A multi-agent autonomous robot system based on an active observation mode

Abstract

The present invention provides a multi-agent autonomous robot system based on an active observation mode. The method includes: generating a task behavior decision according to the current environmental information and the task objective; generating a task behavior sequence according to the task behavior decision, and controlling the autonomous robot to execute the task behaviors in the task behavior sequence; during the execution of the task behaviors, sending an observation behavior decision request according to the current environmental information, and generating an observation behavior based on the active observation mode according to the observation behavior decision request; determining the scheduling timing for generating the observation behavior, and sending a scheduling request; evaluating the scheduling request; when the current environmental information meets the triggering event, controlling the autonomous robot to execute the observation behavior and generating an observation result; updating the current environmental information according to the observation result; and controlling the autonomous robot to continue to execute the task behaviors according to the updated environmental information. The present invention can enable the autonomous robot to obtain environmental information comprehensively and in a timely manner, and improve the effectiveness of the observation of the autonomous robot.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and particularly to a multi-agent autonomous robot system based on an active observation mode. Background Art

[0002] Autonomous robots operate in an open environment. The dynamics and partial observability of the open environment limit the perception of the autonomous robot for the environment, making it difficult to obtain complete and timely information about the environment, and thus making it difficult to make decisions to generate effective task behaviors to achieve tasks. For example, in the library service scenario, the autonomous robot needs to navigate to the book location according to the reader's needs to pick up books for the reader. However, in this scenario, the book location may change, and the autonomous robot cannot comprehensively and timely know the location of the target book, resulting in the inability of the decision-making task behavior to effectively deliver the book. The limited environmental information of the autonomous robot has become a key factor restricting the effective implementation of tasks.

[0003] In existing work, autonomous robots mainly adopt domain-related observation modes and reactive observation modes.

[0004] The domain-related observation mode refers to a special observation mode adopted by autonomous robots for specific domain problems. For example, in the grasping domain, the grasping of an object highly depends on the environmental state, which requires the autonomous robot to have a comprehensive and clear understanding of the current environment and perform a large amount of calculations based on this understanding. Therefore, autonomous robots usually follow the "sense-plan-grasp" observation mode. The domain-related observation mode has excellent performance in its focused domain, but it is difficult to generalize to other problems because it is too focused on the characteristics of specific domains.

[0005] The reactive observation mode refers to a type of observation mode in which an autonomous robot triggers observations according to rules set by the user, and is widely used in tasks that require emergency handling. For example, in the navigation task, when there is a deviation between the environmental model and the perception data, the autonomous robot triggers an observation behavior to observe the surrounding environment and correct the environmental model. The reactive observation mode enables the autonomous robot to quickly respond to environmental changes and has strong real-time performance, but it lacks the decision-making and execution of observing behaviors considering long-term goals.

[0006] The above two types of observation modes lack initiative, that is, the autonomous robot cannot spontaneously make decisions, schedule, and execute observation behaviors according to task requirements during the task process, and it is difficult to obtain task-related environmental information as needed and efficiently. This causes the autonomous robot to make decisions based on one-sided and outdated environmental information, and the task behavior sequence determined by the decision is difficult to effectively complete the task. Therefore, the autonomous robot needs to observe the environment more actively: that is, on the premise of clarifying the observation motivation, spontaneously make decisions, schedule, and execute observation behaviors, so that the autonomous robot can conduct targeted observations at the appropriate time and place, comprehensively and timely obtain task-related environmental information, and reduce the impact of the restrictive factor of limited environmental information on the effective implementation of the task.

[0007] The above requirements for active observation pose challenges to the observation mode and software architecture design of autonomous robots. On the one hand, the requirement for active observation requires the autonomous robot to clarify which scenarios in the open environment will lead to limited environmental information; on this basis, it is necessary to further clarify what targets the autonomous robot should observe, and when, where to make decisions, schedule, and execute observation behaviors, which makes the design of the autonomous robot's observation mode more complex and difficult. On the other hand, the complex mechanism brought by the active observation mode requires the autonomous robot software system to provide software components that support functions such as observation and decision-making, and support the flexible interaction between these software components, which makes it complex to abstract and design an autonomous robot software architecture that effectively adapts to the active observation mode. Summary of the Invention

[0008] The purpose of the present invention is to provide a multi-agent autonomous robot system based on an active observation mode to solve the problem that the active robot has limited perception of environmental information and is difficult to obtain environmental information comprehensively and in a timely manner.

[0009] To achieve the above purpose, the present invention provides the following solutions:

[0010] An active observation method for a multi-agent autonomous robot system based on an active observation mode, including:

[0011] Obtain the current environmental information and task objectives, and generate task behavior decisions according to the current environmental information and the task objectives;

[0012] Generate a task behavior sequence according to the task behavior decision, and control the autonomous robot to execute the task behaviors in the task behavior sequence;

[0013] During the execution of the task behaviors, send an observation behavior decision request according to the current environmental information;

[0014] Generate an observation behavior based on the active observation mode according to the observation behavior decision request; the active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation;

[0015] Decide the scheduling timing for generating the observation behavior, and send a scheduling request based on the scheduling timing;

[0016] Evaluate the scheduling request;

[0017] Based on the scheduling request, when the current environmental information meets the trigger event, control the autonomous robot to execute the observation behavior and generate an observation result;

[0018] Update the current environmental information according to the observation result;

[0019] Control the autonomous robot to continue to execute the task behavior according to the updated environmental information.

[0020] Optionally, the execution process of the active observation mode for one-sided observation specifically includes:

[0021] When the precondition of the task behavior is missing, send the missing environmental information;

[0022] Based on the missing environmental information, decide to generate an observation behavior;

[0023] Autonomously decide the scheduling timing based on the observation behavior. When the current environmental information meets the execution condition of the observation behavior, send a scheduling request;

[0024] Online evaluate the scheduling request and return a scheduling signal;

[0025] When the scheduling signal is a scheduling permission signal, execute the observation behavior to search for the missing environmental information, generate an observation result, and send an execution end signal.

[0026] Optionally, the execution process of the active observation mode for outdated observation specifically includes:

[0027] Obtain the confidence level. When the confidence level is lower than the confidence level threshold, decide to generate an observation behavior;

[0028] Autonomously decide the scheduling timing based on the observation behavior. When the current environmental information meets the scheduling situation, send a scheduling request;

[0029] Online evaluate the scheduling request and return a scheduling signal; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal;

[0030] When the scheduling signal is a scheduling permission signal, execute the observation behavior, generate an observation result, and send an execution end signal.

[0031] Optionally, online evaluate the scheduling request and return a scheduling signal, specifically including:

[0032] Obtain the maximum execution time for executing the task behavior and the consumption time for executing the observation behavior;

[0033] Based on the maximum execution time, determine whether the consumption time is less than the remaining execution time for executing the task behavior;

[0034] If so, return the scheduling permission signal;

[0035] If not, return the scheduling rejection signal.

[0036] A multi-agent autonomous robot system based on an active observation mode, including:

[0037] A task behavior decision-making agent, configured to obtain current environment information and a task target, and generate a task behavior decision according to the current environment information and the task target;

[0038] The task behavior decision-making agent is further configured to generate a task behavior sequence according to the task behavior decision, and control an autonomous robot to execute the task behavior in the task behavior sequence;

[0039] A task behavior execution agent, configured to send an observation behavior decision request according to current environment information during the execution of the task behavior;

[0040] An observation behavior decision-making agent, configured to generate an observation behavior based on an active observation mode according to the observation behavior decision request; the active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation;

[0041] An observation behavior execution agent, configured to decide the scheduling timing for generating the observation behavior, and send a scheduling request based on the scheduling timing;

[0042] A behavior scheduling agent, configured to evaluate the scheduling request;

[0043] The observation behavior execution agent is further configured to, based on the scheduling request, when the current environment information meets a trigger event, control the autonomous robot to execute the observation behavior and generate an observation result;

[0044] A knowledge management agent, configured to update the current environment information according to the observation result;

[0045] The task behavior execution agent further controls the autonomous robot to continue executing the task behavior according to the updated environment information.

[0046] Optionally, the execution process of the active observation mode for one-sided observation specifically includes:

[0047] When the precondition of the task behavior is missing, the task behavior execution agent sends the missing environmental information to the observation behavior decision-making agent; based on the missing environmental information, the observation behavior decision-making agent makes a decision to generate an observation behavior and sends the observation behavior to the observation behavior execution agent; the observation behavior agent autonomously makes a decision to generate a scheduling opportunity based on the observation behavior. When the current environmental information meets the execution condition of the observation behavior, it sends a scheduling request to the behavior scheduling agent, and through online evaluation of the scheduling request, returns a scheduling signal; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior to search for the missing environmental information, generates an observation result, updates the missing environmental information to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

[0048] Optionally, the execution process of the active observation mode for outdated observation specifically includes:

[0049] The observation behavior decision-making agent receives the confidence level sent by the intelligent management agent. When the confidence level is lower than the confidence level threshold, it makes a decision to generate an observation behavior; the observation behavior execution agent autonomously makes a decision to generate a scheduling opportunity based on the observation behavior. When the current environmental information meets the scheduling situation, it sends a scheduling request to the behavior scheduling agent; the behavior scheduling agent returns a scheduling signal through online evaluation of the scheduling request; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior, generates an observation result, updates the observation result to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

[0050] Optionally, the behavior scheduling agent specifically includes:

[0051] A maximum execution time and consumption time acquisition module for acquiring the maximum execution time of executing the task behavior and the consumption time of executing the observation behavior;

[0052] A judgment module for judging whether the consumption time is less than the remaining execution time of executing the task behavior based on the maximum execution time;

[0053] A scheduling permission signal return module for, if so, returning the scheduling permission signal;

[0054] A scheduling rejection signal return module for, if not, returning the scheduling rejection signal.

[0055] According to the specific embodiments provided by the present invention, the following technical effects are disclosed: The present invention provides a multi-agent autonomous robot system based on an active observation mode. Based on two types of motivations, namely, the lack of environmental information caused by partial observation and the inconsistency of environmental information caused by outdated observation, two types of active observation modes are specifically proposed. The coordination mechanism between task behaviors and observation behaviors is described, corresponding observation behavior decision-making and scheduling are proposed, and timely feedback of observation results and behavior adaptive adjustment are realized, enabling the autonomous robot to obtain environmental information comprehensively and in a timely manner and improving the effectiveness of the autonomous robot's observation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0057] Figure 1 It is a flowchart of the active observation method for the multi-agent autonomous robot system based on the active observation mode provided by the present invention;

[0058] Figure 2 It is a structure diagram of the multi-agent autonomous robot system based on the active observation mode provided by the present invention;

[0059] Figure 3 It is a flowchart of the two-stage observation behavior decision-making for the multi-agent autonomous robot system based on the active observation mode provided by the present invention;

[0060] Figure 4 It is an interaction diagram of the multi-agent implementation of the active observation mode for partial observation provided by the present invention;

[0061] Figure 5 It is an interaction diagram of the multi-agent implementation of the active observation mode for partial observation provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0063] The object of the present invention is to provide a multi-agent autonomous robot system based on an active observation mode, which can enable the autonomous robot to obtain environmental information comprehensively and in a timely manner and improve the effectiveness of the autonomous robot's observation.

[0064] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0065] Embodiment 1

[0066] As Figure 1 shown, the present invention provides an active observation method for a multi-agent autonomous robot system based on an active observation mode, including:

[0067] Step 101: Obtain the current environmental information and task objectives, and generate a task behavior decision based on the current environmental information and the task objectives.

[0068] Step 102: Generate a task behavior sequence based on the task behavior decision, and control the autonomous robot to execute the task behaviors in the task behavior sequence.

[0069] Step 103: During the execution of the task behavior, send an observation behavior decision request according to the current environmental information.

[0070] Step 104: Generate an observation behavior based on the active observation mode according to the observation behavior decision request; the active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation.

[0071] In practical applications, the execution process of the active observation mode for one-sided observation specifically includes: when the preconditions of the task behavior are missing, send the missing environmental information; based on the missing environmental information, decide to generate an observation behavior; autonomously decide a scheduling opportunity based on the observation behavior, and when the current environmental information meets the execution conditions of the observation behavior, send a scheduling request; online evaluate the scheduling request and return a scheduling signal; when the scheduling signal is a scheduling permission signal, execute the observation behavior to search for the missing environmental information, generate an observation result, and send an execution end signal.

[0072] In practical applications, the execution process of the active observation mode for outdated observation specifically includes: obtain a confidence level, and when the confidence level is lower than a confidence level threshold, decide to generate an observation behavior; autonomously decide a scheduling opportunity based on the observation behavior, and when the current environmental information meets the scheduling situation, send a scheduling request; online evaluate the scheduling request and return a scheduling signal; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; when the scheduling signal is a scheduling permission signal, execute the observation behavior, generate an observation result, and send an execution end signal.

[0073] Step 105: Decide the scheduling opportunity of the observation behavior, and send a scheduling request based on the scheduling opportunity.

[0074] Step 106: Evaluate the scheduling request.

[0075] Step 107: Based on the scheduling request, when the current environmental information meets the triggering event, control the autonomous robot to perform the observation behavior and generate an observation result.

[0076] Step 108: Update the current environmental information according to the observation result.

[0077] Step 109: Control the autonomous robot to continue to perform the task behavior according to the updated environmental information.

[0078] In practical applications, online evaluate the scheduling request and return a scheduling signal, specifically including: obtaining the maximum execution time for performing the task behavior and the consumption time for performing the observation behavior; based on the maximum execution time, determine whether the consumption time is less than the remaining execution time for performing the task behavior; if so, return the scheduling permission signal; if not, return the scheduling rejection signal.

[0079] The present invention verifies the effectiveness of the active observation mode by comparing it with the reactive observation and concomitant observation modes in the book transfer task: in the comparison of the action path and time consumption, the active observation mode has a movement path of 32.7 m, which is shorter than 61.1 m of the reactive observation mode and 69.7 m of the concomitant observation mode; and a time consumption of 365.6 s, which is lower than 525.3 s of the reactive observation mode and 483.7 s of the concomitant observation mode.

[0080] Embodiment 2

[0081] To execute the method corresponding to Embodiment 1 above to achieve the corresponding functions and technical effects, the following provides a multi-agent autonomous robot system based on the active observation mode.

[0082] As Figure 2 shown, a multi-agent autonomous robot system based on the active observation mode includes:

[0083] A task behavior decision-making agent, configured to obtain the current environmental information and the task objective, and generate a task behavior decision according to the current environmental information and the task objective.

[0084] The task behavior decision-making agent is further configured to generate a task behavior sequence according to the task behavior decision and control the autonomous robot to perform the task behavior in the task behavior sequence.

[0085] In practical applications, the task behavior decision-making agent is responsible for making decisions on task behaviors. It generates a task behavior sequence based on the current environmental information and task objectives, and sends the task behavior sequence to the behavior scheduling agent and the corresponding task behavior execution agent.

[0086] The task behavior execution agent is used to send the observation behavior decision request according to the current environmental information during the execution of the task behavior.

[0087] The observation behavior decision-making agent is used to generate an observation behavior based on the active observation mode according to the observation behavior decision request; the active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation.

[0088] In practical applications, the observation behavior decision-making agent implements an observation behavior decision function. It receives observation behavior decision requests from the knowledge management agent and the task behavior execution agent, generates an observation behavior based on the active observation mode, and sends the observation behavior information to the corresponding observation behavior execution agent.

[0089] There are mainly two types of scenarios that limit the environment of autonomous robots: one-sided observation and outdated observation. One-sided observation refers to the fact that in an open environment, the partial observability of the environment and the limited sensing resources make it difficult for autonomous robots to comprehensively perceive the environment, lacking environmental information related to the task and unable to execute task behaviors. Outdated observation refers to the fact that in an open environment, the dynamic nature of the environment makes it difficult for autonomous robots to timely perceive changes in the environment, and the obtained environmental information is inconsistent with the current environmental state, and the task behaviors generated by the decision-making are difficult to complete the task. Figure 3 The two-stage observation behavior decision flowchart provided by the present invention is as Figure 3 shown.

[0090] Algorithm 1: Observation behavior decision in the offline stage

[0091]

[0092]

[0093] In the online stage, when the observation behavior β v for outdated observation is executed and its observation result τ shows that the precondition α(μ) of the task behavior α is inconsistent with the current environmental state s, α(μ) is missing. At this time, the autonomous robot generates an observation behavior β s for one-sided observation with α(μ) as the observation target based on the active observation mode π s (as shown in Algorithm 2).

[0094] As the task is executed, the confidence level conf of the precondition α(μ) of α will gradually decrease. When it is lower than the threshold thres, the autonomous robot will be based on the active observation mode π for the outdated observation. v Generate the observation behavior β for the outdated observation. v (as shown in Algorithm 3).

[0095] In practical applications, the execution process of the active observation mode for one-sided observation specifically includes: when the precondition of the task behavior is missing, the task behavior execution agent sends the missing environmental information to the observation behavior decision agent; the observation behavior decision agent makes a decision on the observation behavior based on the missing environmental information and sends the observation behavior to the observation behavior execution agent; the observation behavior agent makes an autonomous decision on the scheduling opportunity, and when the current environmental information meets the execution condition of the observation behavior, it sends a scheduling request to the behavior scheduling agent, and returns a scheduling signal by online evaluating the scheduling request; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior to search for the missing environmental information, generates an observation result, updates the missing environmental information to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

[0096] As an optional implementation manner of the present invention, the active observation mode for one-sided observation occurs among the knowledge management agent, the task behavior execution agent, the observation behavior decision agent, the observation behavior execution agent, and the behavior scheduling agent.

[0097] Figure 4 The multi-agent interaction diagram of the active observation mode for one-sided observation provided by the present invention is as Figure 4 shown. When the task behavior execution agent receives the autonomous robot knowledge sent by the knowledge management agent, it judges whether the precondition of the task behavior is missing. When the precondition is missing, the task behavior execution agent sends the missing environmental information to the observation behavior decision agent. The observation behavior decision agent makes a decision on the observation behavior based on this message and sends the observation behavior to the observation behavior execution agent. The observation behavior execution agent makes an autonomous decision on the scheduling opportunity. When the environment meets the execution condition of the observation behavior, it sends a scheduling request to the behavior scheduling agent. The behavior scheduling agent returns a scheduling signal by online evaluating the scheduling request. After obtaining the scheduling permission signal, the observation behavior execution agent drives the physical device to execute the observation behavior to search for the missing environmental information. When the missing environmental information is obtained, the observation behavior execution agent updates the environmental information to the knowledge management agent and sends an execution end signal to the behavior scheduling agent to release the physical device.

[0098] Algorithm 2: Observation Behavior Decision for One-sided Observation in the Online Phase

[0099] Input: (α, β v )

[0100] Output: β s

[0101] If β v verifies that α(μ) is inconsistent with the environment;

[0102] β s ← observationPlan(π s , α(μ));

[0103] End;

[0104] Return β s .

[0105] In practical applications, the execution process of the active observation mode for outdated observations specifically includes: the observation behavior decision agent receives the confidence level sent by the intelligent management agent, and when the confidence level is lower than the confidence level threshold, a decision is made to generate an observation behavior; the observation behavior execution agent autonomously makes a decision based on the observation behavior to generate a scheduling opportunity, and when the current environmental information meets the scheduling situation, a scheduling request is sent to the behavior scheduling agent; the behavior scheduling agent returns a scheduling signal by online evaluating the scheduling request; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior, generates an observation result, updates the observation result to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

[0106] As an optional implementation manner of the present invention, the active observation mode for outdated observations occurs among the knowledge management agent, the observation behavior decision agent, the observation behavior execution agent, and the behavior scheduling agent. Figure 5 For the multi-agent interaction diagram of the active observation mode for outdated observations provided by the present invention, as shown in Figure 5As shown, after the observation behavior decision-making agent receives the knowledge with a confidence level lower than the threshold sent by the knowledge management agent, it decides to generate an observation behavior and sends it to the observation behavior execution agent. After receiving the observation behavior, the observation behavior execution agent makes an online decision on the scheduling timing. When the environment meets the scheduling conditions, the observation behavior execution agent sends a scheduling request to the behavior scheduling agent. The behavior scheduling agent returns a scheduling signal by online evaluating the scheduling request. After obtaining the scheduling permission signal, the observation behavior execution agent drives the physical device to execute the observation behavior. After the observation behavior is completed, the observation behavior execution agent sends the observation result to the knowledge management agent and sends an execution end signal to the behavior scheduling agent to release the physical device.

[0107] Algorithm 3: Observation Behavior Decision for Obsolete Observations in the Online Phase

[0108]

[0109] The observation behavior execution agent is used to decide the scheduling timing of the observation behavior and send a scheduling request based on the scheduling timing.

[0110] The behavior scheduling agent is used to evaluate the scheduling request.

[0111] In practical applications, the behavior scheduling agent specifically includes: a maximum execution time and consumption time acquisition module for acquiring the maximum execution time of executing the task behavior and the consumption time of executing the observation behavior; a judgment module for judging whether the consumption time is less than the remaining execution time of executing the task behavior based on the maximum execution time; a scheduling permission signal return module for returning the scheduling permission signal if so; and a scheduling rejection signal return module for returning the scheduling rejection signal if not.

[0112] In practical applications, the behavior scheduling agent implements an online evaluation scheduling algorithm. It receives the task behavior sequence sent by the task behavior decision-making agent to schedule the task behavior; receives the scheduling request from the observation behavior execution agent, evaluates whether to suspend the task behavior and execute the observation behavior, and returns a scheduling signal.

[0113] During the execution process, both the task behavior and the observation behavior will control the physical devices of the autonomous robot, and resource conflicts may occur. Therefore, it is necessary to design a scheduling strategy to coordinate the operation of the behaviors. Usually, the task behavior α has a maximum execution time, which is common and reasonable in the real world. For example, the autonomous robot needs to deliver a book to a specified location within a specified time. Therefore, the time consumption caused by executing the observation behavior β should not cause α to time out. This requires the autonomous robot to online evaluate whether the time consumption cost of executing β is less than the remaining execution time residual_time of α, so as to decide whether to schedule the observation behavior. The online evaluation observation behavior scheduling algorithm is shown in Algorithm 4.

[0114] Algorithm 4: Observation Behavior Scheduling with Online Evaluation

[0115]

[0116] The observation behavior execution agent is also used to control the autonomous robot to execute the observation behavior based on the scheduling request when the current environmental information meets the trigger event, and generate an observation result.

[0117] In practical applications, each observation behavior execution agent corresponds to a specific type of observation behavior. It receives the observation behavior information sent by the observation behavior decision agent, decides the scheduling timing of the observation behavior; when the environment meets the trigger event, it sends a scheduling request to the behavior scheduling agent; when the scheduling request is permitted, it controls the autonomous robot to execute the observation behavior.

[0118] The knowledge management agent is used to update the current environmental information according to the observation result.

[0119] The knowledge management agent is responsible for the management of the internal knowledge of the autonomous robot. It receives the knowledge update signals of the task behavior and the observation behavior agents, regularly broadcasts the environmental information, and evaluates whether it is outdated.

[0120] The task behavior execution agent also controls the autonomous robot to continue to execute the task behavior according to the updated environmental information.

[0121] In practical applications, each task behavior execution agent corresponds to a specific type of task behavior. It receives the task behavior information sent by the task behavior decision agent, monitors the preconditions of the task behavior, and judges the executability of the task behavior; when the task behavior execution agent monitors that the preconditions are missing, it sends an observation behavior decision request; when it receives the scheduling signal sent by the behavior scheduling agent and the preconditions are met, it controls the autonomous robot to execute the task behavior.

[0122] The above six types of agents respectively implement the decision-making of observation behaviors, the management of knowledge, the decision-making of task behaviors, the scheduling of behaviors, and the execution of observation behaviors and task behaviors; the agents complete tasks through communication and have the basic functions to implement the active observation mode.

[0123] The present invention implements the above active observation mode based on the autonomous robot software architecture of a multi-agent system. This architecture uses six types of software agents to form the autonomous robot software. Each type of agent implements specific functions, and the control logic of the active observation mode is realized through the communication of the agents.

[0124] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same and similar parts among the various embodiments, reference can be made to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and reference can be made to the description in the method part for the relevant parts.

[0125] Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An active observation method for a multi-agent autonomous robot system based on an active observation mode, characterized in that, Including: Obtain the current environmental information and task objectives, and generate task behavior decisions according to the current environmental information and the task objectives; Generate a task behavior sequence according to the task behavior decision, and control the autonomous robot to execute the task behaviors in the task behavior sequence; During the execution of the task behaviors, send an observation behavior decision request according to the current environmental information; Generate an observation behavior based on the active observation mode according to the observation behavior decision request; The active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation; Decide the scheduling timing for generating the observation behavior, and send a scheduling request based on the scheduling timing; Evaluate the scheduling request; Based on the scheduling request, when the current environmental information meets the trigger event, control the autonomous robot to execute the observation behavior and generate an observation result; Update the current environmental information according to the observation result; Control the autonomous robot to continue to execute the task behavior according to the updated environmental information.

2. The active observation method of the multi-agent autonomous robot system based on the active observation mode according to claim 1, characterized in that The execution process of the active observation mode for one-sided observation specifically includes: When the preconditions of the task behavior are missing, send the missing environmental information; Based on the missing environmental information, decide to generate an observation behavior; Autonomously decide the scheduling timing based on the observation behavior. When the current environmental information meets the execution conditions of the observation behavior, send a scheduling request; Online evaluate the scheduling request and return a scheduling signal; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; When the scheduling signal is a scheduling permission signal, execute the observation behavior to search for the missing environmental information, generate an observation result, and send an execution end signal.

3. The active observation method of the multi-agent autonomous robot system based on the active observation mode according to claim 1, characterized in that, The execution process of the active observation mode for outdated observation specifically includes: Obtain the confidence level of the preconditions of the task behavior. When the confidence level is lower than the confidence level threshold, decide to generate an observation behavior; Autonomously decide the scheduling timing based on the observation behavior. When the current environmental information meets the scheduling situation, send a scheduling request; Online evaluate the scheduling request and return a scheduling signal; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; When the scheduling signal is a scheduling permission signal, execute the observation behavior, generate an observation result, and send an execution end signal.

4. The active observation method of the multi-agent autonomous robot system based on the active observation mode according to claim 2 or 3, characterized in that Online evaluate the scheduling request and return a scheduling signal, specifically including: Obtain the maximum execution time for executing the task behavior and the consumption time for executing the observation behavior; Based on the maximum execution time, judge whether the consumption time is less than the remaining execution time for executing the task behavior; If so, return the scheduling permission signal; If not, return the scheduling rejection signal.

5. A multi-agent autonomous robot system based on an active observation mode, characterized in that, Including: A task behavior decision-making agent, used to obtain the current environmental information and task objectives, and generate task behavior decisions according to the current environmental information and the task objectives; The task behavior decision-making agent is also used to generate a task behavior sequence according to the task behavior decision, and control the autonomous robot to execute the task behaviors in the task behavior sequence; A task behavior execution agent, configured to send an observation behavior decision request according to current environment information during the execution of the task behavior; An observation behavior decision agent, configured to generate an observation behavior based on an active observation mode according to the observation behavior decision request; the active observation mode includes an active observation mode for one-sided observation and an active observation mode for outdated observation; An observation behavior execution agent, configured to determine a scheduling opportunity for generating the observation behavior and send a scheduling request based on the scheduling opportunity; A behavior scheduling agent, configured to evaluate the scheduling request; The observation behavior execution agent is further configured to, based on the scheduling request, when the current environment information meets a trigger event, control the autonomous robot to execute the observation behavior and generate an observation result; A knowledge management agent, configured to update the current environment information according to the observation result; The task behavior execution agent further controls the autonomous robot to continue executing the task behavior according to the updated environment information.

6. The multi-agent autonomous robot system based on the active observation mode according to claim 5, characterized in that, The execution process of the active observation mode for one-sided observation specifically includes: When a precondition of the task behavior is missing, the task behavior execution agent sends the missing environment information to the observation behavior decision agent; the observation behavior decision agent generates an observation behavior based on the missing environment information and sends the observation behavior to the observation behavior execution agent; the observation behavior agent autonomously determines a scheduling opportunity based on the observation behavior, and when the current environment information meets the execution condition of the observation behavior, sends a scheduling request to the behavior scheduling agent, and returns a scheduling signal by online evaluating the scheduling request; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior to search for the missing environment information, generates an observation result, updates the missing environment information to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

7. The multi-agent autonomous robot system based on the active observation mode according to claim 5, characterized in that, The execution process of the active observation mode for outdated observation specifically includes: The observation behavior decision agent receives the confidence level of the precondition of the task behavior sent by the intelligent management agent, and when the confidence level is lower than the confidence level threshold, generates an observation behavior; the observation behavior execution agent autonomously determines a scheduling opportunity based on the observation behavior, and when the current environment information meets the scheduling situation, sends a scheduling request to the behavior scheduling agent; the behavior scheduling agent returns a scheduling signal by online evaluating the scheduling request; the scheduling signal includes a scheduling permission signal and a scheduling rejection signal; when the scheduling signal is a scheduling permission signal, the observation behavior execution agent executes the observation behavior, generates an observation result, updates the observation result to the knowledge management agent, and sends an execution end signal to the behavior scheduling agent.

8. The multi-agent autonomous robot system based on the active observation mode according to claim 6 or 7, characterized in that The behavior scheduling agent specifically includes: The maximum execution time and consumption time acquisition module is used to acquire the maximum execution time for executing the task behavior and the consumption time for executing the observation behavior; The judgment module is used to judge whether the consumption time is less than the remaining execution time for executing the task behavior based on the maximum execution time; The scheduling permission signal return module is used to return the scheduling permission signal if so; The scheduling rejection signal return module is used to return the scheduling rejection signal if not.

Images