness and noise resistance of speed extraction.

[0161] As an optional implementation, the dispatching system is not limited to determining congestion segments and congestion points based solely on vehicle quantity and average speed thresholds. To adapt to more flexible traffic flow patterns, the determination logic can also comprehensively incorporate vehicle density on the segment elements, median deviation of vehicle passage time on the segment, or the element passage probability predicted by the dispatching system's backend AI model, thereby achieving a multi-dimensional comprehensive congestion determination.

[0162] As another optional embodiment, in addition to the aforementioned exponential decay formula, more complex time-series or nonlinear modeling methods such as task path-based equivalent feature vector regression, Gaussian process regression, or Support Vector Regression (SVR) can also be used for modeling distance forgetting models and time prediction. As long as the core input features of the model retain path length, number of turns, and congestion overlap, and the output provides a prediction of the remaining task time, it conforms to the computational intent of differentiated quantification of congestion impact in this invention.

[0163] As an optional implementation, to further improve prediction accuracy under complex and variable working conditions, additional auxiliary features can be introduced on top of the core features mentioned above, regarding the feature input dimensions of the task remaining time prediction model. For example, the ground slope of the road segment primitive, the current remaining battery power of the target automated guided vehicle, the task priority of the currently assigned task (e.g., whether it includes expedited or queue-jumping attributes), and the historical success rate of the road segment during the corresponding time period can be input as auxiliary features into the prediction model.

[0164] Figure 10 This is a schematic diagram of the task remaining time prediction device provided by the present invention, as shown below. Figure 10 As shown, it mainly includes, but is not limited to: The data acquisition module 101 is used to acquire the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location. The distance calculation module 102 is used to calculate the path distance between each target congestion point and the current position, wherein the target congestion point is a node in the serialized node that overlaps with the set of congestion points; The turning time determination module 103 is used to obtain the turning time based on the number of path turns on the planned path. The number of path turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes. The weight determination module 104 is used to determine the penalty weight corresponding to each of the target congestion points based on the path distance between each target congestion point and the current position, wherein the penalty weight decreases as the path distance increases; The additional time determination module 105 is used to determine the additional congestion time based on the penalty weight of each target congestion point and the preset basic congestion penalty; The duration prediction module 106 is used to obtain the remaining mission duration of the target automated guided vehicle based on the basic travel time of the planned route, the turning time, and the additional congestion time.

[0165] It should be noted that the task remaining time prediction device provided in this embodiment can execute the task remaining time prediction method provided in any of the above embodiments during actual operation, which will not be described in detail here.

[0166] The task remaining time prediction device provided by this invention accurately assesses the dynamic impact of spatial road network details and real-time congestion on vehicle operation by introducing turning time and constructing a congestion time quantification mechanism based on distance decay. It overcomes the distortion defects of traditional uniform speed prediction and significantly improves the accuracy of multi-vehicle task prediction and the overall collaborative efficiency of the scheduling system in complex and constrained scenarios.

[0167] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 11As shown, the electronic device may include: a processor 110, a communications interface 120, a memory 130, and a communications bus 140, wherein the processor 110, the communications interface 120, and the memory 130 communicate with each other through the communications bus 140. The processor 110 can call logical instructions in the memory 130 to execute a method for predicting the remaining task duration. This method includes: acquiring the planned path of the target automated guided vehicle (AGV) and a set of congestion points in the current road network, the planned path including serialized nodes from the current location to the target location; calculating the path distance between each target congestion point and the current location, the target congestion point being a node in the serialized nodes that overlaps with the set of congestion points; obtaining the turning time based on the number of turns on the planned path, the number of turns being determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes; determining the penalty weight corresponding to each target congestion point based on the path distance between each target congestion point and the current location, the penalty weight decreasing as the path distance increases; determining additional congestion time based on the penalty weight of each target congestion point and a preset basic congestion penalty; and obtaining the remaining task duration of the target AGV based on the basic travel time of the planned path, the turning time, and the additional congestion time.

[0168] Furthermore, the logical instructions in the aforementioned memory 130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0169] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the task remaining time prediction method provided in the above embodiments, the method comprising: obtaining the planned path of the target automated guided vehicle and the set of congestion points in the current road network, the planned path comprising serialized nodes from the current location to the target location; calculating the path distance between each target congestion point and the current location, wherein the target congestion point is one of the serialized nodes that coincides with the set of congestion points. The nodes are defined as follows: Turning time is obtained based on the number of turns on the planned path, where the number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes; Penalty weights are determined for each target congestion point based on the path distance between each target congestion point and the current position, with the penalty weights decreasing as the path distance increases; Additional congestion time is determined based on the penalty weights for each target congestion point and a preset basic congestion penalty; The remaining mission time of the target automated guided vehicle is obtained based on the basic travel time of the planned path, the turning time, and the additional congestion time.

[0170] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the task remaining time prediction method provided in the above embodiments. The method includes: obtaining a planned path of a target automated guided vehicle and a set of congestion points in the current road network, wherein the planned path includes serialized nodes from the current location to the target location; calculating the path distance between each target congestion point and the current location, wherein the target congestion point is a node in the serialized nodes that overlaps with the set of congestion points; obtaining the turning time based on the number of turns on the planned path, wherein the number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes; determining the penalty weight corresponding to each target congestion point based on the path distance between each target congestion point and the current location, wherein the penalty weight decreases as the path distance increases; determining the additional congestion time based on the penalty weight of each target congestion point and a preset basic congestion penalty; and obtaining the remaining task time of the target automated guided vehicle based on the basic travel time of the planned path, the turning time, and the additional congestion time.

[0171] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0172] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the remaining duration of a task, characterized in that, Applications include multi-autonomous guided vehicle (MAV) logistics systems, including: Obtain the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location. Calculate the path distance between each target congestion point and the current location, where the target congestion point is a node in the serialized nodes that overlaps with the set of congestion points; The turning time is obtained based on the number of turns on the planned path. The number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes. Based on the path distance between each target congestion point and the current location, a penalty weight corresponding to each target congestion point is determined, and the penalty weight decreases as the path distance increases; Based on the penalty weights of each target congestion point and the preset basic congestion penalty, the additional congestion time is determined. The remaining mission duration of the target automated guided vehicle is obtained based on the basic travel time of the planned route, the turning time, and the additional time spent due to congestion.

2. The method for predicting the remaining duration of a task according to claim 1, characterized in that, The step of determining the penalty weight corresponding to each target congestion point based on the path distance between each target congestion point and the current location includes: Obtain a preset congestion forgetting factor, which is used to characterize the spatial decay rate of the impact of congestion on vehicle driving in the road network. Multiply the congestion forgetting factor by the path distance from the target congestion point to the current location to obtain the impact attenuation degree of the target congestion point; Based on the impact attenuation degree, the penalty weight corresponding to the target congestion point is determined so that the penalty weight decreases exponentially with the increase of the path distance.

3. The method for predicting the remaining duration of a task according to claim 1 or 2, characterized in that, The determination of additional congestion time based on the penalty weights of each target congestion point and the preset basic congestion penalty includes: By aggregating the penalty weights of all the target congestion points on the planned path, the overall congestion impact coefficient of the planned path is obtained. Obtain a preset base congestion penalty, which represents the baseline delay time under single-point congestion conditions; The overall congestion impact coefficient is weighted and calculated with the basic congestion penalty to determine the additional congestion time corresponding to the planned path.

4. The method for predicting the remaining duration of a task according to claim 1, characterized in that, The step of obtaining the turning time based on the number of turns on the planned path includes: In the serialization nodes of the planned path, extract at least three consecutive path nodes in sequence; Based on the first two and last two path nodes of the three path nodes, the entry direction vector and exit direction vector of the target automated guided vehicle are determined respectively. Obtain the steering angle between the inbound direction vector and the outbound direction vector; If the steering angle is within a preset turning determination range, the middle path node among the three path nodes is marked as the turning position, so as to count the number of path turns. The number of turns on the path is correlated with the preset baseline time for a single turn to obtain the turning time corresponding to the planned path.

5. The method for predicting the remaining duration of a task according to claim 1, characterized in that, After determining the additional congestion time based on the penalty weights for each target congestion point and the preset basic congestion penalty, the following steps are also included: Determine whether the current location is a target congestion point in the set of congestion points; If the current location is the target congestion point, the additional congestion time is updated to a preset constant penalty value, which is greater than the threshold of additional congestion time accumulated along the planned path. Based on the updated additional congestion time, the scheduling system of the multi-autonomous vehicle logistics system is triggered to implement a task avoidance strategy for the target automated guided vehicle.

6. The method for predicting the remaining duration of a task according to claim 1, characterized in that, Before obtaining the planned route of the target automated guided vehicle and the set of congested points in the current road network, the method further includes: Obtain the number of vehicles on each road segment element in the current road network, and the smooth driving speed of the vehicles on each road segment element; If the number of vehicles corresponding to the road segment element is greater than or equal to a preset first threshold, and the average smooth driving speed is less than a preset second threshold, then the location of the vehicle that meets the preset low speed condition on the road segment element is determined as a multi-vehicle congestion point. If it is detected that the dwell time of a single vehicle at a non-permitted long-term dwell node exceeds a preset third threshold, the location of the single vehicle will be determined as an abnormal congestion point for the single vehicle. The congestion points are aggregated from the multi-vehicle congestion points and the single-vehicle abnormal congestion points to generate the congestion point set.

7. The method for predicting the remaining duration of a task according to claim 6, characterized in that, Obtaining the smooth driving speed of vehicles on each of the aforementioned road segment primitives in the current road network includes: Extract multiple frames of historical state data for each vehicle within a preset sliding time window, wherein the historical state data includes timestamps and location coordinates; Based on the timestamp difference and position coordinate difference between adjacent frames in the multi-frame historical state data, the single-step time difference and the single-step displacement difference based on Manhattan distance are calculated for each vehicle. The cumulative time is obtained by summing the time differences of each step, and the cumulative displacement is obtained by summing the displacement differences of each step. Based on the cumulative displacement and the cumulative time, the smooth driving speed of each vehicle is determined.

8. The method for predicting the remaining duration of a task according to claim 6, characterized in that, Before obtaining the number of vehicles on each road segment primitive in the current road network, the process also includes: Obtain the map topology data of the current road network and construct an undirected graph with path nodes as vertices and node connections as the connection relationships; Count the number of connecting lines between each path node in the undirected graph to determine the degree of each path node; Identify path nodes in the undirected graph whose degree is not equal to 2, and mark the identified path nodes as branch nodes; Along the path node link with a node degree of 2, path tracing is performed between two adjacent branch nodes, and the resulting continuous non-branching path segments are determined as each of the path segment primitives.

9. The method for predicting the remaining duration of a task according to claim 1, characterized in that, After obtaining the remaining mission duration of the target automated guided vehicle, the process further includes: Get the waiting time of the currently pending tasks and the preset latest assignment time threshold; Determine whether the waiting time has reached the latest allocation time threshold, and determine whether the additional congestion time exceeds the preset time limit threshold; If the waiting time reaches the latest allocation time threshold and the additional congestion time exceeds the time limit threshold, then the additional congestion time is removed from the remaining task time. Based on the remaining time of the task after removing the additional time spent due to congestion, the scheduling system of the multi-autonomous vehicle logistics system is triggered to forcibly assign the currently unassigned task to the target automated guided vehicle.

10. The method for predicting the remaining duration of a task according to claim 1, characterized in that, The base driving time is determined based on the following method: Determine the total path length of the planned path; The current load status of the target automated guided vehicle is obtained, including both empty and fully loaded states. Based on the current load status, determine the corresponding target driving speed from the preset unloaded driving speed and the preset fully loaded driving speed; The base travel time of the planned route is obtained by calculating the ratio of the total path length to the target travel speed.

11. A device for predicting the remaining duration of a task, characterized in that, Applications include multi-autonomous guided vehicle (MAV) logistics systems, including: The data acquisition module is used to acquire the planned path of the target automated guided vehicle and the set of congestion points in the current road network. The planned path includes serialized nodes from the current location to the target location. The distance calculation module is used to calculate the path distance between each target congestion point and the current location, wherein the target congestion point is a node in the serialized node that overlaps with the set of congestion points; The turning time determination module is used to obtain the turning time based on the number of turns on the planned path. The number of turns is determined after identifying the turning positions in the planned path based on the distribution characteristics of the serialized nodes. The weight determination module is used to determine the penalty weight corresponding to each of the target congestion points based on the path distance between each target congestion point and the current position, wherein the penalty weight decreases as the path distance increases; The additional time determination module is used to determine the additional congestion time based on the penalty weight of each target congestion point and the preset basic congestion penalty; The duration prediction module is used to obtain the remaining mission duration of the target automated guided vehicle based on the basic travel time of the planned route, the turning time, and the additional time spent due to congestion.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the task remaining duration prediction method as described in any one of claims 1 to 10.

13. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the task remaining duration prediction method as described in any one of claims 1 to 10.

14. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the task remaining duration prediction method as described in any one of claims 1 to 10.

Images