Intention-based multi-robot distributed asynchronous frontier exploration method

Abstract

The present application relates to the technical field of robot autonomous exploration, and in particular to a multi-mobile robot distributed asynchronous frontier exploration method based on intention bidding, which comprises: each robot performs positioning and mapping, extracts a frontier grid cluster as a candidate task, and maintains its own state; after entering an idle state based on an event condition, the robot forms a current decision group with a communication neighbor; the robots in the group are divided into two categories: idle and busy, and the busy robots generate and publish intention bids; the idle robots calculate net benefits by combining intention costs and their immediate benefits, and determine the task executors through local multi-round bidding. The method effectively solves the task conflict and repeated exploration problem without global synchronization and a central node, and significantly improves the stability, coverage efficiency and system scalability of distributed collaborative exploration through event-driven, intention expression and local asynchronous bidding.

Description

Technical Field

[0001] This invention relates to the field of autonomous robot exploration technology, specifically to a distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding. Background Technology

[0002] Collaborative exploration by multiple mobile robots is an important research direction in the field of robotics, aiming to efficiently build environmental maps in unknown environments through collaboration among multiple mobile robots. A key challenge in this process lies in how to assign different exploration tasks to each robot in real time to minimize repetitive work and improve overall coverage efficiency.

[0003] To address this challenge, existing technical solutions primarily follow this approach: first, employing a frontier-based exploration strategy to transform the continuous spatial exploration problem into a selection problem of a limited number of candidate frontier points, providing clear targets for subsequent task allocation; then, introducing a specialized coordination mechanism to allocate suitable robots to these candidate frontier points. For example, patents CN117168437A, CN114137955A, and CN114859375A all use a market-bid-based coordination mechanism to allocate frontier targets. However, such coordination mechanisms have the following limitations: First, they assume excessive synchronicity, assuming all participating robots are in a state of synchronous negotiation, failing to fully consider the inherent asynchronicity of robots due to differences in task execution time; second, they lack the ability to express intent, as allocation decisions are based solely on the current preferences of participating robots, completely ignoring the potential future coverage intentions of robots currently performing tasks towards unallocated frontiers. In a dynamically changing frontier environment, these deficiencies directly lead to multi-robot task conflicts, redundant exploration, and system instability.

[0004] Furthermore, analysis reveals that the shortcomings of the aforementioned coordination mechanisms are difficult to effectively address under different system architectures. In a centralized architecture, although global coordination can theoretically be achieved by a central node, its strong dependence on the central node and global communication leads to problems such as high communication overhead and poor robustness. In a distributed architecture, while theoretically more adaptable to dynamic environments, the synchronous negotiation rules typically employed in existing distributed coordination schemes fail to overcome the inherent flaws of market mechanisms, making it difficult to achieve efficient and stable task coordination under conditions of local communication and asynchronous execution.

[0005] Therefore, there is an urgent need in this field for a novel collaborative exploration method that can adapt to a fully distributed asynchronous environment and effectively integrate the robot's future intentions. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding. This method aims to solve the problems of frequent task allocation conflicts, high rate of repeated exploration, and poor system stability during multi-robot collaborative exploration, which are caused by asynchronous robot execution states, dynamic evolution of frontier targets, and limited communication conditions.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] This invention proposes a distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding, comprising the following steps:

[0009] S1. Each mobile robot performs localization and mapping synchronously based on its own sensors, generates a partially occupied grid map, and extracts a set of candidate tasks from the partially occupied grid map; each mobile robot maintains its own state information and obtains the state information of other mobile robots in the communication neighborhood, wherein its own state information includes at least its current position and its current task state.

[0010] S2. Each mobile robot determines whether it has entered a state that can accept new tasks based on a predetermined event triggering condition. If a mobile robot enters a state that can accept new tasks, then the mobile robot forms a communication cooperation group with other mobile robots in the communication neighborhood.

[0011] S3. Based on the asynchronous bidding mechanism of intended bids, tasks are allocated within the current communication collaboration group, specifically including the following sub-steps:

[0012] S301. Divide the mobile robots in the current communication collaboration group into two categories: idle robots and busy robots. The idle robots are in a state where they can accept new tasks, and the busy robots are in a state where they are performing tasks and have not triggered the event triggering condition.

[0013] S302. For each task in the candidate task set, the idle robot calculates the immediate benefit evaluation value of the task for itself based on its current position and the directional consistency coefficient between itself and the task.

[0014] S303. The busy robot calculates its potential intention evaluation value for each task in the candidate task set based on its current position or its current target position. For tasks with a positive potential intention evaluation value, the robot publishes the potential intention evaluation value of the task as an intention bid within the current communication collaboration group.

[0015] S304. The idle robot determines the initial cost of each task in the candidate task set based on the intended bid published in the current communication cooperation group; and calculates the net benefit of each task for itself based on the immediate benefit evaluation value and the initial cost.

[0016] S305. Based on the net revenue, perform multiple rounds of local distributed bidding among the idle robots in the current communication cooperation group to determine the final robot to execute each task;

[0017] S4. The final execution robot performs its corresponding task, repeating S1 to S4 until the exploration task termination condition is met.

[0018] Furthermore, in S1, the current task status includes the status of currently executing a task and the status of accepting new tasks, and the candidate task set is the set of leading grid clusters extracted from the locally occupied grid map.

[0019] Furthermore, the feature is that, in S2, the event triggering conditions include at least: the mobile robot arrives at the currently assigned task, the execution time of the current task exceeds a preset threshold, or it detects that its own target task overlaps with the target tasks of other mobile robots in the communication neighborhood.

[0020] Furthermore, in S302, the formula for calculating the immediate benefit evaluation value is as follows:

[0021]

[0022] In the formula, The first in the candidate task set within the current communication collaboration group The first task, namely the... A front-end grille cluster, for For idle robots The immediate benefit evaluation value, yes The estimated information gain, for Current location for arrive European distance, , is the distance smoothing term; for and The directional consistency coefficient between them.

[0023] Furthermore, the directional consistency coefficient The method for determining it is as follows:

[0024] Idle Robot First, calculate the current direction vector of its own motion and its direction. The cosine of the angle between the direction vectors Next, determine if the direction information is valid; finally, determine the direction using the following formula. :

[0025]

[0026]

[0027] In the formula, for Current direction vector of motion, for point to directional vector, It is a very small positive number; This is a function indicating the validity of the direction information; if the direction information is invalid, then... =0; if the direction information is valid, then =1;

[0028] The method to determine whether direction information is valid is: if itself and If the distance is less than the distance threshold or the displacement of itself within a preset time is less than the displacement threshold, the direction information is deemed invalid; otherwise, the direction information is deemed valid.

[0029] Furthermore, in S305, the process of multi-round locally distributed bidding specifically includes:

[0030] S3051. Each idle machine takes the task with the highest net profit as its local optimal task.

[0031] S3052. When the same task is selected as the local best task by multiple idle robots at the same time, a bidding for the task is triggered; the idle robots participating in the bidding calculate their bids based on the net revenue of their local best task.

[0032] S3053. Compare all offers, update the highest offer to the new price for the task, and mark the idle robot that submitted the highest offer as the temporary owner of the task.

[0033] Repeat steps S3051 to S3053 until there are no more competing conflicts within the current communication collaboration group. At this point, the temporary owner of each task is the robot that will ultimately execute the task.

[0034] Furthermore, in S305, after determining the final robot to perform each task, the task allocation results are published within the current communication collaboration group; after allocating the task allocation results, the updated prices of each task and the quotes of each idle robot are cleared.

[0035] Furthermore, the calculation method for the potential intention evaluation value described in S303 is the same as the calculation method for the immediate benefit evaluation value described in S302.

[0036] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-mobile robot distributed asynchronous frontier exploration method based on intention bidding as described in any one of claims 1 to 8.

[0037] The present invention also proposes 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 multi-mobile robot distributed asynchronous frontier exploration method based on intention bidding as described in any one of claims 1 to 8.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] (1) This invention adopts an event-driven triggering mechanism, in which the robot participates in or initiates a new round of local allocation asynchronously only when its own task state changes. At the same time, by introducing the intention bid of busy robots, this invention enables robots in the execution state to express their potential future coverage intentions to the system. This mechanism effectively coordinates the behavior of robots in different states, namely idle and busy, and avoids the problem of multiple robots choosing the same frontier area due to incomplete information and mismatched decision timing. Stable and conflict-free task allocation is achieved in a fully distributed asynchronous environment.

[0040] (2) This invention forms a virtual cost for the task by aggregating intended bids, guiding the robot to prioritize areas not potentially covered. At the same time, a directional consistency coefficient is introduced into the utility function to optimize the allocation results and reduce ineffective turns. The two work together to significantly reduce redundant exploration and ineffective movements. In addition, all decisions are limited to local communication cooperation groups, and the state is immediately cleared after bidding, realizing a lightweight communication mode of on-demand negotiation. While improving spatial coverage efficiency, it significantly reduces system communication overhead.

[0041] (3) The present invention adopts a fully distributed architecture, which does not rely on any central node. The failure of any robot only affects its local neighborhood. The system can adapt to dynamically changing network topology and the number of robots, and has fault tolerance for communication interruption, node addition or removal. This makes the present invention exhibit excellent robustness and scalability in large-scale real-world application scenarios with limited communication and complex environments. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the overall execution flow of the distributed asynchronous frontier exploration method for multiple mobile robots in an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the event-driven local task allocation process based on intention bidding in an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of local frontal detection and information gain of a single robot in an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of the asynchronous distributed collaborative exploration effect of multiple robots in an embodiment of the present invention, where (a) is the environment in which the robots are located, and (b) is the exploration result and the movement trajectory of each robot. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example

[0048] refer to Figure 1 This embodiment proposes a distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding, which is implemented according to the following steps:

[0049] S1. Obtain the candidate task set and the local states of multiple mobile robots, including the following sub-steps:

[0050] S101. Multiple mobile robots are randomly deployed in an unknown environment. Each robot is equipped with a lidar, a wireless communication module and a computing unit. The robots form a communication neighborhood by limiting the communication radius.

[0051] Each robot acquires distance point cloud data of the surrounding environment through LiDAR and uses a simultaneous localization and mapping algorithm to construct a local occupancy grid map. This map is represented in probabilistic form, where the grid state includes known free areas, known obstacle areas, and unobserved areas.

[0052] S102. Each robot, within its locally occupied grid map, detects all known free grids and the boundary grids adjacent to unobserved grids, i.e., leading edge points. These leading edge points are then spatially clustered based on Euclidean distance to form a set of leading edge grid clusters. Each leading edge grid cluster is represented by the centroid coordinates of all its leading edge points, serving as the navigation target point for the robot to explore. The number of unobserved grids within a certain radius (e.g., the maximum range of the robot's sensors) around this cluster is used as its estimated information gain. The set of leading edge grid clusters formed in this step constitutes the current set of candidate tasks to be assigned.

[0053] At the same time, each robot maintains a local state vector:

[0054]

[0055] in, For the first A robot Maintained local state vector, for Current location for Current target location for The set of visible neighborhoods, i.e. All other robots within the communication neighborhood; for A locally occupied raster map is constructed.

[0056] robot visible neighborhood set for:

[0057]

[0058] in for Other robots in the communication neighborhood Current location for The communication radius is used to represent The maximum distance at which direct communication is possible. It will share information with other robots in the communication neighborhood.

[0059] S2. Each robot continuously monitors its own status information and determines whether it should transition from a busy (task-performing) state to an idle (accepting new tasks) state based on the following event triggering conditions:

[0060] a. Target completion event: The mobile robot reaches the currently assigned task. , for Location tolerance range;

[0061] b. Task Failure Event: The execution time of the current task exceeds a preset threshold. ;

[0062] c. Target conflict events: , Threshold for determining target conflict. for The current target point is when the robot detects that its own target task overlaps with the target tasks of other robots within the communication neighborhood. The robot... The robot marks its own task status as idle and broadcasts this status to robots in its communication neighborhood. Robots that receive this broadcast communicate with each other. Form a communication collaboration group and trigger entry into the S3 phase.

[0063] S3. An asynchronous bidding mechanism based on intended bids is used to allocate tasks within the current communication collaboration group. Figure 2 The task assignment process is demonstrated:

[0064] S301. Divide the mobile robots in the current communication cooperation group into a set of idle robots. Collection with Busy Robots All robots in the idle robot set are in an idle state, while all robots in the busy robot set are in a busy state and have not triggered the event triggering condition.

[0065] S302. Idle robots within the current communication cooperation group. For each task in the candidate task set, calculate the immediate benefit evaluation value of the task for the user based on the user's current position and the directional consistency coefficient between the user and the task:

[0066]

[0067] In the formula, The first in the candidate task set within the current communication collaboration group The first task, namely the... A front-end grille cluster, for for The immediate benefit evaluation value, yes The estimated information gain, for Current location for arrive European distance, , is the distance smoothing term, to prevent The time-dependent divergence leads to an infinite preference for "close-up targets"; at the same time, it prevents close-range differences from being excessively amplified, thus enhancing stability. for and The directional consistency coefficient between them.

[0068] Directional consistency coefficient The method for determining it is as follows:

[0069] Idle Robot First, calculate the current direction vector of its own motion and its direction. The cosine of the angle between the direction vectors Next, determine if the direction information is valid; finally, determine the direction using the following formula. :

[0070]

[0071]

[0072] In the formula, for Current direction vector of motion, for point to directional vector, For a very small positive number, in this embodiment, ; This is a function indicating the validity of the direction information; if the direction information is invalid, then... =0; if the direction information is valid, then =1.

[0073] The method to determine whether direction information is valid is: if itself and If the distance is less than the distance threshold or the displacement of itself within a preset time is less than the displacement threshold, the direction information is deemed invalid; otherwise, the direction information is deemed valid.

[0074] In this step, when calculating the immediate benefit evaluation value of the task for itself, the directional consistency coefficient between the robot and the task is introduced, which can effectively reduce the target jitter and frequent replanning problems caused by local greedy decision-making during multi-robot exploration.

[0075] S303, Busy Robot Based on its current position or its current target position, calculate its potential intention evaluation value for each task in the candidate task set:

[0076]

[0077] In the formula, for for Potential intention evaluation value for Current location Or the current target location , for arrive European distance, for and The directional consistency coefficient between them, its calculation method is the same as The same.

[0078] when When >0, Will The potential intention evaluation value is published as an intention bid within the current communication collaboration group;

[0079] S304, Each idle robot receives... Sum all the intended bids to get The initial cost; subsequently, each idle robot uses Evaluation of one's own immediate earnings minus The initial cost of the leading edge grid cluster was calculated. Net profit for itself;

[0080] In this step, by aggregating the intended bids of busy robots into the initial competitive price of the task, idle robots can be guided to prioritize frontier areas that have not yet been potentially covered, without increasing the number of actual bidding participants, thereby reducing the probability of repeated exploration and reducing communication overhead.

[0081] S305. Based on the obtained net profit, determine the final robot to execute each task. The specific process is as follows:

[0082] S3051. Each idle machine will take the front grid cluster with the highest net profit as its local optimal task.

[0083] S3052. When the same task is simultaneously selected as the local best task by multiple idle robots, an auction for that task is triggered. The idle robots participating in the auction are the multiple idle robots that simultaneously selected the same task as the local best task. During this round of auction, the participating idle robots... Calculate the quote based on the net revenue of the current best local task:

[0084]

[0085] In the formula, for The current quote for the best local task. The current local optimal task pair Net income, This is for bidding to increase volume.

[0086] S3053. Compare all bids, update the highest bid to the new price for the task, and mark the idle robot that submitted the highest bid as the temporary owner of the task.

[0087] Repeat steps S3051 to S3053 until there are no more competing conflicts within the current communication collaboration group. At this point, the temporary owner of each task becomes the final executing robot.

[0088] S306. After determining the final executor of each task, the tasks obtained by each winning robot will be announced within the current communication collaboration group. All robots will update their local records accordingly. Subsequently, the current communication collaboration group will be disbanded, and all temporary prices, quotations, and other data related to this bidding will be completely cleared. The system will not retain any cross-event state to ensure that each allocation is independent and clean.

[0089] In S4, the robot, having received a task, takes its assigned leading edge grid cluster as the target task. Based on its local grid map occupancy and real-time sensor information, it independently runs path planning and motion control algorithms (such as D*Lite, TEB, etc.) to achieve autonomous movement, obstacle avoidance, and exploration towards the target task. During navigation, the robot continuously performs map updates and state maintenance as described in S1. Once the task is completed (reaching the target, timeout, or conflict occurs), the event judgment in S2 is triggered again, thus starting a new cycle of S1-S4 until there is no more explorable leading edge in the environment or the preset coverage rate is reached.

[0090] Figure 4 This illustration shows the effect of multi-robot collaborative exploration in this embodiment. Figure 4 (a) shows the initial deployment locations of two mobile robots in an unknown indoor environment. Figure 4 (b) shows the respective motion trajectories of the two robots after collaborative exploration using the method proposed in this embodiment. As can be seen from the figure, the two trajectories cover the main areas of the environment with very little overlap, indicating that the method of the present invention effectively allocates different exploration areas to the robots, avoiding task conflicts and repeated exploration, and achieving efficient distributed asynchronous collaboration.

[0091] The specific embodiments of the present invention are provided to enable those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention.

[0092] It should be understood that the present invention is not limited to the content already described above, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding, characterized in that, Includes the following steps: S1. Each mobile robot performs localization and mapping synchronously based on its own sensors, generates a partially occupied grid map, and extracts a set of candidate tasks from the partially occupied grid map; each mobile robot maintains its own state information and obtains the state information of other mobile robots in the communication neighborhood, wherein its own state information includes at least its current position and its current task state. S2. Each mobile robot determines whether it has entered a state that can accept new tasks based on a predetermined event triggering condition. If a mobile robot enters a state that can accept new tasks, then the mobile robot forms a communication cooperation group with other mobile robots in the communication neighborhood. S3. Based on the asynchronous bidding mechanism of intended bids, tasks are allocated within the current communication collaboration group, specifically including the following sub-steps: S301. Divide the mobile robots in the current communication collaboration group into two categories: idle robots and busy robots. The idle robots are in a state where they can accept new tasks, and the busy robots are in a state where they are performing tasks and have not triggered the event triggering condition. S302. For each task in the candidate task set, the idle robot calculates the immediate benefit evaluation value of the task for itself based on its current position and the directional consistency coefficient between itself and the task. S303. The busy robot calculates its potential intention evaluation value for each task in the candidate task set based on its current position or its current target position. For tasks with a positive potential intention evaluation value, the robot publishes the potential intention evaluation value of the task as an intention bid within the current communication collaboration group. S304. The idle robot determines the initial cost of each task in the candidate task set based on the intended bids published within the current communication cooperation group. Based on the immediate benefit evaluation value and the initial cost, calculate the net benefit of each task for itself. S305. Based on the net revenue, perform multiple rounds of local distributed bidding among the idle robots in the current communication cooperation group to determine the final robot to execute each task; S4. The final execution robot performs its corresponding task, repeating S1 to S4 until the exploration task termination condition is met.

2. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, In S1, the current task status includes the status of executing a task and the status of accepting new tasks, and the candidate task set is the set of leading grid clusters extracted from the locally occupied grid map.

3. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, In S2, the event triggering conditions include at least: the mobile robot arrives at the currently assigned task, the execution time of the current task exceeds a preset threshold, or it detects that its own target task overlaps with the target tasks of other mobile robots in the communication neighborhood.

4. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, In S302, the formula for calculating the immediate benefit evaluation value is as follows: In the formula, The first in the candidate task set within the current communication collaboration group The first task, namely the first A front-end grille cluster, for For idle robots The immediate benefit evaluation value, yes The estimated information gain, for Current location for arrive European distance, , is the distance smoothing term; for and The directional consistency coefficient between them.

5. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 4, characterized in that, Directional consistency coefficient The method for determining it is as follows: Idle Robot First, calculate the current direction vector of its own motion and its direction. The cosine of the angle between the direction vectors ; Next, determine if the direction information is valid; finally, determine it using the following formula. : In the formula, for Current direction vector of motion, for point to directional vector, It is a very small positive number; This is a function indicating the validity of the direction information; if the direction information is invalid, then... =0; if the direction information is valid, then =1; The method to determine whether direction information is valid is: if itself and If the distance is less than the distance threshold or the displacement of itself within a preset time is less than the displacement threshold, the direction information is deemed invalid; otherwise, the direction information is deemed valid.

6. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, In S305, the process of multi-round locally distributed bidding specifically includes: S3051. Each idle machine takes the task with the highest net profit as its local optimal task. S3052. When the same task is selected as the local best task by multiple idle robots at the same time, a bidding for the task is triggered; the idle robots participating in the bidding calculate their bids based on the net revenue of their local best task. S3053. Compare all offers, update the highest offer to the new price for the task, and mark the idle robot that submitted the highest offer as the temporary owner of the task. Repeat steps S3051 to S3053 until there are no more competing conflicts within the current communication collaboration group. At this point, the temporary owner of each task is the robot that will ultimately execute the task.

7. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, In S305, after determining the final robot to perform each task, the task allocation results are published within the current communication collaboration group; after allocating the task allocation results, the updated prices for each task and the quotes for each idle robot are cleared.

8. The distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in claim 1, characterized in that, The calculation method for the potential intention evaluation value described in S303 is the same as the calculation method for the immediate benefit evaluation value described in S302.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the multi-mobile robot distributed asynchronous frontier exploration method based on intention bidding as described in any one of claims 1 to 8.

10. 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 program, it implements the distributed asynchronous frontier exploration method for multiple mobile robots based on intention bidding as described in any one of claims 1 to 8.

Images