A method and system for queue management of a mobile robot dense parking area

Abstract

The application discloses a kind of mobile robot dense parking area queue management method and system, the method includes: parsing scene map, extract multiple parking points and the path connection between each parking point;Based on the direction of path and the in-degree and out-degree of each parking point, automatically identify the queue structure of dense parking area, and establish the topological model including the head position, tail point and point index;For the mobile robot to be parked, allocate parking point;After robot parking, detect whether there is a vacancy in front of the queue where it is, if there is, calculate the push benefit, when the benefit exceeds the preset threshold, generate push task, control robot to move to the head direction, make the head position be occupied preferentially.

Description

Technical Field

[0001] This invention relates to the field of mobile robot scheduling technology, and in particular to a queue management method and system for densely packed mobile robot docking areas. Background Technology

[0002] In automated operations such as smart warehousing and flexible manufacturing, automated guided vehicles (AGVs / AMRs) typically perform material handling tasks in swarms under the unified control of a scheduling system. To shorten the task response chain, mobile robots that have completed their tasks are generally not returned to fixed parking areas, but instead enter the nearest docking area to wait for further instructions, enabling them to quickly reach the work site upon receiving new task commands. Due to space constraints, docking areas generally adopt a dense layout, such as parking lot-style aisle structures or deep charging station-style layouts, with multiple docking points arranged closely along the aisles, their relative positions forming a specific topology. In such dense docking areas, the docking points are spatially constrained: if a docking point located on the inner side of the aisle is occupied, the entry and exit of robots on the outer docking points will be restricted. Therefore, the selection of docking locations and the maintenance of the docking order directly affect the scheduling efficiency of the robot swarm.

[0003] Currently, the dispatching system manages the stopping area primarily through the following methods:

[0004] One method is decentralized docking, where mobile robots randomly select docking locations from available spots. This method is simple to implement, but the robots are distributed haphazardly, the utilization rate of docking points is low, the number of standby robots in different areas is uneven, and some work areas cannot receive timely responses.

[0005] Second, the nearest docking point is selected, where the mobile robot chooses the nearest available docking point to stop. While this method can reduce the travel distance for a single stop, the determination of the docking location does not take into account the topological constraints between docking points, and the docking order cannot be guaranteed. In densely populated docking areas, it is easy for robots that arrive earlier to be blocked inside the alley by robots that arrive later.

[0006] Thirdly, there is the pre-dock mechanism, such as the cargo movement scheme disclosed in Chinese patent CN109885041A, which achieves the pre-layout of robots by pre-setting docking points and docking groups. This type of scheme pre-writes the queue structure of the docking area (including topological information such as entrance, exit, and docking point order) as a static configuration item into the scheduling system, and allocates docking points according to the predetermined configuration during scheduling. Its drawbacks are: the configuration of the queue structure depends on manual completion, the configuration process is cumbersome and prone to errors; when the scene changes (such as map modification, addition or removal of docking points), the configured queue structure becomes invalid and needs to be manually reconfigured, and cannot automatically adapt to scene changes; at the same time, this type of scheme only solves the static allocation problem of robots, and the queue state is no longer maintained after the robot has docked, and when the robot at the front of the queue takes an order and leaves, leaving an empty space, the robot behind will not automatically fill the space.

[0007] However, existing technologies still face the following insurmountable problems in managing densely populated parking areas:

[0008] (1) The parking order cannot be maintained. When multiple mobile robots arrive at the dense parking area one after another, their parking positions are randomly determined by the idle state at the time of arrival. If the robot that arrives later occupies the outer position of the alley while the robot that arrives earlier is stuck on the inner side of the alley, the robot on the inner side must wait for the robot on the outer side to leave first when it receives a task, resulting in additional vehicle movement and waiting time. The disorder of the parking order directly weakens the rapid response capability pursued by the dense layout.

[0009] (2) Empty spaces at the front of the queue cannot be automatically filled, resulting in a waste of space resources. After a robot takes an order and leaves, the empty parking space remains idle for a long time. The robots behind the queue do not actively fill the empty spaces, resulting in a high vacancy rate at the front of the queue—that is, the parking position closest to the work point and with the highest response value. The overall utilization rate and unit area capacity of the dense parking area are both limited.

[0010] (3) The acquisition of queue structure depends on manual configuration and lacks scene adaptation capability. The existing scheduling system itself does not have the ability to identify the topology of the docking area. Key information such as the queue entrance, exit and docking point order must be manually surveyed and entered. In actual warehousing scenarios with frequent business adjustments and frequent map updates, the cost of manual configuration and maintenance is high and the error rate is high. The configuration lag will also cause the scheduling system to allocate based on the wrong queue structure, causing scheduling anomalies.

[0011] Therefore, how to enable the scheduling system to automatically identify the queue structure in densely packed docking areas and automatically maintain the compactness and orderliness of the queue during robot docking is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0012] Therefore, the main objective of this invention is to provide a queue management method and system for densely packed mobile robot docking areas, so that the scheduling system can automatically identify the queue structure in densely packed mobile robot docking areas without the need for manual configuration of the queue topology, and automatically maintain the compactness and orderliness of the queue during robot docking. This solves the technical problems of chaotic docking order, inability to automatically fill empty spaces at the front of the queue, and complex queue structure configuration that cannot adapt to scene changes in the prior art.

[0013] To achieve the above objectives, according to one aspect of the present invention, a queue management method for densely packed docking areas of mobile robots is provided, the steps of which include:

[0014] Analyze the scene map and extract multiple stops in the scene map and the path connection relationships between each stop;

[0015] Based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stop point, the queue structure of the dense stop area is automatically identified from the multiple stop points, and a topology model of the queue structure is established. The topology model includes at least the head point, the tail point, and the point index of each stop point.

[0016] Assign a docking point in the queue structure to the mobile robot to be docked, and control the mobile robot to move to the assigned docking point;

[0017] After the mobile robot stops, it checks whether there is an empty spot in the queue structure in front of it. If there is an empty spot, it calculates the forward push benefit of the mobile robot to the empty spot. When the forward push benefit exceeds a preset forward push benefit threshold, it generates a forward push task and controls the mobile robot to move towards the head of the queue to the empty spot, so that the stopping spot at the head of the queue is occupied first.

[0018] In a possible preferred embodiment, the step of identifying the queue structure includes:

[0019] When a group of stops connected by a path has a unique entry point with an in-degree of 0 and a unique tail point with an out-degree of 0, and all other stops have an in-degree and out-degree of 1, and all paths within the group of stops are bidirectional paths, the group of stops is identified as a stacked queue, and the entry point is the head of the queue.

[0020] When a group of stops connected by a path has a unique head stop with an out-degree of 0 and a unique tail stop with an in-degree of 0, and all other stops have an in-degree and out-degree of 1, and all paths within the group of stops are unidirectional, the group of stops is identified as a queue.

[0021] The in-degree and out-degree are calculated based on a subgraph consisting of the group of stops and the paths between the group of stops; for bidirectional paths, the in-degree and out-degree are calculated after orienting the path from the entry point to the interior of the group of stops.

[0022] In a possible preferred embodiment, the calculation step of the forward gain includes:

[0023] calculate:

[0024]

[0025] Where r is the mobile robot, This is the current docking point of the mobile robot. The parking point corresponding to the empty space; The mobile robot is moved to the target work point by pushing forward and backward. The amount of shortening of the distance, ; This is the amount by which the waiting time is reduced by pushing forward. The cost of moving forward; These are the preset weighting coefficients.

[0026] In a possible preferred embodiment, the generation of the forward task also satisfies the following condition:

[0027] The docking point corresponding to the empty space is in an idle state, and the path between the mobile robot's current docking point and the docking point corresponding to the empty space is reachable.

[0028] In a possible preferred embodiment, the steps of the queue management method further include:

[0029] Record the time when each mobile robot last performed the forward task;

[0030] The mobile robot is only allowed to generate a forward push task if the difference between the current time and the time of the most recent execution of the forward push task is greater than the preset minimum forward push interval, in order to avoid system jitter caused by frequent forward pushes.

[0031] In a possible preferred embodiment, the step of allocating docking points in the queue structure to the mobile robot to be docked includes:

[0032] Calculate the path cost from each mobile robot to its available docking point;

[0033] A greedy algorithm is used to assign the lowest path cost among the available docking points to each mobile robot waiting to dock, in order of path cost from low to high. When there are multiple available docking points with the same path cost, the docking point closer to the end of the queue is assigned first.

[0034] In a possible preferred embodiment, when multiple queue structures are automatically identified, before assigning a docking point from the queue structure to the mobile robot to be docked, the method further includes:

[0035] Calculate the selection score for each queue using the following formula:

[0036]

[0037] in, For the i-th dense parking area, Let its queue length be... This indicates whether the front of the team is available. The number of its available docking points. The total number of its stops. The preset team leader reward coefficient;

[0038] The densely packed docking area with the highest selection score is selected as the target docking area for the mobile robot to be docked.

[0039] In a possible preferred embodiment, the queue management method further includes the following steps:

[0040] Mobile robots are grouped according to their robot group identifier and the area identifier of their location, so that mobile robots from different robot groups can be parked in their corresponding dense parking areas, thus isolating resources of mobile robots across groups.

[0041] In a possible preferred embodiment, the queue management method further includes the following steps:

[0042] When the scene map changes, the changed scene map is re-parsed, and the automatic identification of the dense parking area and the establishment of the topology model are re-executed to adapt to the scene change.

[0043] To achieve the above objectives, corresponding to the above method, according to another aspect of the present invention, a queue management system for densely packed mobile robot docking areas is also provided, comprising:

[0044] The map parsing module is used to parse the scene map and extract multiple stops in the scene map and the path connection relationships between each stop.

[0045] The queue identification module is used to automatically identify the queue structure of dense stopping areas from the multiple stopping points based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stopping point, and to establish a topology model of the queue structure. The topology model includes at least the head point, the tail point, and the point index of each stopping point. The module also determines the queue type of the identified dense stopping area, which includes stacked queues and queued queues.

[0046] The docking allocation module is used to allocate docking points in the identified dense docking areas to the mobile robot to be docked, and to control the mobile robot to move to the allocated docking point.

[0047] The forward control module is used to detect whether there is an empty spot in the queue structure in front of the mobile robot after the mobile robot stops; if there is an empty spot, it calculates the forward push benefit of the mobile robot to the empty spot; when the forward push benefit exceeds the preset forward push benefit threshold, it generates a forward push task and controls the mobile robot to move towards the head of the queue to the empty spot, so that the stopping spot at the head of the queue is occupied first.

[0048] The queue management method and system for densely packed docking areas provided by this invention cleverly decomposes queue management in densely packed docking areas into two levels: "automatic recognition of queue structure" and "forward maintenance of queue status." First, it no longer relies on manual configuration of queue topology. Instead, by analyzing the directionality of paths between docking points in the scene map and the in-degree and out-degree of each docking point, and using a combination of graph theory's degree property and edge directionality criteria, the system autonomously identifies the queue structure of densely packed docking areas and establishes a topological model including the queue head, queue tail, and point indexes. This frees queue management from dependence on manual configuration and enables automatic re-identification as the scene map changes. Simultaneously, based on this, a benefit-driven decision-making approach maintains the compactness of the queue. When an empty space is detected in front of the robot, the benefit of pushing forward is quantified and compared with a preset threshold. The forward pushing task is triggered only when it is profitable, driving the robot to move towards the queue head. This ensures that the queue head position is occupied preferentially in a controlled manner, avoiding ineffective movements.

[0049] This design gives the present invention at least the following advantages over the prior art:

[0050] (1) No manual configuration of queue structure is required, greatly reducing the configuration workload and adapting to scene changes. The system automatically identifies the queue structure of dense parking areas and establishes a topology model, eliminating the tedious process of manually configuring the entrance, exit and point order, and avoiding configuration errors; the system can automatically re-identify after scene map adjustment, which significantly improves the system's adaptability to scene changes.

[0051] (2) It ensures the orderly parking sequence and avoids blocking the robot at the front of the queue. Through the forward filling mechanism, the robots that arrive later automatically move towards the front of the queue to fill the gap. The robots that arrive first and stop first are always at the front of the queue. The robot at the front of the queue can drive away directly after receiving an order, avoiding unnecessary movement and waiting caused by the robot at the back of the queue blocking the robot at the front of the queue in the existing technology.

[0052] (3) The utilization rate of docking points is significantly improved and the waste of space resources is reduced. When there is an empty space in front of the queue, the robot automatically moves forward to fill it. The head of the queue is no longer left empty for a long time. The overall utilization rate of docking points in the queue is significantly improved and the number of robots that can be accommodated per unit area increases.

[0053] (4) Faster order response speed and increased system throughput. Since the head position is occupied first, the robot that receives the task drives directly away from the head of the queue, the average order receiving distance is shortened, the task response time is reduced, and the overall system throughput is increased accordingly.

[0054] (5) The forward push decision is based on the quantification of benefits, taking into account both the benefits of forward push and the cost of movement. The forward push task is triggered when the forward push benefits exceed a preset threshold, which avoids ineffective movement and energy waste caused by blind forward push, so that the forward push mechanism can improve the compactness of the queue while maintaining the economy and stability of the system operation. Attached Figure Description

[0055] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0056] Figure 1 This is a schematic diagram illustrating the steps of the queue management method for densely packed docking areas of mobile robots according to the present invention;

[0057] Figure 2 This is a schematic diagram of a stacked queue topology (bidirectional connection) in the method of the present invention, wherein the bidirectional arrows indicate that the robot can move in both directions, and the black dots indicate the current docking position of the robot;

[0058] Figure 3 This is a schematic diagram of a queue-type queue topology (one-way connection) in the method of the present invention, wherein the one-way arrow indicates the unidirectional flow of the robot (first-in, first-out), and the black dot indicates the current docking position of the robot;

[0059] Figure 4 This is a schematic diagram of the multi-queue selection decision process in the method of the present invention;

[0060] Figure 5 This is a conceptual diagram illustrating the forward filling process in the method of the present invention;

[0061] Figure 6 This is a schematic diagram of the queue management system for densely packed mobile robot docking areas according to the present invention. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions of the present invention, the specific technical solutions of the present invention will be clearly and completely described below in conjunction with embodiments, so as to help those skilled in the art further understand the present invention. Obviously, the embodiments described in this application are merely some embodiments of the present invention, and not all embodiments. It should be noted that, for those skilled in the art, the embodiments and features in the embodiments of this application can be combined with each other without departing from the concept of the present invention and without conflict. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the disclosure and protection scope of the present invention.

[0063] Furthermore, the terms "first," "second," "S1," "S2," etc., used in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such features can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those described herein. At the same time, the stages described in each step are not necessarily to be implemented in the same step; it should be understood that the implementation order of the contents of each step stage can be adjusted and interchanged without violating the inventive concept, so that embodiments of the invention described herein can be implemented in orders other than those described herein. Additionally, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. Unless otherwise expressly specified and limited, the terms "set," "arrange," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to 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 case based on the specific circumstances and in conjunction with existing technology.

[0064] To enable the scheduling system to automatically identify the queue structure in densely populated areas of mobile robots without requiring manual configuration of the queue topology, and to automatically maintain the compactness and orderliness of the queues during robot docking. For example... Figures 1 to 5 As shown, this embodiment provides a queue management method for densely packed mobile robot docking areas, which is applied to a scheduling system that uniformly manages mobile robot clusters (AGVs / AMRs) in a work scenario. The overall process of this method is as follows: Figure 1 As shown, the example includes the following steps:

[0065] Step S1: Scene Map Analysis Step

[0066] The example process for this step is as follows: parse the scene map, extract multiple docking points in the scene map and the path connections between each docking point.

[0067] Specifically, in this example, the scene map pre-marks the locations of various stations and the path information between them. Examples of station types include: Park Points (PP points), used for mobile robots to dock and wait; Charge Points (CP points), used for charging mobile robots; and Landmark Points (LM points), which are ordinary service stations. In this embodiment, the scheduling system parses the scene map during scene initialization (map loading), extracts all parking points to form a parking point set P, and extracts the paths between parking points to form an edge set E, thereby abstracting the parking area into a directed graph in graph theory.

[0068] It should be noted that for bidirectional paths, when constructing the edge set E, it can be represented as two unidirectional edges in opposite directions to unify the calculation of in-degree and out-degree in the future.

[0069] This step provides a graph structure data foundation for the subsequent automatic identification of queue structures. Since all the information required for identification (location, path, direction) is already included in the scene map's own data, unlike existing pre-docking schemes, this invention does not require manual input of any additional queue topology information.

[0070] Step S2: Automatic identification and topology modeling of queue structure

[0071] The example process for this step is as follows: Based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stop point, the queue structure of the dense stop area is automatically identified from multiple stop points, and a topology model of the queue structure is established. The topology model includes at least the head point, the tail point, and the point index of each stop point.

[0072] Specifically, as described in the background section, existing scheduling systems lack the ability to recognize the topological structure of docking areas. The entry, exit, and point order of queues all require manual configuration, which is cumbersome, error-prone, and must be reconfigured after scene changes. Therefore, this invention considers that although the queue structures in dense docking areas vary in form, their topological characteristics can be uniquely characterized by a combination of "vertex degree characteristics + edge directionality" in graph theory. That is, stacked docking areas and queued docking areas have distinct and unique combinations of in-degree, out-degree distribution, and path directionality. Therefore, the system can automatically distinguish them from map data, replacing "manual configuration" with "automatic recognition."

[0073] Automatic structure identification steps

[0074] In this example, based on the directionality of each path and the in-degree and out-degree of each stop, the queue structure of dense stopping areas is automatically identified from multiple stops, mainly including the following two scenarios:

[0075] Scenario 1: Identification of stacked queues

[0076] A set of stops connected by a path is identified as a stacked queue when it meets the following criteria:

[0077]

[0078] in: This represents the in-degree of point p; This is the unique entry point for the queue; Let be the set of edges inside the queue; Used to determine edges Whether it is a bidirectional edge. That is, there exists a unique entry point with an in-degree of 0, and all other docking points have an in-degree of 1, and all paths within the queue are bidirectional paths. In a stacked queue, all docking points are connected by bidirectional paths, with only the entry side connected to the outside of the queue, suitable for scenarios such as charging stations and temporary storage areas. It should be noted that in this invention, stacked and queued are merely names for two types of topological structures used to distinguish path directional characteristics, and do not limit the entry and exit order of the mobile robot.

[0079] Scenario 2: Identification of queue-type queues

[0080] A group of stops connected by a path is identified as a queue when it meets the following criteria:

[0081]

[0082] in: This represents the out-degree of point p; This is the only team's first position (exit); This is the only point at the back of the queue (entrance). Used to determine edges Is it a unidirectional edge? That is, there exists a unique head point with an out-degree of 0 and a unique tail point with an in-degree of 0, all intermediate stops have an in-degree and out-degree of 1, and all paths within the queue are unidirectional. In a queue-like queue, all stops are connected by unidirectional paths, with the tail being the entrance and the head being the exit, suitable for scenarios such as production line delivery and picking aisles.

[0083] It should be noted that the in-degree and out-degree values ​​mentioned above are calculated based on the subgraph formed by each stop point and the paths between them, excluding the edges between stops and external stations (such as LM points). In other words, the identification criterion describes the internal topology of the stop point cluster and is unrelated to external connections. For stacked queues, all paths within the subgraph are bidirectional. In this case, the in-degree is calculated by orienting each path from the entry point to the depth of the queue: the entry point has no upstream points from within the queue, so its in-degree is 0; all other points have only one adjacent point close to the entry point, so their in-degree is 1.

[0084] The following are examples of identification.

[0085] Example 1 (Stacked): In a certain warehouse scenario, there is a stacked docking area, such as... Figure 2 As shown, the map contains six stops, PP1 to PP6, connected by a bidirectional path. PP1 is the sole entry point and is connected to an external LM point. The system parses the map to obtain the set of stops. Given an edge set E (a bidirectional edge is represented as two unidirectional edges), calculate the in-degree and out-degree of each point: PP1 in-degree = 0 (queue entrance), out-degree = 1; PP2~PP6 in-degree = 1, out-degree = 1 (or 0). After verification, the above stacked queue criteria are met, and the identification result is a stacked queue. .

[0086] Example 2 (Queue-style): In a certain production line delivery scenario, there is a queue-style stopping area, such as... Figure 3 As shown, there are 5 stops, PP51 to PP55, connected by a one-way path. The in-degree and out-degree are calculated as follows: PP55 in-degree = 0, out-degree = 1 (tail of the queue); PP51 in-degree = 1, out-degree = 0 (head of the queue); the in-degree and out-degree of the remaining stops are both 1. After verification, the queue meets the above criteria, and the queue is identified as a queue. .

[0087] Model building steps

[0088] After identification, the system establishes a topology model for each queue, including at least the head of the queue, the tail of the queue, intermediate nodes, and the indexes of each stopping point. Taking Example 1 above as an example, the established topology model is as follows:

[0089] { "queue_id": "Q_stack_001",

[0090] "type": "stack",

[0091] "points": [PP1, PP2, PP3, PP4, PP5, PP6],

[0092] "entry": PP1,

[0093] "head_index": 0,

[0094] "tail_index": 5}

[0095] This topology model provides a structured basis for subsequent docking point allocation and forward filling: the head point determines the forward direction and the priority target position, the tail point determines the priority direction for allocation of new robots, and the point index determines the forward order.

[0096] Furthermore, in an optional implementation, when the scene map changes (such as adding or removing stops, or adjusting the route), the system can re-parse the changed scene map and re-execute the aforementioned automatic recognition and topology model establishment process, adapting to scene changes without any manual intervention. In one implementation, queue recognition is a one-time calculation completed during map loading, typically taking less than 1 second.

[0097] Through this design, the system can automatically identify stacked and queued docking areas without manual configuration, eliminating the workload of manually configuring queue topology (verified in actual applications, this can reduce configuration workload by more than 90% and reduce configuration error rate by about 80%), and supports automatic re-identification after dynamic scene adjustments, fundamentally solving the technical problem that existing queue structure configurations are complex and cannot adapt to scene changes.

[0098] Step S3, Stop Point Allocation Steps

[0099] The example process for this step is as follows: assign a docking point in the queue structure to the mobile robot to be docked, and control the mobile robot to move to the assigned docking point.

[0100] Specifically, the scheduling system first filters mobile robots that need to dock (e.g., robots that are idle, non-faulty, or not performing tasks), calculates the path cost C(R, PP) from each robot to available docking points (path cost can be represented by travel time), and then uses a greedy algorithm to allocate docking points: robots are allocated the lowest path cost among currently available docking points in ascending order of path cost; when multiple available docking points have the same path cost, the docking point closer to the end of the queue (entrance) is prioritized. It should be noted that in this embodiment, when a docking point is occupied by a mobile robot, it does not block the corresponding path. Other mobile robots can move along the path from the occupied docking point to an available docking point within the queue. Therefore, even after the docking point near the head of the queue is prioritized for occupation, subsequent mobile robots can still enter the queue to dock.

[0101] For queue-type queues, since the only entrance is located at the rear of the queue, after a new robot enters the queue from the rear, it gradually becomes more compact towards the front of the queue through the forward pushing mechanism in step S4. Therefore, the priority allocation tendency of the rear of the queue is naturally realized by the unidirectional entry structure of the queue in the queue-type scenario.

[0102] The design intention of prioritizing the rear position of the queue is that: when a new robot enters the queue from the rear, it does not hinder the exit of robots already parked at the front of the queue, and leaves complete queue space for subsequent push-forward filling. Together with the push-forward mechanism, it maintains the parking order in which the first to arrive occupies the front of the queue.

[0103] It should be noted that in this embodiment, when each docking point is occupied by a mobile robot, it does not block the passage of the corresponding path. Other mobile robots can pass through the occupied docking points and move along the path to an empty docking point inside the queue.

[0104] Example: Continuing from Example 1 above, suppose there are 3 robots R1, R2 and R3 that need to dock, and the current queue is empty.

[0105] Calculate path cost (unit: seconds):

[0106] C(R1, PP1)=10, C(R1, PP2)=15, ..., C(R1, PP6)=35

[0107] C(R2, PP1)=8, C(R2, PP2)=13, ..., C(R2, PP6)=33

[0108] C(R3, PP1)=12, C(R3, PP2)=17, ..., C(R3, PP6)=37

[0109] Greedy allocation (sorted by cost, prioritizing allocation to the tail of the queue):

[0110] R2→PP1 (cost 8, lowest);

[0111] R1→PP2 (PP1 is already in use, select PP2);

[0112] R3→PP3 (PP1 and PP2 are already in use).

[0113]

[0114] Furthermore, when multiple queue structures are automatically identified in the scene, such as Figure 4 As shown, a queue selection step is included before assigning docking points: An example is shown where the selection score for each queue is calculated using the following formula:

[0115]

[0116] in, For the i-th queue; The length is the queue length, measured by the number of stops within the queue. In this embodiment, the queues are arranged in a linear series. and With the same value, in other implementations, the queue length can also be measured by physical distance to adapt to scenarios with uneven spacing between stopping points; Indicates whether the head of the queue is free (1 for free, 0 for occupied). This represents the number of available docking stations. This represents the total number of stops. This is the preset reward coefficient for the first queue (default value is 2.0, range is 0.0~10.0). The system selects the queue with the highest score as the robot's target docking area.

[0117] The design principle of this scoring function is as follows: the shorter the queue length, the shorter the forward distance of the robot after entering the queue; an empty head of the queue means that the robot has the opportunity to directly push forward to the position with the highest response value, so the weight of this item is amplified by the head of the queue reward coefficient λ; the higher the proportion of empty positions, the stronger the queue's capacity.

[0118] Example: If there are 3 docking queues in the scenario, Q1 has a length of 4, occupies 3 queues, and the head of the queue is free; Q2 has a length of 6, occupies 2 queues, and the head of the queue is occupied; Q3 has a length of 3 and is full. Calculate the following for each queue:

[0119] Queue Q1:

[0120] Queue Q2:

[0121] Queue Q3:

[0122] Decision result: Choose Q1. Since the front of the queue is empty, robot R will move directly to the front of the queue.

[0123] Furthermore, it should be noted that when there is only one mobile robot waiting to dock (e.g., when robots arrive at the queue-style docking area sequentially), the above greedy allocation degenerates into direct allocation: Continuing from Example 2 above, robot R1 arrives at the queue-style docking area... When the robot arrives, the system directly assigns it to the rear entry point PP55. R1 then moves towards the front of the queue according to the forward-pushing mechanism in step S4 below. That is, regardless of whether the robots arrive in batches or one by one, the new robots enter the queue from the rear, and the order within the queue is maintained by the forward-pushing filling mechanism.

[0124] Furthermore, in an optional implementation, in complex scenarios with multiple maps and multiple robot groups, the scheduling system can group and manage each mobile robot according to its robot group identifier (groupId) and its area identifier (areaId), allowing robots from different robot groups to dock in their respective densely populated docking areas, thus achieving resource isolation and orderly docking across robot groups. In addition, this embodiment supports mixed docking management of robots of different models (sizes, shapes).

[0125] Step S4, Forward Filling Mechanism Step

[0126] The example process for this step is as follows: After the mobile robot stops, it checks whether there is an empty spot in the queue structure in front of the mobile robot. If there is an empty spot, it calculates the forward push benefit of the mobile robot to the empty spot. When the forward push benefit exceeds the preset forward push benefit threshold, it generates a forward push task and controls the mobile robot to move towards the head of the queue to the empty spot so that the parking spot at the head of the queue is occupied first.

[0127] Specifically, this step primarily addresses the problem of empty spaces at the front of the queue failing to be automatically filled due to disordered order. The inventive concept lies in the fact that after a robot docks, the queue state is not static but continuously maintained in a profit-driven manner. That is, whenever an empty space appears in the queue, the system quantitatively evaluates the gains and losses of pushing the robot forward behind that empty space, triggering a push only when it is profitable. Thus, the queue can spontaneously maintain a compact position without manual intervention, with the head position (the closest to the work point and the position with the highest response value) always being prioritized for occupation; simultaneously, a threshold is used to filter out ineffective pushes where the benefits do not outweigh the costs, avoiding wasted movement.

[0128] Specifically, when an empty stop is detected in the queue ahead of the mobile robot, the forward push gain of the robot towards the empty stop is calculated according to the following formula:

[0129]

[0130] in:

[0131] r represents a mobile robot;

[0132] This is the robot's current docking point;

[0133] The parking spot corresponding to the empty space;

[0134] Δd represents the distance the robot travels from the forward-pushing robot to the target work point. The amount of shortening of the distance,

[0135] ;

[0136] The cost of moving forward;

[0137] α, β, and γ are preset weighting coefficients. In the example, the default values ​​are α=1.0, β=0.5, and γ=0.3.

[0138] Δt represents the reduction in waiting time after the push-forward action. Δt is the reduction in the expected waiting time of the mobile robot after the push-forward action. The expected waiting time can be estimated based on the number of occupied docking points in front of the mobile robot and the expected departure time of each preceding mobile robot. For example, for every occupied docking point removed in front of the robot after the push-forward action, the expected waiting time is reduced by the average departure time of the preceding robot. In a simplified implementation, Δt is set to 0, meaning the push-forward decision is made solely based on the distance reduction and movement cost (as shown in the calculation examples below).

[0139] In the revenue function, Δd is the main revenue term. The fundamental purpose of forward pushing is to bring the robot closer to the future work point, thereby shortening the response distance after receiving the order. This is a cost penalty term used to mitigate insufficient revenue forwarding. In one implementation, the target operation point... The determination method is as follows: when there are already assigned or predicted tasks in the scheduling system, the job point of that task is taken as... When the robot is in an idle standby state and its future mission is unknown, a representative target point in the area is taken as... For example, the geometric center of the historical mission points in the area, high-frequency mission points, or the LM point connected to the alley exit.

[0140] The generation of the forward task also needs to meet the following conditions, namely, the forward trigger function is:

[0141]

[0142] in:

[0143] This is the threshold for forward earnings (0.5 by default in the example, with a range of 0.0 to 10.0).

[0144] Used to determine stopping points Is it empty?

[0145] Used to determine from arrive Is it reachable?

[0146] In other words, a forward push task is generated only when all three conditions are met simultaneously: the profit exceeds a threshold, the target point is available, and the path is reachable. This forward push decision is calculated in real time, typically taking less than 10ms per iteration.

[0147] The following is for reference Figure 5 As shown, an example illustrates the forward push example.

[0148] Example (Stack-based): Continuing from the allocation result of Example 1 above (R2→PP1, R1→PP2, R3→PP3). When R2 leaves PP1, the system detects that PP1 is idle, as follows:

[0149]

[0150] Now calculate the benefit of moving R1 from PP2 to PP1: distance reduction. The amount of time reduction Mobile costs ,but:

[0151]

[0152] If the triggering condition is met, generate the forward push task of R1 from PP2 to PP1. Post-push state:

[0153]

[0154] After the forward push is completed, continue to check the forward push yield of R3. If the condition is also met, R3 is pushed forward to PP2.

[0155]

[0156] As a result, the queue spontaneously becomes more compact towards the front, and the front PP1 is always occupied first.

[0157] Example (queue type): Continuing from Example 2 above, after robot R1 reaches the tail of the queue PP55, it detects an empty space in front and calculates:

[0158]

[0159] Trigger forward push, R1 moves along a one-way path to the head of the queue PP51;

[0160]

[0161] Then R2, R3, and R4 arrive in sequence and push forward.

[0162]

[0163] When R1 receives a task and leaves PP51, it triggers the queue to move forward again. R2, R3, and R4 move forward in sequence to ensure that the robots that arrive later do not block the departure path of the robots that arrive earlier. The robots that arrive first have priority to accept orders and leave directly.

[0164]

[0165] Furthermore, in an optional implementation, to avoid system jitter caused by frequent forward pushes, a forward push cooling time (i.e., minimum forward push interval) constraint can be introduced: for example, recording the time of each robot's most recent forward push task. The robot's forward task is only allowed to be generated if the difference between the current time and that time is greater than the preset minimum forward interval.

[0166]

[0167] in, The minimum advance interval is 5 seconds (the preferred value range is 1 to 60 seconds).

[0168] After the forward push task is completed, the system checks the empty space in the queue again and repeats the above forward push decision until there is no forward push in the queue that meets the triggering condition, thereby realizing the gradual compaction of the queue towards the head (such as the process of R1 pushing forward and triggering R3 to continue pushing forward in the example above).

[0169] In summary, through the above design and practical application verification, the following results were achieved after adopting the forward-filling mechanism: the vacancy rate at the head of the queue can be reduced from 42% to 12%, the overall queue utilization rate can be increased from 58% to 89%, the number of robots that can be accommodated per unit area can be increased by about 30%, and the effective utilization rate of the docking point can be increased by 40% to 60%; the average order receiving distance can be shortened from 28m to 14m, the task response time can be reduced from 45s to 28s, the overall system throughput can be increased from 850 orders / h to 1120 orders / h, and the task queuing time during peak periods can be reduced by about 40%.

[0170] The configurable parameters involved in the above push-forward mechanism example are shown in Table 1 below: Table 1 Parameter Configuration Table

[0171] pushBenefitThreshold Forward return threshold 0.5 0.0~10.0 pushCooldownTime Pre-cooldown time (seconds) 5 1~60 queueScoreLambda Team leader reward coefficient 2.0 0.0~10.0 safeDistance Safe distance (multiples) 1.5 1.0~3.0

[0172] Furthermore, to meet different scenarios and needs, in the optional implementation, a task prediction module can be added to the above basic solution. This module can be used to statistically analyze the historical task frequency of each region based on historical data and establish a time-series prediction model (such as LSTM or ARIMA). This allows the robot to be moved to the high-demand area in advance before the task arrives. This solution can further shorten the response time.

[0173] Furthermore, to meet different scenarios and needs, in optional implementations, multi-priority queues can be designed. For example, the queue can be divided into two sub-queues: a high-priority queue and a low-priority queue. Priorities are determined based on robot performance (speed, load capacity), allowing high-priority robots to jump to the front of the queue, while low-priority robots wait at the back. This solution can improve the utilization rate of high-performance robots.

[0174] Furthermore, to address different scenarios and needs, the optional implementation can also design dynamic queue length adjustment, such as monitoring task arrival rate and dynamically adjusting the effective queue length based on real-time task load, closing or opening some docking points: shortening the queue to reduce the number of push-forwards when there are few tasks, and extending the queue to increase capacity when there are many tasks. This solution can save energy consumption.

[0175] Furthermore, depending on different scenarios and needs, in optional implementations, cross-queue load balancing can be configured, such as establishing an inter-queue communication mechanism, allowing multiple queues to share queue status information in real time, and dynamically allocating robots across queues when a queue becomes saturated, guiding them to other queues. This solution can improve the overall system resilience.

[0176] On the other hand, corresponding to the above method embodiments, such as Figure 6 As shown, this embodiment also provides a queue management system for densely populated areas of mobile robots, examples of which include:

[0177] The map parsing module is used to parse the scene map and extract multiple stops in the scene map and the path connection relationships between each stop.

[0178] The queue identification module is used to automatically identify the queue structure of dense stopping areas from multiple stopping points based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stopping point, and to establish a topological model of the queue structure. The topological model includes at least the head point, the tail point, and the point index of each stopping point.

[0179] The docking allocation module is used to allocate docking points in the queue structure to mobile robots waiting to be docked, and to control the mobile robots to move to the allocated docking points.

[0180] The forward control module is used to detect whether there is an empty spot in the queue structure in front of the mobile robot after the mobile robot stops. If there is an empty spot, it calculates the forward push benefit of the mobile robot to the empty spot. When the forward push benefit exceeds the preset forward push benefit threshold, it generates a forward push task and controls the mobile robot to move towards the head of the queue to the empty spot so that the stopping spot at the head of the queue is occupied first.

[0181] The specific implementation methods of the above modules correspond to steps S1~S4 and their extended implementation methods in the method embodiment, and will not be repeated here. This system is used to seamlessly integrate with existing robot scheduling systems to mark PP points and CP points in the scene configuration and enable queue docking parameters (AutoQueuePark), thereby automatically identifying and managing queues without modifying existing business logic and task allocation algorithms.

[0182] In summary, the queue management method and system for densely packed docking areas provided by this invention cleverly deconstructs queue management into two interconnected levels: automatic recognition of queue structure and forward maintenance of queue state. The former utilizes a combination of vertex degree and edge directionality criteria from graph theory to enable the system to autonomously identify the queue structure of densely packed docking areas from scene map data and establish a topological model, replacing manual configuration with automatic recognition. The latter quantifies the gains and losses of each forward push using a forward reward function, and uses a threshold judgment to drive the robot to spontaneously fill empty spaces in the queue towards the head. The combination of these two approaches frees the management of densely packed docking areas from dependence on manual configuration and continuously maintains the orderliness and compactness of the queue throughout the robot's docking process.

[0183] Based on this solution, the workload of queue structure configuration can be reduced by more than 90%, and it can automatically re-identify as the scene changes; the docking order is maintained, and robots that arrive first are no longer blocked by robots that arrive later; the vacancy rate at the head of the queue is significantly reduced, and the overall utilization rate of the queue increases from 58% to 89%; the average order receiving distance is shortened by 50%, the task response time is reduced by 38%, and the system throughput is increased by 32%; at the same time, with the constraint of revenue threshold and cooldown time, the economy and stability of the push-forward mechanism can be guaranteed, and it is particularly suitable for mobile robot operation scenarios with dense docking needs, such as e-commerce warehouse picking aisles, manufacturing production line workstation delivery, medical logistics pharmacy delivery, and airport baggage handling.

[0184] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The present invention is limited only by the claims and their full scope and equivalents. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

[0185] Those skilled in the art will understand that, besides implementing the system, apparatus, unit, and its modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and its modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0186] Furthermore, all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0187] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A queue management method for densely packed docking areas of mobile robots, comprising the following steps: Analyze the scene map and extract multiple stops in the scene map and the path connection relationships between each stop; Based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stop point, the queue structure of the dense stop area is automatically identified from the multiple stop points, and a topology model of the queue structure is established. The topology model includes at least the head point, the tail point, and the point index of each stop point. Assign a docking point in the queue structure to the mobile robot to be docked, and control the mobile robot to move to the assigned docking point; After the mobile robot stops, it is checked whether there is an empty spot at the stopping point in front of the mobile robot in the queue structure. If there is an empty space, calculate the forward push benefit of the mobile robot to the empty space. When the forward push benefit exceeds the preset forward push benefit threshold, generate a forward push task and control the mobile robot to move towards the head of the queue to the empty space, so that the docking point at the head of the queue is occupied first.

2. The queue management method according to claim 1, wherein the step of identifying the queue structure includes: When a group of stops connected by a path has a unique entry point with an in-degree of 0 and a unique tail point with an out-degree of 0, and all other stops have an in-degree and out-degree of 1, and all paths within the group of stops are bidirectional paths, the group of stops is identified as a stacked queue, and the entry point is the head of the queue. When a group of stops connected by a path has a unique head stop with an out-degree of 0 and a unique tail stop with an in-degree of 0, and all other stops have an in-degree and out-degree of 1, and all paths within the group of stops are unidirectional, the group of stops is identified as a queue. The in-degree and out-degree are calculated based on a subgraph consisting of the group of stops and the paths between the group of stops; for bidirectional paths, the in-degree and out-degree are calculated after orienting the path from the entry point to the interior of the group of stops.

3. The queue management method according to claim 1, wherein the calculation step of the forward push benefit includes: calculate: ； Where r is the mobile robot, This is the current docking point of the mobile robot. The parking point corresponding to the empty space; The mobile robot is moved to the target work point by pushing forward and backward. The amount of shortening of the distance, ; This is the amount by which the waiting time is reduced by pushing forward. The cost of moving forward; These are the preset weighting coefficients.

4. The queue management method according to claim 1 or 3, wherein the generation of the forward task further satisfies the following condition: The docking point corresponding to the empty space is in an idle state, and the path between the mobile robot's current docking point and the docking point corresponding to the empty space is reachable.

5. The queue management method according to claim 1, further comprising the following steps: Record the time when each mobile robot last performed the forward task; The mobile robot is only allowed to generate a forward push task if the difference between the current time and the time of the most recent execution of the forward push task is greater than the preset minimum forward push interval, in order to avoid system jitter caused by frequent forward pushes.

6. The queue management method according to claim 1, wherein the step of allocating a docking point in the queue structure to a mobile robot to be docked includes: Calculate the path cost from each mobile robot to its available docking point; A greedy algorithm is used to assign the lowest path cost among the available docking points to each mobile robot waiting to dock, in order of path cost from low to high. When there are multiple available docking points with the same path cost, the docking point closer to the end of the queue is assigned first.

7. The queue management method according to claim 1, when multiple queue structures are automatically identified, before assigning a docking point in the queue structure to the mobile robot to be docked, the step further includes: Calculate the selection score for each queue using the following formula: ； in, For the i-th dense parking area, Let its queue length be... This indicates whether the front of the team is available. The number of its available docking points. The total number of its stops. The preset team leader reward coefficient; The densely packed docking area with the highest selection score is selected as the target docking area for the mobile robot to be docked.

8. The queue management method according to claim 1, further comprising the following steps: Mobile robots are grouped according to their robot group identifier and the area identifier of their location, so that mobile robots from different robot groups can be parked in their corresponding dense parking areas, thus isolating resources between mobile robots in different groups.

9. The queue management method according to claim 1, further comprising the following steps: When the scene map changes, the changed scene map is re-parsed, and the automatic identification of the dense parking area and the establishment of the topology model are re-executed to adapt to the scene change.

10. A queue management system for densely populated mobile robot docking areas, comprising: The map parsing module is used to parse the scene map and extract multiple stops in the scene map and the path connection relationships between each stop. The queue identification module is used to automatically identify the queue structure of dense stopping areas from the multiple stopping points based on the directionality of each path in the path connection relationship and the in-degree and out-degree of each stopping point, and to establish a topology model of the queue structure. The topology model includes at least the head point, the tail point, and the point index of each stopping point. The module also determines the queue type of the identified dense stopping area, which includes stacked queues and queued queues. The docking allocation module is used to allocate docking points in the identified dense docking areas to the mobile robot to be docked, and to control the mobile robot to move to the allocated docking point. The forward control module is used to detect whether there is an empty space at the docking point in front of the mobile robot in the queue structure after the mobile robot stops. If there is an empty space, calculate the forward push benefit of the mobile robot to the empty space. When the forward push benefit exceeds a preset forward push benefit threshold, generate a forward push task and control the mobile robot to move towards the head of the queue to the empty space so that the stopping point at the head of the queue is occupied first.

Images