Intelligent agent behavior decision method, device and electronic equipment

Abstract

The present application discloses a behavior decision method, device and electronic device for an intelligent agent, belonging to the field of artificial intelligence technology. The method includes: determining the behavior set of the intelligent agent in the current state according to the behavior capability function of the intelligent agent; determining multiple behavior trajectories of the intelligent agent from the current state to the target state according to the behavior set; calculating the costs corresponding to the multiple behavior trajectories based on the current cognition of the intelligent agent according to the cognitive architecture function, and calculating the values ​​corresponding to the multiple behavior trajectories based on the current preference of the intelligent agent according to the value preference function; taking maximizing value and minimizing cost as the optimization goal, selecting the target behavior trajectory that meets the optimization goal from multiple behavior trajectories. The present application comprehensively considers the behavior capability, cognitive architecture and value preference of the intelligent agent, so that the intelligent agent can calmly face complex and changing environments, enhance the autonomous learning ability and adaptability of the intelligent agent, and improve the decision-making accuracy and efficiency of the intelligent agent.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and in particular, relates to a behavior decision-making method, device and electronic equipment for an intelligent entity. Background Art

[0002] In recent years, agent technology has made significant progress in a variety of application fields, especially in reinforcement learning, autonomous driving, game design, social simulation and other fields. The core of agent technology is to simulate the decision-making and behavior of humans or other entities, enabling them to perform tasks autonomously in complex environments. The development of these technologies has not only pushed the boundaries of artificial intelligence, but also provided new tools and methods for solving complex problems in the real world.

[0003] In related technologies, most intelligent agent systems are based on traditional behavior models and target optimization algorithms, such as Q-learning, Deep Reinforcement Learning (DRL), etc. These methods usually rely on clear reward signals and optimization goals to guide the agent to choose the optimal behavior. However, when faced with complex, dynamic and incomplete information environments, these methods do not consider how to simulate the agent's cognitive biases, value preferences, and autonomous exploration behaviors and evolution laws based on them. When the agents based on these methods have cognitive biases and lack autonomous exploration and evolution, they often make wrong decisions. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a behavior decision method, device and electronic device for an intelligent agent to improve the decision accuracy of the intelligent agent.

[0005] In a first aspect, the present application provides a behavior decision method for an intelligent agent, comprising: Determine a behavior set of the agent in the current state according to the behavior capability function of the agent; Determine multiple behavior trajectories of the agent from a current state to a target state according to the behavior set; Calculating the costs corresponding to the plurality of behavior trajectories based on the current cognition of the agent according to the cognitive architecture function, and calculating the values ​​corresponding to the plurality of behavior trajectories based on the current preference of the agent according to the value preference function; Taking maximizing value and minimizing cost as optimization goals, a target behavior trajectory that meets the optimization goals is selected from the multiple behavior trajectories.

[0006] According to the behavior decision method of the agent of the present application, the behavior set of the agent in the current state is determined according to the behavior capability function of the agent; multiple behavior trajectories of the agent from the current state to the target state are determined according to the behavior set; the costs corresponding to the multiple behavior trajectories are calculated according to the cognitive architecture function based on the current cognition of the agent, and the values ​​corresponding to the multiple behavior trajectories are calculated according to the value preference function based on the current preference of the agent; with maximizing value and minimizing cost as the optimization goal, the target behavior trajectory that meets the optimization goal is selected from the multiple behavior trajectories. The behavior set determined by the behavior capability function of the embodiment of the present application enables the agent to consider actions within the scope of its ability, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated by the cognitive architecture function. Considering the current cognition of the agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The value preference function is introduced to evaluate the preference value of the agent for different behavior trajectories, and the behavior capability, cognitive architecture and value preference of the agent are comprehensively considered, so that the agent can select the target behavior trajectory that is both in line with its ability and can maximize value and minimize cost from multiple possible behavior trajectories, thereby improving the accuracy and efficiency of the decision-making of the agent.

[0007] According to one embodiment of the present application, the cost corresponding to the plurality of behavior trajectories is calculated based on the current cognition of the agent according to the cognitive architecture function, including: According to the formula

[0008] Calculating costs corresponding to a plurality of the behavior trajectories; in, Indicates that the agent is in the current state To the target state A behavioral trajectory, Indicates behavior trajectory The corresponding cost, and Indicates different times, is the behavioral capability function, which indicates that the agent t Status at all times The following set of behaviors, is a cognitive architecture function, which is used to evaluate the agent from state Transition to state difficulty.

[0009] In this embodiment, the current cognitive state of the agent is taken into account through the above-mentioned cost calculation formula, and the difficulty of the agent's state transition can be evaluated in combination with the agent's behavioral capabilities and the dynamic changes of the environment, so that the potential costs of different behavioral trajectories can be predicted based on current cognition.

[0010] According to an embodiment of the present application, taking maximizing value and minimizing cost as the optimization goal, selecting a target behavior trajectory that meets the optimization goal from a plurality of behavior trajectories includes: Taking maximizing value and minimizing cost as optimization goals, constructing an objective function according to the cognitive architecture function and the value preference function; Substitute the costs corresponding to the plurality of behavior trajectories and the values ​​corresponding to the plurality of behavior trajectories into the objective function, and select a target behavior trajectory that meets the optimization objective from the plurality of behavior trajectories.

[0011] In this embodiment, an objective function that comprehensively considers value maximization and cost minimization is constructed, and the cognitive architecture function and value preference function are utilized, so that the objective function can quantify the expected value and execution cost of each behavior trajectory. The cost and value of each behavior trajectory are substituted into the objective function. By comparing the scores of different trajectories, the optimal trajectory that maximizes value while minimizing cost can be selected, thereby improving the efficiency of the intelligent agent's decision-making.

[0012] According to an embodiment of the present application, the optimization goal of maximizing value and minimizing cost is to construct an objective function according to the cognitive architecture function and the value preference function, including: According to the formula

[0013] Construct the objective function; in, is the value preference function, which represents the agent's behavior trajectory The value given.

[0014] In this embodiment, by constructing the objective function through the above formula, it is possible to clearly identify those behavioral trajectories that satisfy value preferences while having the lowest cost, thereby making the best decision among the many behavioral trajectories, reducing the number of options that need to be evaluated, and not only improving the efficiency of the intelligent agent's decision-making, but also being able to flexibly find the best balance between value and cost, further enhancing the adaptability of the intelligent agent in complex environments.

[0015] According to an embodiment of the present application, the optimization goal of maximizing value and minimizing cost is to construct an objective function according to the cognitive architecture function and the value preference function, including: According to the formula

[0016] Construct the objective function; in, is the intention probability function, which represents the behavior trajectory attempted by the agent willingness to explore.

[0017] In this embodiment, the objective function is constructed by the above formula, which not only takes into account the value and cost of the behavior trajectory, but also takes into account the willingness to explore new behavior trajectories. This ensures that when facing unknown situations, the intelligent agent is not limited to the known optimal solution, but adjusts the exploration willingness and tries new behavior trajectories in order to find better solutions. This ensures that while pursuing the maximization of immediate value and the minimization of costs, the intelligent agent also invests in exploring new behaviors that may bring long-term benefits, which helps the intelligent agent learn and adapt to complex and changing environments in the long run.

[0018] According to one embodiment of the present application, the method further includes: Controlling the behavior of the agent according to the target behavior trajectory and determining the behavior result of the agent; The behavioral ability function, the cognitive architecture function and the value preference function are optimized according to the behavioral results.

[0019] In this embodiment, the behavior of the intelligent agent is controlled according to the target behavior trajectory and the behavior results are observed, thereby providing the intelligent agent with actual feedback information, allowing the intelligent agent to use the feedback information to optimize its behavioral capability function, cognitive architecture function and value preference function, thereby achieving closed-loop optimization of the intelligent agent's behavioral decision-making process. Through this closed-loop optimization, the intelligent agent can continuously evolve. This evolution not only stems from the intelligent agent's own practical accumulation, but can also be directly affected by external teaching or information input. During the evolution process, cognitive biases and value judgments can be gradually corrected, thereby achieving the intelligent agent's evolutionary process from unknown to known, from immature to mature, so as to adapt to the ever-changing environment and task requirements.

[0020] According to one embodiment of the present application, the optimizing the behavior capability function, the cognitive architecture function and the value preference function according to the behavior result includes: According to the formula

[0021] Or the formula

[0022] Optimizing the behavioral capability function, the cognitive architecture function, and the value preference function; in, represents the optimization function, are the parameters of the intention probability function, is the parameter of the value preference function, are the parameters of the cognitive architecture function, Parameters of the behavior capability function.

[0023] In a second aspect, the present application provides a behavior decision method for an intelligent agent, characterized by comprising: A first determination module, used to determine a behavior set of the agent in a current state according to the behavior capability function of the agent; A second determination module is used to determine multiple behavior trajectories of the agent from a current state to a target state according to the behavior set; A calculation module, used to calculate the costs corresponding to the plurality of behavior trajectories based on the current cognition of the intelligent agent according to the cognitive architecture function, and to calculate the values ​​corresponding to the plurality of behavior trajectories based on the current preference of the intelligent agent according to the value preference function; The selection module is used to take maximizing value and minimizing cost as optimization goals, and select a target behavior trajectory that meets the optimization goals from multiple behavior trajectories.

[0024] According to the behavior decision device of the intelligent agent of the present application, the behavior set of the intelligent agent in the current state is determined according to the behavior capability function of the intelligent agent; multiple behavior trajectories of the intelligent agent from the current state to the target state are determined according to the behavior set; the costs corresponding to the multiple behavior trajectories are calculated according to the cognitive architecture function based on the current cognition of the intelligent agent, and the values ​​corresponding to the multiple behavior trajectories are calculated according to the value preference function based on the current preference of the intelligent agent; with maximizing value and minimizing cost as the optimization goal, the target behavior trajectory that meets the optimization goal is selected from the multiple behavior trajectories. The behavior set determined by the behavior capability function of the embodiment of the present application enables the intelligent agent to consider actions within the scope of its ability, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated by the cognitive architecture function. Considering the current cognition of the intelligent agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The value preference function is introduced to evaluate the preferred value of the intelligent agent for different behavior trajectories, and the behavior capability, cognitive architecture and value preference of the intelligent agent are comprehensively considered, so that the intelligent agent can select the target behavior trajectory that is both in line with its ability and can maximize value and minimize cost from multiple possible behavior trajectories, thereby improving the decision-making accuracy and efficiency of the intelligent agent.

[0025] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the behavior decision method of the intelligent agent as described in the first aspect above is implemented.

[0026] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the behavior decision-making method of an intelligent agent as described in the first aspect above.

[0027] In a fifth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the behavior decision-making method of the intelligent agent as described in the first aspect above.

[0028] In a sixth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the behavior decision-making method of an intelligent agent as described in the first aspect above.

[0029] The above one or more technical solutions in the embodiments of the present application have at least one of the following technical effects: According to the behavior decision method of the agent of the present application, the behavior set of the agent in the current state is determined according to the behavior capability function of the agent; multiple behavior trajectories of the agent from the current state to the target state are determined according to the behavior set; the costs corresponding to the multiple behavior trajectories are calculated according to the cognitive architecture function based on the current cognition of the agent, and the values ​​corresponding to the multiple behavior trajectories are calculated according to the value preference function based on the current preference of the agent; with maximizing value and minimizing cost as the optimization goal, the target behavior trajectory that meets the optimization goal is selected from the multiple behavior trajectories. The behavior set determined by the behavior capability function of the embodiment of the present application enables the agent to consider actions within the scope of its ability, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated by the cognitive architecture function. Considering the current cognition of the agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The value preference function can be introduced to evaluate the preference value of the agent for different behavior trajectories, and the behavior capability, cognitive architecture and value preference of the agent are comprehensively considered, so that the agent can select the target behavior trajectory that is both in line with its ability and can maximize value and minimize cost from multiple possible behavior trajectories, thereby improving the accuracy and efficiency of the decision-making of the agent.

[0030] Furthermore, in some embodiments, the above-mentioned cost calculation formula takes into account the current cognitive state of the agent, and can evaluate the difficulty of the agent's state transition in combination with the agent's behavioral capabilities and the dynamic changes of the environment, thereby being able to predict the potential costs of different behavioral trajectories based on current cognition.

[0031] Furthermore, in some embodiments, by constructing an objective function that comprehensively considers value maximization and cost minimization, the cognitive architecture function and the value preference function are utilized, so that the objective function can quantify the expected value and execution cost of each behavior trajectory, and the cost and value of each behavior trajectory are substituted into the objective function. By comparing the scores of different trajectories, the optimal trajectory that maximizes value while minimizing cost can be selected, thereby improving the efficiency of the intelligent agent's decision-making.

[0032] Furthermore, in some embodiments, by constructing the objective function through the above formula, it is possible to clearly identify those behavioral trajectories that have the lowest cost while satisfying value preferences, thereby making the best decision among the many behavioral trajectories, reducing the number of options that need to be evaluated, and not only improving the efficiency of the intelligent agent's decision-making, but also being able to flexibly find the best balance between value and cost, further enhancing the adaptability of the intelligent agent in complex environments.

[0033] Furthermore, in some embodiments, the objective function is constructed through the above formula, which not only takes into account the value and cost of the behavior trajectory, but also takes into account the willingness to explore new behavior trajectories. This allows the intelligent agent to not only be limited to the known optimal solution when facing unknown situations, but to try new behavior trajectories by adjusting the exploration willingness in order to find better solutions. This allows the intelligent agent to invest in exploring new behaviors that may bring long-term benefits while pursuing the maximization of immediate value and minimization of costs, which helps the intelligent agent learn and adapt to complex and changing environments in the long run.

[0034] Furthermore, in some embodiments, by controlling the behavior of the intelligent agent according to the target behavior trajectory and observing the behavior results, actual feedback information is provided to the intelligent agent, so that the intelligent agent can use the feedback information to optimize its behavioral capability function, cognitive architecture function and value preference function, thereby realizing closed-loop optimization of the intelligent agent's behavioral decision-making process. Through this closed-loop optimization, the intelligent agent can continue to evolve. This evolution not only stems from the intelligent agent's own practical accumulation, but can also be directly affected by external teaching or information input. During the evolution process, cognitive biases and value judgments can be gradually corrected, realizing the evolutionary process of the intelligent agent from unknown to known, from immature to mature, so as to adapt to the ever-changing environment and task requirements.

[0035] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a flowchart of the behavior decision-making method of an intelligent agent provided in an embodiment of the present application; Figure 2 It is a structural diagram of a behavior decision-making device of an intelligent agent provided in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0039] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0040] As a frontier field of artificial intelligence, intelligent agent technology is gradually penetrating into every aspect of our lives. Intelligent agents refer to intelligent systems designed to perceive the environment, make decisions, and perform actions. Intelligent agents can be in the form of software, such as virtual assistants, recommendation algorithms, etc., or physical entities, such as self-driving cars and industrial robots. The core of intelligent agents lies in their autonomy, that is, the ability to independently control their behavior and internal state without direct human intervention.

[0041] The intelligent agent system can usually include three parts: perception module, decision module and execution module. The perception module is responsible for collecting environmental information, the decision module uses artificial intelligence algorithms to analyze and model the collected information and make corresponding decisions, and the execution module is responsible for executing these decisions and taking actual actions. This layered architecture enables the intelligent agent system to complete various complex tasks flexibly and efficiently.

[0042] In related technologies, most intelligent agent systems are based on traditional behavior models and target optimization algorithms, such as Q-learning, Deep Reinforcement Learning (DRL), etc. These methods usually rely on clear reward signals and optimization goals to guide the agent to choose the optimal behavior. However, when faced with complex, dynamic and incomplete information environments, these methods do not consider how to simulate the agent's cognitive biases, value preferences, and autonomous exploration behaviors and evolution laws based on them. When the agents based on these methods have cognitive biases and lack autonomous exploration and evolution, they often make wrong decisions.

[0043] Based on this, the embodiments of the present application provide a behavior decision-making method, device and electronic device for an intelligent agent. The behavior set determined by the behavior capability function enables the intelligent agent to consider actions within its capability, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated through the cognitive architecture function. Taking into account the current cognition of the intelligent agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The introduction of a value preference function can evaluate the preferred value of the intelligent agent for different behavior trajectories. The intelligent agent's behavioral capability, cognitive architecture and value preference are comprehensively considered, so that the intelligent agent can select a target behavior trajectory from multiple possible behavior trajectories that is consistent with its capability and maximizes value and minimizes cost, thereby improving the decision-making accuracy and efficiency of the intelligent agent.

[0044] The intelligent agent behavior decision-making method, device and electronic device provided in the embodiments of the present application can be applied in various fields, such as smart home, security monitoring, smart assistant, automatic driving and games.

[0045] The following, in conjunction with the accompanying drawings, describes in detail the behavior decision-making method, device and electronic device of the intelligent agent provided in the embodiments of the present application through specific embodiments and their application scenarios.

[0046] The agent's behavior decision method may be applied to a terminal, and may be specifically executed by hardware or software in the terminal.

[0047] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or a tablet computer having a touch-sensitive surface (e.g., a touch screen display and / or a touch pad). It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer having a touch-sensitive surface (e.g., a touch screen display and / or a touch pad).

[0048] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse and a joystick.

[0049] The embodiment of the present application provides a behavior decision method for an intelligent entity. The executor of the behavior decision method for the intelligent entity can be an electronic device or a functional module or functional entity in the electronic device that can implement the behavior decision method for the intelligent entity. The electronic devices mentioned in the embodiment of the present application include but are not limited to mobile phones, tablet computers, computers, cameras, and wearable devices. The behavior decision method for the intelligent entity provided in the embodiment of the present application is described below using an electronic device as an example of the executor.

[0050] like Figure 1 As shown, the behavior decision method of the intelligent agent includes: step 110, step 120, step 130 and step 140.

[0051] Step 110: Determine the behavior set of the agent in the current state according to the behavior capability function of the agent.

[0052] In the embodiments of the present application, an intelligent agent refers to an intelligent agent equipped with artificial intelligence and having the ability to perceive and interact with the outside world, which can interact with the environment in real time through perception and interaction, and can perform various tasks in a virtual environment or a real physical environment. For example, in a virtual environment, an intelligent agent can be a virtual assistant that can provide user consultation and answer questions; in a real physical environment, taking autonomous driving as an example, an intelligent agent can be an unmanned vehicle that can achieve autonomous navigation and safe driving.

[0053] In the embodiment of the present application, the behavior capability function is a mathematical function that describes the set of behaviors that the agent can take in the current state, and defines the behavior options that the agent can take in a specific environment and conditions. , Indicates that the agent is t Status at all times The following set of behaviors.

[0054] The behavioral capability function can be described based on factors such as the agent's perception, decision-making mechanism, and execution capability. For example, the behavioral capability function can be represented by a finite set, where each element in the set represents an optional discrete behavior. For an unmanned vehicle, the behavioral capability function can include a set of behavioral options such as "driving straight", "changing lanes", and "slowing down". Given the current state, a finite set can be output to represent all the behavioral options that the agent can take in this state.

[0055] The behavioral capability function can also be represented by modeling it using a continuous behavioral space. For example, a multidimensional real number vector can be used to represent continuous acceleration, steering angle, etc. The behavioral capability function describes the range of continuous behaviors that the agent can take in the current state. Given the current state, a continuous behavior space can be output to represent the continuous behaviors that the agent can take.

[0056] The behavior capability function can also be represented by a probability distribution model, such as a Gaussian distribution. The probability distribution model defines the probability of occurrence of various possible behaviors of the agent in the current state. Given the current state, a probability distribution can be output to represent the probability of various behavior choices of the agent in this state. Of course, the behavior capability function can also be represented and constructed in other ways, which is not limited in the embodiments of the present application.

[0057] Taking the intelligent agent as an unmanned vehicle as an example, the behavior set may include straight driving, lane keeping, obstacle avoidance, lane changing, parking, acceleration, deceleration, etc. The embodiment of the present application does not limit the behavior set.

[0058] In the embodiment of the present application, the state of the agent can be the internal attributes and external environment information describing the agent at a specific moment. For example, the internal attributes may include position, speed, acceleration, power, internal system state parameters, etc., and the external environment information may include weather, temperature, the position and motion state of other surrounding objects, roads, and traffic lights.

[0059] In the embodiment of the present application, the behavior set of the agent in the current state can be determined according to the behavior capability function and the current state. Taking the agent as an unmanned vehicle as an example, if the vehicle approaches an intersection under a green light, the behavior set may include "accelerate" and "maintain speed". If a pedestrian is detected crossing the road ahead, the behavior set may include "slow down" and "stop".

[0060] Step 120: Determine multiple behavior trajectories of the agent from the current state to the target state according to the behavior set.

[0061] In the embodiments of the present application, the behavior trajectory refers to the path that the agent takes from the current state to the target state through a series of continuous behavior selections. Taking the agent as an unmanned vehicle as an example, the current state can be the current position of the vehicle, the target state can be the target position that the vehicle needs to reach, and the behavior trajectory can be the path that the vehicle plans to reach the target position from the current position. According to the different behaviors of the vehicle, the path from the current position to the target position can include multiple paths, that is, different behavior combinations in the behavior set can determine different behavior trajectories, so that the agent can transition from the current state to the target state.

[0062] In the embodiment of the present application, after determining the behavior set, multiple behavior trajectories of the agent from the current state to the target state can be determined according to different behavior combinations in the behavior set. Taking the agent as an unmanned vehicle as an example, the current state can be the current position of the vehicle, the target state can be the target position that the vehicle needs to reach, and the straight line behavior at the current position can determine the path A that can reach the target position, and the right turn behavior at the current position can determine the path B that can reach the target position.

[0063] Step 130: Calculate the costs corresponding to the multiple behavior trajectories based on the current cognition of the intelligent agent according to the cognitive architecture function, and calculate the values ​​corresponding to the multiple behavior trajectories based on the current preference of the intelligent agent according to the value preference function.

[0064] In the embodiment of the present application, the cognitive architecture function is a mathematical model that describes the perception and decision-making mechanism of the intelligent agent, defines how the intelligent agent perceives and understands based on the current state information, and evaluates the difficulty of the intelligent agent transitioning from the current state to the target state. , which can be used to evaluate the agent from state Transition to state When the agent lacks experience in a certain situation, even if there is a feasible action plan, it may think that the difficulty is too high (or even infinite) due to insufficient cognition, which means that the agent's current cognition believes that it does not have the behavioral ability to support state transition in this situation.

[0065] The cognitive architecture function can realize the cognition of the intelligent agent according to the behavioral ability function. For example, according to the behavioral ability function, it can be determined that the maximum speed of the intelligent agent in the current state is 120km / h. If the distance for the intelligent agent to transition from the current state to the target state is 120km, then it will take at least 1h. If the intelligent agent needs to transition from the current state to the target state within 1.5h, the cognitive architecture function can evaluate that the difficulty is low. If the intelligent agent needs to transition from the current state to the target state within 0.5h, the cognitive architecture function can evaluate that the difficulty is close to infinity, because in the current cognition of the intelligent agent, the maximum speed of the intelligent agent is 120km / h, and it is almost impossible to travel 120km within 0.5h.

[0066] In some embodiments, the cognitive architecture function can be a difficulty assessment model that can output the difficulty of state transfer based on the current state and behavioral capability function. The cognitive architecture function can also take the form of a probability distribution function to output the probability distribution of the agent successfully completing the state transfer under the current state and behavioral capability. It is also possible to use a set of behaviors under different states to train a machine learning model as a cognitive architecture function. The machine learning model can be a neural network, a decision tree, etc., which inputs the current state and the behavioral capability set corresponding to the current state, and outputs the difficulty of state transfer. Of course, the cognitive architecture function can also combine the difficulty assessment model and probability distribution, or introduce expert knowledge, etc. The embodiment of the present application does not limit the cognitive architecture function.

[0067] In the embodiment of the present application, the value preference function is a function used by the agent to evaluate the value of different behavior trajectories, reflecting the value preferences and goals of the agent. Value judgments can come from endogenous factors (such as endocrine and nervous systems determined by genes) or external factors (such as acquired education). If a higher value is given to unexplored trajectory paths in the environment, the agent will be more inclined to try and expand, thereby stimulating innovation and adaptability under new conditions; if a lower value is given to unknown trajectory paths, the agent will avoid trying unknown things and focus on known and robust directions.

[0068] In some embodiments, the preferences of the agent can be understood based on the agent's attributes, historical behaviors, etc., so as to construct a value preference function. Taking the agent as an unmanned vehicle as an example, the behavior trajectory of the agent from the current state to the target state determined according to the behavior set may include behavior trajectory 1 and behavior trajectory 2, wherein the path of behavior trajectory 1 is shorter, but the road is narrower, and the driving risk is higher, and the path of behavior trajectory 2 is longer, but the road is wider, and the driving risk is lower. If the agent prefers safety, the value preference function will assign a lower value to behavior trajectory 1 and a higher value to behavior trajectory 2; if the agent prefers speed, the value preference function will assign a higher value to behavior trajectory 1 and a lower value to behavior trajectory 2.

[0069] In some embodiments, the value preference function can be a utility function, in which the preferences of the agent can be quantified as numerical representations. For example, different weights can be assigned to attributes such as time, cost, and risk. These weights can reflect the preferences of the agent. Different preferences can set different weight values ​​for these attributes. For example, if the agent attaches more importance to time efficiency, the weight of the time attribute in the utility function will be higher. The value preference function can also be a multi-objective optimization model, which involves the weighing of multiple objectives, such as cost, efficiency, safety, etc., and weighs the cost, efficiency, safety, etc. of different behavior trajectories to evaluate the value of different behavior trajectories. The value preference function can also be a probability distribution model, which can evaluate the expected value of the agent taking different behavior trajectories in the current state according to the preferences of the agent. Of course, it is also possible to train a machine learning model as a value preference function, such as a neural network or a decision tree, and predict the value of different behavior trajectories based on the historical data and preferences of the agent. The embodiment of the present application does not limit the value preference function.

[0070] Step 140: Taking maximizing value and minimizing cost as optimization goals, a target behavior trajectory that meets the optimization goals is selected from multiple behavior trajectories.

[0071] In the embodiment of the present application, the higher the cost of the behavior trajectory, the more loss the agent will suffer when transitioning from the current state to the target state through the behavior trajectory. For example, the path of behavior trajectory 1 is shorter, and the amount of oil or electricity lost may be lower, so the cost may be smaller. The higher the value of the behavior trajectory, the more the agent prefers to adopt the behavior trajectory to transition from the current state to the target state. For example, the path of behavior trajectory 1 is shorter but the risk is higher. If the agent prefers safety, the value of behavior trajectory 1 will be lower. In this way, when determining the target behavior trajectory to be selected, it is necessary to use the behavior trajectory with the lowest cost and the highest value as possible as the target behavior trajectory.

[0072] For example, the value of each behavior trajectory can be subtracted from the cost to obtain the score of each behavior trajectory, and the behavior trajectory with the highest score can be determined as the target behavior trajectory that meets the optimization goal. After selecting the target behavior trajectory, a behavior action can be selected from the behavior set to control the actual behavior of the agent, so that the agent transitions from the current state to the target state.

[0073] According to the behavior decision method of the agent of the present application, the behavior set of the agent in the current state is determined according to the behavior capability function of the agent; multiple behavior trajectories of the agent from the current state to the target state are determined according to the behavior set; the costs corresponding to the multiple behavior trajectories are calculated according to the cognitive architecture function based on the current cognition of the agent, and the values ​​corresponding to the multiple behavior trajectories are calculated according to the value preference function based on the current preference of the agent; the target behavior trajectory that meets the optimization goal is selected from multiple behavior trajectories with maximizing value and minimizing cost as the optimization goal. The behavior set determined by the behavior capability function in the embodiment of the present application enables the agent to consider actions within its capabilities, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated by the cognitive architecture function. Considering the current cognition of the agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The value preference function can be introduced to evaluate the preference value of the agent for different behavior trajectories. The behavior capability, cognitive architecture and value preference of the agent are comprehensively considered, so that the agent can select the target behavior trajectory that is consistent with its capabilities and maximizes value and minimizes cost from multiple possible behavior trajectories, thereby improving the accuracy and efficiency of the decision-making of the agent.

[0074] In some embodiments, the costs corresponding to the plurality of behavior trajectories are calculated based on the current cognition of the agent according to the cognitive architecture function, including: According to the formula

[0075] Calculate the costs corresponding to multiple behavior trajectories; in, Indicates that the agent is in the current state To the target state A behavioral trajectory, Indicates behavior trajectory The corresponding cost, and Indicates different times, is the behavioral capability function, which indicates that the agent t Status at all times The following set of behaviors, is a cognitive architecture function, which is used to evaluate the agent from state Transition to state difficulty.

[0076] In this embodiment, if Smaller, , that is, it is believed that the agent changes from the current state through this behavior trajectory To the target state The difficulty is low; if you encounter infinite difficulty, that is, an unfeasible behavior trajectory, then .

[0077] In this embodiment, the current cognitive state of the agent is taken into account through the above-mentioned cost calculation formula, and the difficulty of the agent's state transition can be evaluated in combination with the agent's behavioral capabilities and the dynamic changes of the environment, so that the potential costs of different behavioral trajectories can be predicted based on current cognition.

[0078] In some embodiments, maximizing value and minimizing cost are optimization goals, and a target behavior trajectory that meets the optimization goals is selected from multiple behavior trajectories, including: Taking maximizing value and minimizing cost as the optimization goal, the objective function is constructed according to the cognitive architecture function and the value preference function; The costs corresponding to the multiple behavior trajectories and the values ​​corresponding to the multiple behavior trajectories are substituted into the objective function, and a target behavior trajectory that meets the optimization goal is selected from the multiple behavior trajectories.

[0079] In this embodiment, the formula

[0080] Construct the objective function; in, is the value preference function, which represents the agent's behavior trajectory The value given.

[0081] Substitute the costs and values ​​corresponding to multiple behavior trajectories into the objective function, calculate the scores of different behavior trajectories under the objective function, and select the trajectory with the highest score from multiple behavior trajectories as the target behavior trajectory.

[0082] In this embodiment, by constructing the objective function through the above formula, it is possible to clearly identify those behavioral trajectories that satisfy value preferences while having the lowest cost, thereby making the best decision among the many behavioral trajectories, reducing the number of options that need to be evaluated, and not only improving the efficiency of the intelligent agent's decision-making, but also being able to flexibly find the best balance between value and cost, further enhancing the adaptability of the intelligent agent in complex environments.

[0083] In this embodiment, an objective function that comprehensively considers value maximization and cost minimization is constructed, and the cognitive architecture function and value preference function are utilized, so that the objective function can quantify the expected value and execution cost of each behavior trajectory. The cost and value of each behavior trajectory are substituted into the objective function. By comparing the scores of different trajectories, the optimal trajectory that maximizes value while minimizing cost can be selected, thereby improving the efficiency of the intelligent agent's decision-making.

[0084] In some embodiments, maximizing value and minimizing cost are optimization goals, and an objective function is constructed according to the cognitive architecture function and the value preference function, including: According to the formula

[0085] Construct the objective function; in, is the intention probability function, which represents the behavior trajectory attempted by the agent willingness to explore.

[0086] In this embodiment, an intention probability function is introduced to construct the objective function. The intention probability function is used to describe the subjective tendency of the agent to try a specific trajectory. Even if the agent is not sure whether the behavior trajectory can be achieved in the real environment, the intention probability function can also reflect the agent's willingness to explore.

[0087] The intention probability function can be used to calculate the probability values ​​of different behavior trajectories. The probability value reflects the agent's willingness to explore different behavior trajectories. The higher the probability value, the stronger the willingness to explore.

[0088] In this embodiment, the objective function is constructed by the above formula, which not only takes into account the value and cost of the behavior trajectory, but also takes into account the willingness to explore new behavior trajectories. This ensures that when facing unknown situations, the intelligent agent is not limited to the known optimal solution, but adjusts the exploration willingness and tries new behavior trajectories in order to find better solutions. This ensures that while pursuing the maximization of immediate value and the minimization of costs, the intelligent agent also invests in exploring new behaviors that may bring long-term benefits, which helps the intelligent agent learn and adapt to complex and changing environments in the long run.

[0089] In some embodiments, the method further comprises: Control the behavior of the agent according to the target behavior trajectory and determine the behavior results of the agent; Optimize behavioral ability function, cognitive architecture function and value preference function according to behavioral results.

[0090] In this embodiment, according to the determined target behavior trajectory, a behavior action can be selected from the behavior set to control the actual behavior of the agent, and the behavior result of the agent can be recorded during the process of the agent performing the behavior action. The behavior result can include information such as the internal changes of the agent and whether it has normally transitioned to the target state.

[0091] For example, based on the behavior results, the maximum speed of the agent can be determined. If the behavior set indicates that the maximum speed of the agent is 120km / h, and the behavior results show that the maximum speed of the agent reaches 125km / h, it can be said that the agent's behavior ability has changed, that is, the behavior ability function has expanded, and the behavior ability function can be optimized. If the agent does not transition to the target state normally, cognitive bias may occur. The cost of the target behavior trajectory is increased to update the cognitive architecture function, and the value of the target behavior trajectory is reduced to update the value preference function.

[0092] In some embodiments, optimizing the behavior capability function, cognitive architecture function, and value preference function according to the behavior results includes: According to the formula

[0093] Or the formula

[0094] Optimize behavioral capability function, cognitive architecture function and value preference function; in, represents the optimization function, are the parameters of the intention probability function, is the parameter of the value preference function, are the parameters of the cognitive architecture function, is the parameter of the behavioral capability function. , , , , Subscript Indicates that during the optimization process, the corresponding parameters Changes can occur.

[0095] In this embodiment, given the behavioral capability function, cognitive architecture function, and value preference function, the intention probability function can be determined by optimizing the objective.

[0096] The optimization goal of the above formula comprehensively considers the value and cost of the behavior trajectory. During the optimization process, a higher intention probability will be assigned to those behavior trajectories that achieve a better balance between value and cost, making it easier for the intelligent agent to adapt to complex and changing environments and improve decision-making efficiency in response to environmental changes.

[0097] In this embodiment, the expansion of the behavioral capability function represents the continuous expansion of the capabilities of the agent. As the agent learns or is given new skills in new situations, more potential explorable behavioral trajectories will be added, enabling the agent to achieve new state transitions and more diverse choice spaces. At this point, the newly added behavioral trajectories are in an unknown area, and their true feasibility and difficulty are still unknown. This situation provides an opportunity for the update of the cognitive architecture function and the value preference function, and it is necessary to conduct subjective cognition and value evaluation of the newly added trajectories, thereby laying the foundation for subsequent decision-making.

[0098] When faced with an untried trajectory, the cognitive architecture function needs to estimate the difficulty of the path. Due to lack of experience, the agent may give an inaccurate difficulty judgment of the unknown behavior trajectory. Optimists may regard the difficulty as low, so they are motivated to try; pessimists may think that the difficulty is too high and avoid trying. Regardless of the initial judgment, when the agent begins to explore these behavior trajectories, it will collect experience, feedback, and lessons from failure in practice. As information accumulates, the evaluation results of the cognitive architecture function will be gradually corrected and refined. Through continuous iteration, the agent's difficulty perception system will be closer to the actual situation, thereby reducing unnecessary attempts or unnecessary conservatism caused by cognitive biases.

[0099] The value preference function is also updated during the evolution process. When the agent successfully experiences or tries a certain behavior trajectory, it will not only deepen its understanding of the difficulty, but also gain a true sense of the value of the behavior trajectory. If a new behavior trajectory is found to be of high value after exploration, the value preference function will give a higher preference to this type of behavior trajectory; if a behavior trajectory is of little value or even harmful to the agent, the value preference function will lower its preference value. By dynamically updating the value preference, the agent can more easily identify paths that are of long-term benefit to itself in future decisions.

[0100] It should be noted that this evolution not only comes from the agent's own practical accumulation, but can also be directly affected by external teaching or information input. When external sources directly give the agent new cognition or value preferences, the agent's cognition and strategy preferences will change in an instant. However, the accuracy of this external empowerment does not always match the actual situation. The agent will still re-evaluate and correct the external input in subsequent practice to ensure the internal consistency of its own cognitive system and the rationality of the evolutionary direction.

[0101] In this embodiment, the behavior of the intelligent agent is controlled according to the target behavior trajectory and the behavior results are observed, thereby providing the intelligent agent with actual feedback information, allowing the intelligent agent to use the feedback information to optimize its behavioral capability function, cognitive architecture function and value preference function, thereby achieving closed-loop optimization of the intelligent agent's behavioral decision-making process. Through this closed-loop optimization, the intelligent agent can continuously evolve. This evolution not only stems from the intelligent agent's own practical accumulation, but can also be directly affected by external teaching or information input. During the evolution process, cognitive biases and value judgments can be gradually corrected, thereby achieving the intelligent agent's evolutionary process from unknown to known, from immature to mature, so as to adapt to the ever-changing environment and task requirements.

[0102] The agent behavior decision method provided in the embodiment of the present application can be executed by an agent behavior decision device. In the embodiment of the present application, the agent behavior decision device is used as an example to illustrate the agent behavior decision method provided in the embodiment of the present application.

[0103] The embodiment of the present application also provides a behavior decision-making device for an intelligent agent.

[0104] like Figure 2 As shown, the behavior decision device of the intelligent agent includes: A first determination module 210, for determining a behavior set of the agent in a current state according to the behavior capability function of the agent; A second determination module 220 is used to determine multiple behavior trajectories of the agent from the current state to the target state according to the behavior set; A calculation module 230, for calculating the costs corresponding to the plurality of behavior trajectories based on the current cognition of the agent according to the cognitive architecture function, and calculating the values ​​corresponding to the plurality of behavior trajectories based on the current preference of the agent according to the value preference function; The selection module 240 is used to select a target behavior trajectory that meets the optimization goal from multiple behavior trajectories, taking maximizing value and minimizing cost as the optimization goal.

[0105] According to the behavior decision device of the intelligent agent of the present application, the behavior set of the intelligent agent in the current state is determined according to the behavior capability function of the intelligent agent; multiple behavior trajectories of the intelligent agent from the current state to the target state are determined according to the behavior set; the costs corresponding to the multiple behavior trajectories are calculated according to the cognitive architecture function based on the current cognition of the intelligent agent, and the values ​​corresponding to the multiple behavior trajectories are calculated according to the value preference function based on the current preference of the intelligent agent; the target behavior trajectory that meets the optimization goal is selected from multiple behavior trajectories with maximizing value and minimizing cost as the optimization goal. The behavior set determined by the behavior capability function of the embodiment of the present application enables the intelligent agent to consider actions within its ability range, thereby improving the feasibility of decision-making. The cost of the behavior trajectory is calculated by the cognitive architecture function. Considering the current cognition of the intelligent agent, the cost consumed by different behavior trajectories can be evaluated, and high-cost decisions can be reduced. The value preference function is introduced to evaluate the preferred value of the intelligent agent for different behavior trajectories. The behavior capability, cognitive architecture and value preference of the intelligent agent are comprehensively considered, so that the intelligent agent can select the target behavior trajectory that is both in line with its ability and can maximize value and minimize cost from multiple possible behavior trajectories, thereby improving the decision-making accuracy and efficiency of the intelligent agent.

[0106] In some embodiments, the computing module 230 is further configured to: According to the formula

[0107] Calculate the costs corresponding to multiple behavior trajectories; in, Indicates that the agent is in the current state To the target state A behavioral trajectory, Indicates behavior trajectory The corresponding cost, and Indicates different times, is the behavioral capability function, which indicates that the agent t Status at all times The following set of behaviors, is a cognitive architecture function, which is used to evaluate the agent from state Transition to state difficulty.

[0108] In some embodiments, the selection module 240 is further configured to: Taking maximizing value and minimizing cost as the optimization goal, the objective function is constructed according to the cognitive architecture function and the value preference function; The costs corresponding to the multiple behavior trajectories and the values ​​corresponding to the multiple behavior trajectories are substituted into the objective function, and a target behavior trajectory that meets the optimization goal is selected from the multiple behavior trajectories.

[0109] In some embodiments, the selection module 240 is further configured to: According to the formula

[0110] Construct the objective function; in, is the value preference function, which represents the agent's behavior trajectory The value given.

[0111] In some embodiments, the selection module 240 is further configured to: According to the formula

[0112] Construct the objective function; in, is the intention probability function, which represents the behavior trajectory attempted by the agent willingness to explore.

[0113] In this embodiment, the objective function is constructed by the above formula, which not only takes into account the value and cost of the behavior trajectory, but also takes into account the willingness to explore new behavior trajectories. This ensures that when facing unknown situations, the intelligent agent is not limited to the known optimal solution, but adjusts the exploration willingness and tries new behavior trajectories in order to find better solutions. This ensures that while pursuing the maximization of immediate value and the minimization of costs, the intelligent agent also invests in exploring new behaviors that may bring long-term benefits, which helps the intelligent agent learn and adapt to complex and changing environments in the long run.

[0114] In some embodiments, the selection module 240 is further configured to: Control the behavior of the agent according to the target behavior trajectory and determine the behavior results of the agent; Optimize behavioral ability function, cognitive architecture function and value preference function according to behavioral results.

[0115] In this embodiment, the behavior of the intelligent agent is controlled according to the target behavior trajectory and the behavior results are observed, thereby providing the intelligent agent with actual feedback information, allowing the intelligent agent to use the feedback information to optimize its behavioral capability function, cognitive architecture function and value preference function, thereby achieving closed-loop optimization of the intelligent agent's behavioral decision-making process. Through this closed-loop optimization, the intelligent agent can continuously evolve. This evolution not only stems from the intelligent agent's own practical accumulation, but can also be directly affected by external teaching or information input. During the evolution process, cognitive biases and value judgments can be gradually corrected, thereby achieving the intelligent agent's evolutionary process from unknown to known, from immature to mature, so as to adapt to the ever-changing environment and task requirements.

[0116] In some embodiments, the selection module 240 is further configured to: According to the formula

[0117] Or the formula

[0118] Optimize behavioral capability function, cognitive architecture function and value preference function; in, represents the optimization function, are the parameters of the intention probability function, is the parameter of the value preference function, are the parameters of the cognitive architecture function, Parameters of the behavior capability function.

[0119] The behavior decision device of the intelligent body in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or it can be other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0120] The behavior decision device of the intelligent agent in the embodiment of the present application may be a device having an operating system. The operating system may be a Microsoft (Windows) operating system, an Android (Android) operating system, an IOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0121] In some embodiments, Figure 3As shown, an embodiment of the present application also provides an electronic device 300, including a processor 301, a memory 302, and a computer program stored in the memory 302 and executable on the processor 301. When the program is executed by the processor 301, each process of the above-mentioned intelligent agent behavior decision method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0122] It should be noted that the electronic devices in the embodiments of the present application include the above-mentioned mobile electronic devices and non-mobile electronic devices.

[0123] An embodiment of the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned intelligent agent behavior decision-making method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0124] The processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0125] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above-mentioned behavior decision-making method of the intelligent agent when executed by a processor.

[0126] The processor is the processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0127] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned intelligent agent behavior decision-making method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0128] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0129] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0130] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0131] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0132] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0133] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A behavior decision method for an intelligent agent, characterized in that: include: Determine a behavior set of the agent in the current state according to the behavior capability function of the agent; Determine multiple behavior trajectories of the agent from a current state to a target state according to the behavior set; Calculating the costs corresponding to the plurality of behavior trajectories based on the current cognition of the agent according to the cognitive architecture function, and calculating the values ​​corresponding to the plurality of behavior trajectories based on the current preference of the agent according to the value preference function; Taking maximizing value and minimizing cost as optimization goals, a target behavior trajectory that meets the optimization goals is selected from the multiple behavior trajectories.

2. The method according to claim 1, characterized in that The calculating the costs corresponding to the plurality of behavior trajectories based on the current cognition of the agent according to the cognitive architecture function includes: According to the formula Calculating costs corresponding to a plurality of the behavior trajectories; in, Indicates that the agent is in the current state To the target state A behavioral trajectory, Indicates behavior trajectory The corresponding cost, and Indicates different times, is the behavioral capability function, which indicates that the agent t Status at all times The following set of behaviors, is a cognitive architecture function, which is used to evaluate the agent from state Transition to state difficulty.

3. The method according to claim 1, characterized in that The optimization goal is to maximize value and minimize cost, and select a target behavior trajectory that meets the optimization goal from multiple behavior trajectories, including: Taking maximizing value and minimizing cost as optimization goals, constructing an objective function according to the cognitive architecture function and the value preference function; Substituting the costs corresponding to the plurality of behavior trajectories and the values ​​corresponding to the plurality of behavior trajectories into the objective function, and selecting a target behavior trajectory that meets the optimization objective from the plurality of behavior trajectories.

4. The method according to claim 3, characterized in that The optimization goal of maximizing value and minimizing cost is to construct an objective function according to the cognitive architecture function and the value preference function, including: According to the formula Construct the objective function; in, is the value preference function, which represents the agent's behavior trajectory The value given.

5. The method according to claim 3, characterized in that: The optimization goal of maximizing value and minimizing cost is to construct an objective function according to the cognitive architecture function and the value preference function, including: According to the formula Construct the objective function; in, is the intention probability function, which represents the behavior trajectory attempted by the agent willingness to explore.

6. The method according to claim 1, characterized in that The method further comprises: Controlling the behavior of the agent according to the target behavior trajectory and determining the behavior result of the agent; The behavioral ability function, the cognitive architecture function and the value preference function are optimized according to the behavioral results.

7. The method according to claim 6, characterized in that The optimizing the behavior capability function, the cognitive architecture function and the value preference function according to the behavior result includes: According to the formula Or the formula Optimizing the behavioral capability function, the cognitive architecture function, and the value preference function; in, represents the optimization function, are the parameters of the intention probability function, is the parameter of the value preference function, are the parameters of the cognitive architecture function, Parameters of the behavior capability function.

8. A behavior decision-making device for an intelligent agent, characterized in that: include: A first determination module, used to determine a behavior set of the agent in a current state according to the behavior capability function of the agent; A second determination module is used to determine multiple behavior trajectories of the agent from a current state to a target state according to the behavior set; A calculation module, used to calculate the costs corresponding to the plurality of behavior trajectories based on the current cognition of the intelligent agent according to the cognitive architecture function, and to calculate the values ​​corresponding to the plurality of behavior trajectories based on the current preference of the intelligent agent according to the value preference function; The selection module is used to take maximizing value and minimizing cost as optimization goals, and select a target behavior trajectory that meets the optimization goals from multiple behavior trajectories.

9. 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, the method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images