WO2026025751A2 - 一种异构移动智能体协同控制与演化方法一种异构移动智能体协同控制与演化方法 - Google Patents

一种异构移动智能体协同控制与演化方法一种异构移动智能体协同控制与演化方法

Info

Publication number
WO2026025751A2
WO2026025751A2 PCT/CN2024/136661 CN2024136661W WO2026025751A2 WO 2026025751 A2 WO2026025751 A2 WO 2026025751A2 CN 2024136661 W CN2024136661 W CN 2024136661W WO 2026025751 A2 WO2026025751 A2 WO 2026025751A2
Authority
WO
WIPO (PCT)
Prior art keywords
sub
vehicle
emotional
emotion
virtual
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2024/136661
Other languages
English (en)
French (fr)
Inventor
丁飞
都秉甲
宋爱国
陈竺
束良军
童恩
朱洪波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Publication of WO2026025751A2 publication Critical patent/WO2026025751A2/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Program-control systems
    • G05B19/02—Program-control systems electric
    • G05B19/04—Program control other than numerical control, i.e. in sequence controllers or logic controllers
    • G05B19/042—Program control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
    • G05B19/0423—Input/output
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/20—Pc systems
    • G05B2219/25—Pc structure of the system
    • G05B2219/25257—Microcontroller

Definitions

  • This invention relates to a method for cooperative control and evolution of heterogeneous mobile intelligent agents, belonging to the field of mobile intelligent agent formation and cooperative control technology.
  • the technical problem to be solved by this invention is that the existing technology does not fully consider the influence of individual emotional factors of artificial mobile intelligent agents and uncertain factors in the local environment on the collaborative control of multiple agents, resulting in the collaborative control and evolution of heterogeneous mobile intelligent agent systems failing to approximate the real situation.
  • the present invention is implemented using the following technical solution.
  • This invention provides a method for cooperative control and evolution of heterogeneous mobile intelligent agents, comprising:
  • the artificial mobile agent cells and the automatic mobile agent cells are mixed in a preset ratio using a continuous cellular automaton, and artificial mobile agents and automatic mobile agents are generated through an iterative process.
  • An emotion recognition module is embedded in the artificial mobile intelligent agent to obtain emotion cues of different granularities based on fuzzy theory.
  • the evolution rules of the artificial mobile agent are constructed, and the state of the artificial mobile agent is updated.
  • the evolution rules of the artificial mobile agent include acceleration rules, deceleration rules, random slowing rules that introduce a random slowing probability function, and position update rules;
  • the position, velocity, and acceleration of the autonomous mobile agent are obtained. Based on the improved interaction potential energy field function, the evolution rules of the autonomous mobile agent are constructed, and the state of the autonomous mobile agent is updated.
  • the evolution rules for the autonomous mobile intelligent agent include acceleration rules, deceleration rules, and position update rules
  • the acceleration rules are used to adjust the speed of automated or artificial mobile agents.
  • the deceleration rule is used to avoid collisions between automated mobile agents and artificial mobile agents
  • the random slowing rule which introduces a random slowing probability function, is used to randomly slow down an artificial mobile agent based on an emotion influence factor and a random slowing probability function.
  • the location update rule is used to determine the location of an automated or manually operated mobile agent.
  • the movement and path of the automated mobile agent are controlled by the field force planning method.
  • the movement path of the artificial mobile agent is planned by combining the movement and path of the automated mobile agent with emotional cues of different granularities in the operation scenario of the artificial mobile agent, and utilizing the perception information sharing mechanism of the automated mobile agent.
  • the emotion cognition module includes an emotion perception layer and a feature fusion network.
  • the emotion perception layer adopts a dual-channel fusion attention mechanism network structure.
  • methods for obtaining emotion cues of different granularities include:
  • the emotion perception layer is used to capture and extract character emotion cues and scene emotion cues in the running scene through the character channel and scene channel respectively, forming a preliminary feature vector;
  • the emotion-influencing factors are input into the feature fusion network for fusion processing to obtain emotion cues at different granularities.
  • a method for quantifying the emotion values of continuous dimensions in the initial feature vector using fuzzy theory to obtain emotion influencing factors includes:
  • a three-valued input and single-valued output fuzzy inference model of emotion is constructed, with emotional pleasure, emotional arousal and emotional dominance as input variables and emotional influence factors as output variables.
  • fuzzy sets of input variables are constructed respectively, including fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional pleasure, fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional arousal, and fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional dominance;
  • Construct fuzzy sets of output variables including fuzzy set 1 of emotion factors, fuzzy set 2 of emotion factors, and fuzzy set 3 of emotion factors;
  • the fuzzy sets of input and output variables are fuzzified using Gaussian membership functions.
  • a fuzzy rule matrix for emotion representation is constructed.
  • the continuous sentiment values in the initial feature vector are converted into the membership degrees of the fuzzy set of the input set through fuzzification.
  • a fuzzy inference engine is used to map the input fuzzy set to the output fuzzy set, and calculations are performed to obtain the final output fuzzy set.
  • centroid method is used to defuzzify the final output fuzzy set to obtain the emotion influence factor.
  • fuzzy rule matrix for emotion representation is constructed.
  • the constructed fuzzy rule matrix for emotion representation is represented as follows:
  • the rule in the first row of the emotion representation fuzzy rule matrix is: when the input emotion pleasure degree belongs to the fuzzy set of emotion pleasure degree one, the input emotion arousal degree belongs to the fuzzy set of emotion arousal degree three, and the input emotion dominance degree belongs to the fuzzy set of emotion dominance degree three, the output emotion factor fuzzy set one is output.
  • the rule weight in the first row is 1, and the logical connector is "AND".
  • the rule in the second row of the fuzzy rule matrix is: when the input emotional pleasure degree belongs to the second fuzzy set of emotional pleasure degree, the input emotional arousal degree belongs to the first fuzzy set of emotional arousal degree, and the input emotional dominance degree belongs to the third fuzzy set of emotional dominance degree, the output emotional factor fuzzy set 2 is given.
  • the rule weight in the second row is 1, and the logical connector is "AND".
  • the rule in the third row of the fuzzy rule matrix is: when the input emotional pleasure degree belongs to the first fuzzy set of emotional pleasure degree, the input emotional arousal degree belongs to the second fuzzy set of emotional arousal degree, and the input emotional dominance degree belongs to the first fuzzy set of emotional dominance degree, the output emotional factor fuzzy set three is given.
  • the rule weight in the third row is 1, and the logical connector is "AND".
  • the continuous cellular automaton uses vehicles as cells and generates manually driven vehicles and autonomous connected vehicles through an iterative process.
  • the state information ⁇ ⁇ sub>i ⁇ /sub> (t) of the continuous cellular automaton is expressed as:
  • x ⁇ sub>n ⁇ /sub> (t), ⁇ ⁇ sub>n ⁇ /sub> (t), and a ⁇ sub>n ⁇ /sub> (t) represent the position, velocity, and acceleration of vehicle n at the current time t, respectively;
  • E ⁇ sub>n ⁇ /sub> represents the ideal state vector of vehicle n;
  • AFF ⁇ sub> n ⁇ /sub> represents the driver's emotional state vector of vehicle n.
  • V (t), AFFn_A(t), and AFFn_D(t) represent the expected speed of vehicle n, the expected following distance of vehicle n, and the safe time interval of vehicle n, respectively.
  • AFFn_V(t), AFFn_A (t), and AFFn_D (t) represent the measures of the driver's emotional pleasure, emotional arousal, and emotional dominance at the current time t, respectively.
  • the continuous cellular automaton uses vehicles as cells and generates manually driven vehicles and autonomous connected vehicles through an iterative process.
  • the manually driven vehicle a ⁇ sub>n ⁇ /sub> (t) is represented as:
  • a ⁇ sub>0 ⁇ /sub> represents the maximum acceleration of the vehicle
  • ⁇ represents the vehicle acceleration exponent
  • s ⁇ sub>n ⁇ /sub> (t) and ⁇ (t) represent the relative distance and relative speed between the preceding vehicle n+1 and the current vehicle n, respectively
  • B represents the absolute value of the vehicle's comfortable deceleration
  • S ⁇ sub>0 ⁇ /sub> represents the stationary safety distance
  • v ⁇ sub>n+1 ⁇ /sub> (t) represents the speed of the preceding vehicle n+1 at the current time t.
  • the centroid method is used to defuzzify the final output fuzzy set to obtain the emotion influence factor.
  • AFF n_V (t), AFF n_A (t), and AFF n_D (t) represent the measures of the driver's emotional pleasure, emotional arousal, and emotional dominance at the current time t, respectively.
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the speed of vehicle n at the current time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the speed of vehicle n at the current time t
  • a ⁇ sub>n ⁇ /sub>(t) represents the acceleration of vehicle n at the current time t
  • v ⁇ sub>max ⁇ /sub> represents the maximum speed limit of the road.
  • d_safe represents the safe following distance
  • s_n (t) represents the relative distance between the preceding vehicle n+1 and the current vehicle n
  • s_0 represents the stationary safe distance
  • the driver of vehicle n has a probability of p ⁇ sub>n ⁇ /sub> (t+1) of randomly slowing down, where p ⁇ sub>n ⁇ /sub>(t+1) represents the random slowing probability of vehicle n.
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the speed of vehicle n at the current time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the speed of vehicle n at the current time t
  • c represents the deceleration value
  • the position x ⁇ sub>n ⁇ /sub> (t) and velocity v ⁇ sub>n ⁇ /sub> (t) of car n at the current time t together determine the position x ⁇ sub>n ⁇ /sub> (t+1) of car n at the next time t+1;
  • x ⁇ sub>n ⁇ /sub> (t+1) represents the position of car n at the next time step (t+1)
  • x ⁇ sub>n ⁇ /sub>(t) represents the position of car n at the current time step t
  • v ⁇ sub>n ⁇ /sub> (t) represents the speed of car n at the current time step t.
  • the evolution rules of the autonomous mobile intelligent agent include acceleration rules, deceleration rules, and position update rules
  • the acceleration rule for the autonomous mobile agent is expressed as follows:
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the velocity of vehicle n at the current time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the velocity of vehicle n at the current time t
  • v ⁇ sub> max ⁇ /sub> represents the maximum speed limit of the road
  • a represents the acceleration of vehicle n calculated according to the double integral equation of motion
  • m represents the mass of vehicle n
  • d ⁇ sub>safe ⁇ /sub> represents the safe following distance
  • F represents the total virtual force.
  • the deceleration rule for the automatic mobile intelligent agent is: when the repulsive force generated by the virtual potential field of the current vehicle n+1 is greater than the attractive force, the total virtual force on the current vehicle n is manifested as a repulsive force;
  • the location update rules for automated mobile agents are consistent with those for human-made mobile agents.
  • a method for controlling the motion and motion path of an autonomous mobile intelligent agent through field force planning includes:
  • the position, velocity, and acceleration of the autonomous mobile agent are acquired in real time through its sensors, and the interaction potential energy field function is defined.
  • the interaction potential energy field function is improved by introducing virtual forces of position, velocity, and acceleration.
  • the virtual forces from other agents acting on each autonomously moving agent are calculated, including position virtual force, velocity virtual force and acceleration virtual force.
  • the total virtual force of each autonomous mobile agent is obtained by vector synthesis of all virtual forces acting on each autonomous mobile agent.
  • the driving speed and direction of the autonomous mobile intelligent agent are adjusted through a control algorithm.
  • the virtual force of the potential energy field is used to plan the driving path of an autonomous mobile intelligent agent.
  • the position virtual force includes position virtual attraction and position virtual repulsion
  • the velocity virtual force includes velocity virtual attraction and velocity virtual repulsion
  • the acceleration virtual force includes acceleration virtual attraction and acceleration virtual repulsion
  • the virtual repulsive force F rep-d at the location is expressed as:
  • kre-d represents the virtual repulsive potential energy coefficient at the location
  • r represents the distance from the automatically connected vehicle to the target point
  • r0 represents the range of influence of the repulsive potential field. This represents the distance vector from the automatically connected vehicle to the target point. This represents the magnitude of the distance vector from the automatically connected vehicle to the target point.
  • ka_att-d represents the virtual gravitational potential energy coefficient of the location
  • r_g represents the distance from the automatically connected vehicle to the target point.
  • the virtual velocity repulsion Frep -v can be expressed as:
  • kre -v represents the velocity virtual repulsive potential energy coefficient
  • r represents the distance from the automatically connected vehicle to the target point
  • r ⁇ sub>0 ⁇ /sub> represents the range of influence of the repulsive potential field. This represents the speed difference vector between the vehicle in front and the vehicle in front. This represents the magnitude of the velocity difference vector between the vehicle in front and the vehicle itself.
  • ka att-v represents the velocity virtual gravitational potential energy coefficient
  • ⁇ ve represents the velocity difference between the desired velocity and the current velocity
  • rg represents the distance from the automatically connected vehicle to the target point
  • tanh represents the hyperbolic tangent function
  • represents the magnitude of the velocity difference vector between the desired velocity and the current velocity.
  • the virtual repulsive force Frep-a due to acceleration is expressed as:
  • kre -a represents the virtual repulsive potential energy coefficient of acceleration
  • r represents the distance from the automatically connected vehicle to the target point
  • r0 represents the range of influence of the repulsive potential field. This represents the vector of acceleration differences between the vehicle in front and the vehicle in front. This represents the magnitude of the acceleration difference vector between the vehicle in front and the vehicle itself.
  • F att-a tanh(k att-a
  • ka att-a represents the virtual gravitational potential energy coefficient of acceleration
  • ⁇ a e represents the difference between the desired acceleration and the current acceleration
  • r g represents the distance from the automatically connected vehicle to the target point
  • represents the magnitude of the vector of the difference between the desired acceleration and the current acceleration.
  • This invention improves the realism and accuracy of collaborative control of heterogeneous mobile intelligent agents by introducing an emotion-influencing factor, accurately quantifying continuous-dimensional emotion values and integrating them into a random slowing probability function, making the behavior of artificial mobile intelligent agents more realistic.
  • the method based on potential energy field virtual force transforms the position, velocity, and acceleration information of the autonomous mobile intelligent agent into virtual field force, and effectively controls its motion through field force planning, thereby enhancing the autonomous decision-making and path planning capabilities of the autonomous mobile intelligent agent.
  • This invention addresses the challenge of collaborative control in complex environments by significantly improving the adaptability and overall intelligence of collaborative control of heterogeneous mobile intelligent agents through emotion recognition, deceleration rules, and potential energy field optimization.
  • This invention embeds an emotion recognition module into an artificial mobile agent, utilizing fuzzy theory to process emotional cues of different granularities. This enables the agent to perceive and understand emotional information in complex environments.
  • the evolutionary rules constructed based on emotional cues of different granularities make the behavior of the artificial mobile agent more flexible and varied.
  • the introduction of a randomized slowing probability function and an emotion influence factor allows for dynamic adjustment of movement speed according to emotional state. This not only increases the diversity of the agent's behavior but also enhances its survivability and adaptability in complex environments.
  • This invention effectively achieves dynamic collision avoidance and cooperative movement among agents by modeling virtual repulsive and gravitational fields and constructing evolution rules for autonomous mobile agents based on an improved interaction potential energy field function. Furthermore, this invention utilizes field force planning methods to control the movement and path of autonomous mobile agents, and combines a sensory information sharing mechanism and emotional cues to plan more rational and efficient movement paths for artificial mobile agents. This solves the problem in existing technologies where the influence of individual emotional factors of artificial mobile agents and uncertainties in the local environment on the cooperative control of multiple agents is not fully considered, resulting in the cooperative control and evolution of heterogeneous mobile agent systems failing to approximate real-world conditions.
  • Figure 1 is a structural schematic diagram of the heterogeneous mobile intelligent agent provided in an embodiment of the present invention.
  • Figure 2 is a schematic diagram of the construction process of the emotion fuzzy reasoning model provided in an embodiment of the present invention.
  • Figure 3 is a simulation diagram of the virtual force model of the potential energy field in the operation scenario of the heterogeneous mobile intelligent agent cooperative control provided in the embodiment of the present invention.
  • this embodiment introduces a method for cooperative control and evolution of heterogeneous mobile intelligent agents, including:
  • Step 1 Use a continuous cellular automaton to mix artificial and automatic mobile agent cells according to a preset ratio, and generate artificial and automatic mobile agents through an iterative process;
  • Step 2 Embed an emotion recognition module into the artificial mobile intelligent agent to obtain emotion cues of different granularities based on fuzzy theory
  • Emotional cues can affect the movement of artificial mobile agents. Considering emotional cues of different granularities can make the behavioral decisions of artificial mobile agents more in line with the actual situation of humans, improve the realism and accuracy of the simulation model, and introduce a new dimension—emotional factors—for the collaborative control of heterogeneous agents.
  • Step 3 Based on the emotional cues of different granularities, construct the evolution rules of the artificial mobile agent and update the state of the artificial mobile agent;
  • the evolution rules of the artificial mobile agent include acceleration rules, deceleration rules, random slowing rules that introduce a random slowing probability function, and position update rules;
  • Step 4 Construct evolutionary rules for the artificial mobile agent based on emotional cues, including acceleration rules, deceleration rules, random slowing rules, and position update rules, in order to update the state of the artificial mobile agent.
  • the aforementioned rules enable the behavior of artificial mobile agents to dynamically adjust according to emotional changes, increasing the complexity and realism of the simulation environment.
  • the stochastic slowing rule introduces uncertainty, making behavior more unpredictable.
  • Step 5 Obtain the position, velocity, and acceleration of the autonomous mobile agent; construct the evolution rules of the autonomous mobile agent based on the improved interaction potential energy field function; and update the state of the autonomous mobile agent.
  • acceleration rules, deceleration rules, and position update rules of the autonomous mobile agent are constructed to optimize the motion path and obstacle avoidance ability of the autonomous mobile agent, thereby improving the adaptability and safety of the autonomous mobile agent in complex environments and enabling the autonomous mobile agent to cope with various situations more intelligently, such as avoiding pedestrians and adjusting speed.
  • the evolution rules for the autonomous mobile intelligent agent include acceleration rules, deceleration rules, and position update rules
  • the acceleration rules are used to adjust the speed of automated or artificial mobile agents.
  • the deceleration rule is used to avoid collisions between automated mobile agents and artificial mobile agents.
  • Acceleration rules are used to adjust the speed of the agent to match its motion requirements, while deceleration rules are used to avoid collisions and ensure safety. Together, they maintain the stability and safety of the simulation environment, reduce the occurrence of collision events, and improve the efficiency of the agent's movement.
  • the introduced random slowing-down rule based on an emotion-influenced factor and the random slowing-down probability function, is used to randomly decelerate the artificial mobile agent.
  • the emotion-influenced factor and the random slowing-down probability function to randomly decelerate the artificial mobile agent, the realism and unpredictability of the simulation environment are increased, simulating the random behavioral patterns of humans or organisms under the influence of emotions.
  • the position update rule is used to determine the position of either an automatically moving agent or a manually moving agent. Updating the agent's position based on speed and direction propels it to move within the simulation environment, enabling continuous movement and interaction of the agent in virtual space and providing a foundation for subsequent path planning and decision-making.
  • Step 6 Control the movement and path of the automated mobile agent through the field force planning method, and use the perception information sharing mechanism of the automated mobile agent, combined with the movement and path of the automated mobile agent and the emotional cues of different granularities in the operation scenario of the artificial mobile agent, to plan the movement path of the artificial mobile agent.
  • this embodiment introduces a specific application example of a heterogeneous mobile intelligent agent cooperative control and evolution method.
  • This embodiment takes traffic flow control as an example.
  • Road traffic flow is composed of a mix of connected and automated vehicles (CAVs) and human-driven vehicles (HDVs), sharing road resources.
  • CAVs connected and automated vehicles
  • HDVs human-driven vehicles
  • CAVs have a smaller reaction time.
  • CAVs automatically maintain a headway with the vehicle in front.
  • CAVs also possess environmental perception and autonomous driving capabilities, supporting more flexible and intelligent decision-making.
  • HDVs after embedding an emotion recognition module, generate driving strategies that incorporate driver emotion factors.
  • the combination of HDVs and CAVs can simulate near-realistic mixed traffic flow characteristics.
  • the continuous cellular automaton uses vehicles as cells and mixes the HDV cells of manually driven vehicles and CAV cells of autonomous connected vehicles according to a preset ratio. Through an iterative process, it generates manually driven vehicles and autonomous connected vehicles.
  • the state information ⁇ ⁇ sub>i ⁇ /sub>(t) of the continuous cellular automaton is expressed as:
  • x ⁇ sub>n ⁇ /sub> (t), ⁇ ⁇ sub>n ⁇ /sub> (t), and a ⁇ sub>n ⁇ /sub> (t) represent the position, velocity, and acceleration of vehicle n at the current time t, respectively;
  • E ⁇ sub>n ⁇ /sub> represents the ideal state vector of vehicle n;
  • AFF ⁇ sub> n ⁇ /sub> represents the driver's emotional state vector of vehicle n.
  • V (t), AFFn_A(t), and AFFn_D(t) represent the expected speed of vehicle n, the expected following distance of vehicle n, and the safe time interval of vehicle n, respectively.
  • AFFn_V(t), AFFn_A (t), and AFFn_D (t) represent the measures of the driver's emotional pleasure, emotional arousal, and emotional dominance at the current time t, respectively.
  • acceleration of vehicle n at the current time t is represented as:
  • a ⁇ sub>0 ⁇ /sub> represents the maximum acceleration of the vehicle
  • ⁇ represents the vehicle acceleration exponent
  • s ⁇ sub>n ⁇ /sub> (t) and ⁇ (t) represent the relative distance and relative speed between the preceding vehicle n+1 and the current vehicle n, respectively
  • B represents the absolute value of the vehicle's comfortable deceleration
  • S ⁇ sub>0 ⁇ /sub> represents the stationary safety distance
  • v ⁇ sub>n+1 ⁇ /sub> (t) represents the speed of the preceding vehicle n+1 at the current time t.
  • n refers to a manually driven vehicle (HDV) or a networked automated vehicle (CAV) that is operating in mixed traffic flow.
  • HDV manually driven vehicle
  • CAV networked automated vehicle
  • An emotion recognition module is embedded in the HDV (High-Definition Vehicle) body of the artificially driven vehicle to obtain emotional cues of different granularities based on fuzzy theory; the specific steps include:
  • the emotion perception layer is used to capture and extract the emotional cues of the characters and the emotional cues of the scene in the running scene through the character channel and the scene channel respectively, forming a preliminary feature vector.
  • a three-valued input and single-valued output fuzzy inference model of emotion is constructed, with emotional pleasure, emotional arousal and emotional dominance as input variables and emotional influence factors as output variables.
  • fuzzy sets of input variables are constructed respectively, including fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional pleasure, fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional arousal, and fuzzy set 1, fuzzy set 2, and fuzzy set 3 for emotional dominance;
  • Construct fuzzy sets of output variables including fuzzy set 1 of emotion factors, fuzzy set 2 of emotion factors, and fuzzy set 3 of emotion factors;
  • the fuzzy sets of input and output variables are fuzzified using Gaussian membership functions.
  • a fuzzy rule matrix for emotion representation is constructed, and the constructed fuzzy rule matrix for emotion representation is represented as follows:
  • the first three columns represent the fuzzy set indices corresponding to the input variables of emotional pleasure, emotional arousal, and emotional dominance, respectively; the fourth column represents the fuzzy set index corresponding to the output variable of emotional influence factor; the fifth column represents the rule weights of the emotion representation fuzzy rule matrix; and the sixth column represents the connectives of the rules in the emotion representation fuzzy rule matrix.
  • the rule weight is 1, the logical connective is "AND”; when the rule weight is 2, the logical connective is "OR”.
  • the rule in the first row of the emotion representation fuzzy rule matrix is: when the input emotion pleasure degree belongs to the fuzzy set of emotion pleasure degree one, the input emotion arousal degree belongs to the fuzzy set of emotion arousal degree three, and the input emotion dominance degree belongs to the fuzzy set of emotion dominance degree three, the output emotion factor fuzzy set one is output.
  • the rule weight in the first row is 1, and the logical connector is "AND".
  • the rule in the second row of the fuzzy rule matrix is: when the input emotional pleasure degree belongs to the second fuzzy set of emotional pleasure degree, the input emotional arousal degree belongs to the first fuzzy set of emotional arousal degree, and the input emotional dominance degree belongs to the third fuzzy set of emotional dominance degree, the output emotional factor fuzzy set 2 is given.
  • the rule weight in the second row is 1, and the logical connector is "AND".
  • the rule in the third row of the fuzzy rule matrix is: when the input emotional pleasure degree belongs to the first fuzzy set of emotional pleasure degree, the input emotional arousal degree belongs to the second fuzzy set of emotional arousal degree, and the input emotional dominance degree belongs to the first fuzzy set of emotional dominance degree, the output emotional factor fuzzy set three is given.
  • the rule weight in the third row is 1, and the logical connector is "AND".
  • the continuous sentiment values in the initial feature vector are converted into the membership degree of the fuzzy set of the input set through fuzzification.
  • a fuzzy inference engine is used to map the input fuzzy set to the output fuzzy set, and calculations are performed to obtain the final output fuzzy set.
  • AFF n_V (t), AFF n_A (t), and AFF n_D (t) represent the measures of the driver's emotional pleasure, emotional arousal, and emotional dominance at the current time t, respectively.
  • the evolution rules of the artificial mobile agent include acceleration rules, deceleration rules, random slowing rules with the introduction of a random slowing probability function, and position update rules.
  • v n (t+1) min(v n (t)+a n (t), v max );
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the speed of vehicle n at time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the speed of vehicle n at time t
  • a ⁇ sub>n ⁇ /sub>(t) represents the acceleration of vehicle n at time t
  • v ⁇ sub>max ⁇ /sub> represents the maximum speed limit of the road.
  • d_safe represents the safe following distance
  • s_n (t) represents the relative distance between the preceding vehicle n+1 and the current vehicle n
  • s_0 represents the stationary safe distance
  • the driver of vehicle n has a probability of p ⁇ sub>n ⁇ /sub> (t+1) of randomly slowing down, where p ⁇ sub>n ⁇ /sub> (t+1) represents the random slowdown probability of vehicle n.
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the velocity of vehicle n at time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the velocity of vehicle n at time t
  • c represents the deceleration value
  • the position x ⁇ sub>n ⁇ /sub>(t) and velocity v ⁇ sub>n ⁇ /sub> (t) of vehicle n at the current time t jointly determine the position x ⁇ sub>n ⁇ /sub> (t+1) of vehicle n at the next time t+1.
  • x ⁇ sub>n ⁇ /sub> (t+1) represents the position of car n at the next time step (t+1)
  • x ⁇ sub>n ⁇ /sub>(t) represents the position of car n at the current time step t
  • v ⁇ sub>n ⁇ /sub> (t) represents the speed of car n at the current time step t.
  • the process involves acquiring the position, velocity, and acceleration of an autonomous mobile agent, constructing evolutionary rules based on an improved interaction potential energy field function, and updating the agent's state. Specific steps include:
  • the virtual potential energy field contains virtual potential energy field forces, which are generated by the potential energy gradient in the potential energy field.
  • a virtual potential energy field typically consists of two parts: an attractive field generated by the target point and a repulsive field generated by the obstacle.
  • the virtual forces of the potential energy field can be classified into position virtual forces, velocity virtual forces and acceleration virtual forces according to their generation mechanism.
  • the virtual forces of the potential energy field can be classified into position virtual attraction and position virtual repulsion, velocity virtual attraction and velocity virtual repulsion, and acceleration virtual attraction and acceleration virtual repulsion according to their interaction relationship.
  • r represents the distance from the automatically connected vehicle to the target point
  • r0 represents the range of influence of the repulsive potential field
  • ⁇ v represents the speed difference between the preceding vehicle and the current vehicle
  • ⁇ a represents the acceleration difference between the preceding vehicle and the current vehicle
  • krep-d represents the position virtual repulsive potential energy coefficient
  • krep-v represents the velocity virtual repulsive potential energy coefficient
  • krep-a represents the acceleration virtual repulsive potential energy coefficient.
  • the vehicle Since the vehicle always expects to travel at its maximum speed, if it does not reach the maximum expected speed, it will accelerate until it reaches the speed and then move at a constant speed.
  • r ⁇ sub> g ⁇ /sub> represents the distance from the automatically connected vehicle to the target point
  • ⁇ v ⁇ sub> e ⁇ /sub> represents the speed difference between the desired speed and the current speed
  • ⁇ a ⁇ sub> e ⁇ /sub> represents the difference between the desired acceleration and the current acceleration
  • k ⁇ sub>att-d ⁇ /sub> represents the virtual gravitational potential energy coefficient of position
  • k ⁇ sub>att-v ⁇ /sub> represents the virtual gravitational potential energy coefficient of velocity
  • k ⁇ sub>att-a ⁇ /sub> represents the virtual gravitational potential energy coefficient of acceleration.
  • F ⁇ sub>i ⁇ /sub> represents the virtual force with index i generated by the potential energy field U ⁇ sub> i ⁇ /sub>
  • the negative sign indicates that the direction of the virtual force of the potential energy field extends from the high potential energy to the low potential energy. This represents the gradient function.
  • the virtual force of position includes virtual attraction and virtual repulsion
  • the virtual force of velocity includes virtual attraction and virtual repulsion
  • the virtual force of acceleration includes virtual attraction and virtual repulsion.
  • the virtual repulsive force F rep-d at the position is expressed as:
  • kre-d represents the virtual repulsive potential energy coefficient at the location
  • r represents the distance from the automatically connected vehicle to the target point
  • r0 represents the range of influence of the repulsive potential field. This represents the distance vector from the automatically connected vehicle to the target point. This represents the magnitude of the distance vector from the automatically connected vehicle to the target point.
  • ka_att-d represents the virtual gravitational potential energy coefficient of the location
  • r_g represents the distance from the automatically connected vehicle to the target point.
  • the virtual velocity repulsion Frep -v can be expressed as:
  • kre -v represents the velocity virtual repulsive potential energy coefficient
  • r represents the distance from the automatically connected vehicle to the target point
  • r ⁇ sub>0 ⁇ /sub> represents the range of influence of the repulsive potential field. This represents the speed difference vector between the vehicle in front and the vehicle in front. This represents the magnitude of the velocity difference vector between the vehicle in front and the vehicle itself.
  • ka att-v represents the velocity virtual gravitational potential energy coefficient
  • ⁇ ve represents the velocity difference between the desired velocity and the current velocity
  • rg represents the distance from the automatically connected vehicle to the target point
  • tanh represents the hyperbolic tangent function
  • represents the magnitude of the velocity difference vector between the desired velocity and the current velocity.
  • the virtual repulsive force Frep-a due to acceleration is expressed as:
  • kre -a represents the virtual repulsive potential energy coefficient of acceleration
  • r represents the distance from the automatically connected vehicle to the target point
  • r0 represents the range of influence of the repulsive potential field. This represents the vector of acceleration differences between the vehicle in front and the vehicle in front. This represents the magnitude of the acceleration difference vector between the vehicle in front and the vehicle itself.
  • F att-a tanh(k att-a
  • ka att-a represents the virtual gravitational potential energy coefficient of acceleration
  • ⁇ a e represents the difference between the desired acceleration and the current acceleration
  • r g represents the distance from the automatically connected vehicle to the target point
  • represents the magnitude of the vector of the difference between the desired acceleration and the current acceleration.
  • the position, velocity, and acceleration of the autonomous mobile agent are acquired in real time through its sensors, and the interaction potential energy field function is defined.
  • the interaction potential energy field function is improved by introducing virtual forces of position, velocity, and acceleration.
  • an evolution rule for the autonomous mobile agent is constructed, and the state of the autonomous mobile agent is updated.
  • the evolution rules of the autonomous mobile intelligent agent include acceleration rules, deceleration rules, and position update rules
  • the acceleration rule for the autonomous mobile agent is expressed as follows:
  • v ⁇ sub>n ⁇ /sub> (t+1) represents the velocity of vehicle n at time t
  • ⁇ ⁇ sub>n ⁇ /sub> (t) represents the velocity of vehicle n at time t
  • v ⁇ sub> max ⁇ /sub> represents the maximum speed limit of the road
  • a represents the acceleration of vehicle n calculated according to the double integral equation of motion
  • m represents the mass of vehicle n
  • d ⁇ sub>safe ⁇ /sub> represents the safe following distance
  • F represents the total virtual force.
  • the deceleration rule for the automatic mobile intelligent agent is: when the repulsive force generated by the virtual potential field of the current vehicle n+1 is greater than the attractive force, the total virtual force on the current vehicle n is manifested as a repulsive force;
  • the location update rules for automated mobile agents are consistent with those for human-made mobile agents.
  • the movement and path of the automated mobile agent are controlled by the field force planning method.
  • the movement path of the artificial mobile agent is planned by combining the movement and path of the automated mobile agent with emotional cues of different granularities in the operation scenario of the artificial mobile agent, and utilizing the perception information sharing mechanism of the automated mobile agent.
  • the methods for controlling the motion and movement path of an automated mobile intelligent agent through field force planning include:
  • the virtual forces from other agents acting on each autonomously moving agent are calculated, including position virtual force, velocity virtual force and acceleration virtual force.
  • the total virtual force of each autonomous mobile agent is obtained by vector synthesis of all virtual forces acting on it.
  • the driving speed and direction of the autonomous mobile intelligent agent are adjusted through a control algorithm.
  • the virtual force of the potential energy field is used to plan the driving path of an autonomous mobile intelligent agent.
  • this invention by introducing an emotion-influencing factor, accurately quantifies continuous-dimensional emotion values and integrates them into a random slowing probability function, thereby improving the realism and accuracy of collaborative control of heterogeneous mobile intelligent agents and making the behavior of artificial mobile intelligent agents more realistic.
  • the method based on potential energy field virtual force transforms the position, velocity, and acceleration information of the autonomous mobile intelligent agent into virtual field forces, effectively controlling its motion through field force planning, thus enhancing the autonomous decision-making and path planning capabilities of the autonomous mobile intelligent agent.
  • this invention significantly improves the adaptability and overall intelligence of collaborative control of heterogeneous mobile intelligent agents through emotion recognition, deceleration rules, and potential energy field optimization.
  • embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
  • a computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and/or one or more block diagrams.
  • These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and/or one or more block diagrams.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
  • Organic Low-Molecular-Weight Compounds And Preparation Thereof (AREA)

Description

一种异构移动智能体协同控制与演化方法 技术领域
本发明涉及一种异构移动智能体协同控制与演化方法，属于移动智能体的编队与协同控制技术领域。
背景技术
移动智能体系统的思想是多智能体在实现期望的编队运动时，综合考虑人工移动智能体和自动移动智能体的工作特性，以期完成一体化协同控制与编队管理任务。因其广泛应用于交通流运行控制、空中机具编队和太空探索等领域，一直以来都受到了广泛的关注。以往的研究中大多数未充分考虑人工移动智能体情绪因素和外部环境中不确定因素对于多智能体协同控制系统运行的影响。在实际的应用中，个体情绪因素、不确定环境的影响不仅会使各智能体偏离期望的运动路径，而且会对多智能体的协同编队运行造成不利影响。因此，针对不确定环境中多智能体协同控制问题的研究具有极其重要的理论和实际价值。
公开于该背景技术部分的信息仅仅旨在增加对本实用新型的总体背景的理解，而不应当被视为承认或以任何形式暗示该信息构成已为本领域普通技术人员所公知的现有技术。
发明内容
本发明要解决的技术问题是：现有技术中未充分考虑人工移动智能体的个体情绪因素以及局部环境中不确定因素对多智能体协同控制的影响，造成异构移动智能体系统的协同控制与演化无法逼近真实情况的问题。
为解决上述技术问题，本发明是采用下述技术方案实现的。
本发明提供一种异构移动智能体协同控制与演化方法，包括：
利用连续元胞自动机将人工移动智能体元胞和自动移动智能体元胞按照预设的比例进行混合，并通过迭代过程生成人工移动智能体和自动移动智能体；
在所述人工移动智能体中嵌入情绪认知模块，根据模糊理论，获取不同粒度的情绪线索；
根据所述不同粒度的情绪线索，构建人工移动智能体的演化规则，更新人工移动智能体的状态；
所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则；
获取自动移动智能体的位置、速度和加速度，基于改进的相互作用势能场函数，构建自动移动智能体演化规则，更新自动移动智能体的状态；
所述自动移动智能体演化规则包括加速规则、减速规则和位置更新规则；
所述加速规则，用于调整自动移动智能体或人工移动智能体的速度；
所述减速规则，用于避免自动移动智能体和人工移动智能体之间发生碰撞；
所述引入随机慢化概率函数的随机慢化规则，用于基于情绪影响因子以及随机慢化概率函数，对人工移动智能体进行随机减速；
所述位置更新规则，用于决定自动移动智能体或人工移动智能体的位置；
通过场力规划方法控制自动移动智能体运动以及运动路径，并利用自动移动智能体的感知信息共享机制，结合自动移动智能体的运动和运动路径以及人工移动智能体运行场景下不同粒度的情绪线索，规划人工移动智能体的运动路径。
基于前述的异构移动智能体协同控制与演化方法，所述情绪认知模块包括情绪感知层和特征融合网络，所述情绪感知层采用双通道融合注意力机制网络结构；根据模糊理论，获取不同粒度的情绪线索的方法包括：
利用所述情绪感知层分别通过人物通道和场景通道捕捉并提取运行场景中的人物情绪线索和场景情绪线索，形成初步特征向量；
利用模糊理论对初步特征向量中连续维度的情绪值进行量化处理，获得情绪影响因子；
将所述情绪影响因子输入所述特征融合网络，进行融合处理，获取不同粒度的情绪线索。
基于前述的异构移动智能体协同控制与演化方法，利用模糊理论对初步特征向量中连续维度的情绪值进行量化处理，获得情绪影响因子的方法，包括：
基于Mamdani模型，将情绪愉悦度、情绪唤醒度和情绪支配度作为输入变量，情绪影响因子作为输出变量，构建三值输入单值输出情绪模糊推理模型；
按照情绪愉悦度、情绪唤醒度和情绪支配度的情绪度量值分别构建输入变量模糊集合，包括情绪愉悦度模糊集合一、情绪愉悦度模糊集合二和情绪愉悦度模糊集合三，情绪唤醒度模糊集合一、情绪唤醒度模糊集合二和情绪唤醒度模糊集合三以及情绪支配度模糊集合一、情绪支配度模糊集合二和情绪支配度模糊集合三；
构建输出变量模糊集合，包括情绪因子模糊集合一、情绪因子模糊集合二和情绪因子模糊集合三；
利用高斯型隶属度函数，对输入变量模糊集合和输出变量模糊集合进行模糊化；
分析Emotic数据集中情绪愉悦度、情绪唤醒度和情绪支配度的数据样本，获得能够反映输入变量与输出变量之间关系的人工经验规则库；
根据人工经验规则库，构建情绪表征模糊规则矩阵；
将初步特征向量中的连续情绪值通过模糊化转换为输入集合模糊集合的隶属度；
根据情绪表征模糊规则矩阵，使用模糊推理器将输入模糊集合映射到输出模糊集合，进行计算，获得最终的输出模糊集合；
使用质心法对最终的输出模糊集合进行去模糊化处理，获得情绪影响因子。
基于前述的异构移动智能体协同控制与演化方法，根据人工经验规则库，构建情绪表征模糊规则矩阵，将构建的情绪表征模糊规则矩阵表示为：
其中，情绪表征模糊规则矩阵前三列分别表示输入变量情绪愉悦度、情绪唤醒度和情绪支配度对应的模糊集合索引，第四列表示输出变量情绪影响因子对应的模糊集合索引，第五列表示情绪表征模糊规则矩阵的规则权重，第六列表示情绪表征模糊规则矩阵规则的连接词，当规则权重为1时，逻辑连接词为“与”，当规则权重为2时，逻辑连接词为“或”；
其中，情绪表征模糊规则矩阵第一行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合三且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合一，第一行的规则权重为1，逻辑连接词为“与”；
模糊规则矩阵第二行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合二且输入情绪唤醒度属于情绪唤醒度模糊集合一且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合二，第二行的规则权重为1，逻辑连接词为“与”；
模糊规则矩阵第三行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合二且输入情绪支配度属于情绪支配度模糊集合一时，输出情绪因子模糊集合三，第三行的规则权重为1，逻辑连接词为“与”。
基于前述的异构移动智能体协同控制与演化方法，当所述运行场景为道路交通流时，所述连续元胞自动机将车辆作为元胞，通过迭代过程生成人工驾驶车辆和自动网联车辆，所述连续元胞自动机的状态信息σi(t)，表示为：
其中，xn(t)、νn(t)和an(t)分别表示本车n在当前时刻t的位置、速度和加速度，En表示本车n的车辆理想状态向量，AFFn表示本车n的驾驶员情绪状态向量，分别表示本车n的车辆期望速度、本车n的车辆期望跟驰距离和本车n的车辆安全时间间隔，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员的情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
所述连续元胞自动机将车辆作为元胞，通过迭代过程生成人工驾驶车辆和自动网联车辆，将人工驾驶车辆an(t)表示为：
其中，A0表示车辆最大加速度，λ表示车辆加速度指数，sn(t)和Δν(t)分别表示前车n+1和本车n的相对距离和相对速度，B表示车辆舒适减速度绝对值，S0表示静止安全距离，vn+1(t)表示前车n+1在当前时刻t的速度。
基于前述的异构移动智能体协同控制与演化方法，使用质心法对最终的输出模糊集合进行去模糊化处理，获得情绪影响因子，将获得的情绪影响因子k表达为：
k＝f(AFFn_V(t),AFFn_A(t),AFFn_D(t)) (4)；
其中，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
基于前述的异构移动智能体协同控制与演化方法，所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则；其中，将人工移动智能体的加速规则表示为：
vn(t+1)＝min(vn(t)+an(t),vmax) (5)；
其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，an(t)表示本车n在当前时刻t的加速度，vmax表示道路的最大限速。
将人工移动智能体的减速规则表示为：
dsafe＝sn(t)-s0 (6)；
其中，dsafe表示安全跟驰距离，sn(t)表示前车n+1和本车n的相对距离，s0表示静止安全距离。
基于本车n驾驶员的情绪影响因子以及随机慢化概率函数，本车n驾驶员有pn(t+1)的概率进行随机减速，其中，pn(t+1)表示本车n的随机慢化概率，将引入随机慢化概率函数的随机慢化规则表示为：
vn(t+1)＝max(vn(t)-c,0) (7)；
其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，c表示减速度数值；
本车n在当前时刻t的位置xn(t)和速度vn(t)共同决定本车n下一时刻t+1的位置xn(t+1)；
将人工移动智能体的位置更新规则的表达式为：
xn(t+1)＝xn(t)+vn(t+1) (8)；
其中，xn(t+1)表示在本车n下一时刻(t+1)的位置，xn(t)表示本车n在当前时刻t的位置，vn(t)表示本车n在当前时刻t的速度。
基于前述的异构移动智能体协同控制与演化方法，所述自动移动智能体的演化规则包括加速规则、减速规则和位置更新规则；
其中，将自动移动智能体的加速规则表示为：
其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，，vmax表示道路的最大限速，a表示根据双积分运动方程进行计算得到的本车n的加速度，m表示本车n的质量，dsafe表示安全跟驰距离，F表示总虚拟力。
自动移动智能体的减速规则为：当前车n+1虚拟势场产生的斥力大于引力时，本车n所受的总虚拟力表现为斥力；
自动移动智能体的的位置更新规则与人工移动智能体的位置更新规则一致。
基于前述的异构移动智能体协同控制与演化方法，通过场力规划方法控制自动移动智能体运动以及运动路径的方法，包括：
通过自动移动智能体的传感器实时获取自动移动智能体的位置、速度和加速度，定义相互作用势能场函数；
引入位置虚拟力、速度虚拟力和加速度虚拟力，改进相互作用势能场函数；
根据改进的相互作用势能场函数，计算每个自动移动智能体受到的来自其他智能体的虚拟力，包括位置虚拟力、速度虚拟力和加速度虚拟力；
将每个自动移动智能体受到的所有虚拟力进行矢量合成，得到每个自动移动智能体的总虚拟力；
利用总虚拟力作为控制信号，通过控制算法调整自动移动智能体的行驶速度和方向；
利用势能场虚拟力规划自动移动智能体的行驶路径。
基于前述的异构移动智能体协同控制与演化方法，所述位置虚拟力包括位置虚拟引力和位置虚拟斥力，所述速度虚拟力包括速度虚拟引力和速度虚拟斥力，所述加速度虚拟力包括加速度虚拟引力和加速度虚拟斥力；
将所述位置虚拟斥力Frep-d表示为：
其中，krep-d表示位置虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示自动联网车辆到目标点的距离向量，表示自动联网车辆到目标点的距离向量的模长。
将所述位置虚拟引力Fatt-d表示为：
Fatt-d＝katt-drg (11)；
其中，katt-d表示位置虚拟引力势能系数，rg表示自动联网车辆到目标点的距离。
将所述速度虚拟斥力Frep-v表示为：
其中，krep-v表示速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的速度差向量，表示前车与本车的速度差向量的模长。
将所述速度虚拟引力Fatt-v表示为：
Fatt-v＝tanh(katt-v||Δve||)rg  (13)；
其中，katt-v表示速度虚拟引力势能系数，Δve表示期望速度和当前速度的速度差，rg表示自动联网车辆到目标点的距离，tanh表示双曲正切函数，||Δve||表示期望速度和当前速度的速度差向量模长。
将所述加速度虚拟斥力Frep-a表示为：
其中，krep-a表示加速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的加速度差值向量，表示前车与本车的加速度差向量的模长。
将所述加速度虚拟引力Fatt-a表示为：
Fatt-a＝tanh(katt-a||Δae||)rg；  (15)；
其中，katt-a表示加速度虚拟引力势能系数，Δae表示期望加速度和当前加速度差值，rg表示自动联网车辆到目标点的距离，||Δae||表示期望加速度和当前加速度差值向量的模长。
本发明通过引入情绪影响因子，精准量化连续维度情绪值并集成至随机慢化概率函数中，提升了异构移动智能体协同控制的真实性和精度，使人工移动智能体的行为更贴近实际。
同时，基于势能场虚拟力的方法将自动移动智能体的位置、速度和加速度信息转化为虚拟场力，通过场力规划有效控制其运动，增强了自动移动智能体的自主决策和路径规划能力。
本发明针对复杂环境协同控制难题，通过情绪认知、减速规则及势能场优化，显著提升了异构移动智能体协同控制的适应性和整体智能性。
与现有技术相比，本发明所达到的有益效果：
本发明在人工移动智能体中嵌入情绪认知模块，利用模糊理论处理不同粒度的情绪线索，使智能体能够感知并理解复杂环境中的情感信息，基于不同粒度的情绪线索构建的演化规则，使人工移动智能体的行为更加灵活多变，特别是引入随机慢化概率函数和情绪影响因子，能够根据情绪状态动态调整移动速度，不仅增加了智能体行为的多样性，还提智能体了其在复杂环境中的生存能力和适应性。
同时本发明通过建模虚拟斥力场和虚拟引力场，并基于改进的相互作用势能场函数构建自动移动智能体的演化规则，有效实现了智能体之间的动态避碰和协同运动。本发明还利用场力规划方法控制自动移动智能体的运动及路径，并结合感知信息共享机制和情绪线索，为人工移动智能体规划出更加合理、高效的运动路径。解决了现有技术中未充分考虑人工移动智能体的个体情绪因素以及局部环境中不确定因素对多智能体协同控制的影响，造成异构移动智能体系统的协同控制与演化无法逼近真实情况的问题。
附图说明
图1是本发明实施例提供的异构移动智能体的结构示意图；
图2是本发明实施例提供的情绪模糊推理模型构建流程示意图；
图3是本发明实施例提供的异构移动智能体协同控制的运行场景中势能场虚拟力模型仿真示意图。
具体实施方式
下面通过附图以及具体实施例对本发明技术方案做详细地说明，应当理解本发明实施例以及实施例中的具体特征是对本发明技术方案的详细的说明，而不是对本发明技术方案的限定，在不冲突的情况下，本发明实施例以及实施例中的技术特征可以相互组合。
实施例1
如图1所示，本实施例介绍一种异构移动智能体协同控制与演化方法，包括：
步骤1：利用连续元胞自动机将人工移动智能体元胞和自动移动智能体元胞按照预设的比例进行混合，并通过迭代过程生成人工移动智能体和自动移动智能体；
通过连续元胞自动机模型，将人工移动智能体和自动移动智能体按预设比例混合，模拟真实交互环境中的异构移动智能体，为后续的协同控制研究提供了基础的仿真环境，使得实验结果更加贴近实际应用场景，增强了研究的实用性和可靠性。
步骤2：在所述人工移动智能体中嵌入情绪认知模块，根据模糊理论，获取不同粒度的情绪线索；
通过嵌入情绪认知模块，利用模糊理论获取并量化人工移动智能体的情绪线索，情绪线索会对人工移动智能体的运动造成影响，考虑不同粒度的情绪线索能够使人工移动智能体的行为决策更加符合人类的实际情况，提高了仿真模型的真实性和准确性，同时也为异构智能体的协同控制引入了新的维度——情感因素。
步骤3：根据所述不同粒度的情绪线索，构建人工移动智能体的演化规则，更新人工移动智能体的状态；
所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则；
步骤4：根据情绪线索构建包括加速规则、减速规则、随机慢化规则和位置更新规则在内的人工移动智能体的演化规则，以更新人工移动智能体的状态。
上述规则使人工移动智能体的行为能够根据情绪变化而动态调整，增加了仿真环境的复杂性和真实性。同时，随机慢化规则引入了不确定性，使行为更加难以预测。
步骤5：获取自动移动智能体的位置、速度和加速度，基于改进的相互作用势能场函数，构建自动移动智能体演化规则，更新自动移动智能体的状态；
基于改进的相互作用势能场函数，构建自动移动智能体的加速规则、减速规则和位置更新规则，以优化自动移动智能体的运动路径和避障能力，提高了自动移动智能体在复杂环境中的适应性和安全性，使自动移动智能体的能够更智能地应对各种情况，如避让行人、调整速度等。
所述自动移动智能体演化规则包括加速规则、减速规则和位置更新规则；
所述加速规则，用于调整自动移动智能体或人工移动智能体的速度。
所述减速规则，用于避免自动移动智能体和人工移动智能体之间发生碰撞。
加速规则用于调整智能体的速度以匹配其运动需求，减速规则用于避免碰撞，确保安全，共同维护仿真环境的稳定性和安全性，减少了碰撞事件的发生，并提高了智能体运动的效率。
所述引入随机慢化概率函数的随机慢化规则，用于基于情绪影响因子以及随机慢化概率函数。对人工移动智能体进行随机减速；通过情绪影响因子和随机慢化概率函数，对人工移动智能体进行随机减速，增加了仿真环境的真实性和不可预测性，模拟了人类或生物在情绪影响下的随机行为模式。
所述位置更新规则，用于决定自动移动智能体或人工移动智能体的位置。根据速度和方向更新智能体的位置，推动其在仿真环境中移动，实现了智能体在虚拟空间中的连续移动和交互，为后续的路径规划和决策提供了基础。
步骤6：通过场力规划方法控制自动移动智能体运动以及运动路径，并利用自动移动智能体的感知信息共享机制，结合自动移动智能体的运动和运动路径以及人工移动智能体运行场景下不同粒度的情绪线索，规划人工移动智能体的运动路径。
通过建模虚拟斥力场和虚拟引力场来控制自动移动智能体的运动路径，并利用感知信息共享机制结合不同粒度的情绪线索来规划人工移动智能体的运动路径。提高了自动移动智能体的路径规划效率和避障能力，同时使人工移动智能体的运动路径更加合理和安全。
实施例2
与实施例1基于相同的发明构思，本实施例介绍一种异构移动智能体协同控制与演化方法的具体应用实例。
本实施例以交通流运行控制为例，道路交通流由自动网联车辆(Connected and Automated Vehicles，CAV)和人工驾驶车辆(Human Driven Vehicles，HDV)混合组成，共享道路资源。相比人工驾驶车辆HDV，自动网联车辆CAV具有更小的反应延迟时间，在行驶过程中自动网联车辆CAV与前车自动保持车头时距，同时自动网联车辆CAV具备环境感知与自主驾驶能力，支撑实现更加灵活、智能的决策。人工驾驶车辆HDV嵌入情绪识别模块后，生成融入驾驶员情绪影响因子的驾驶策略。人工驾驶车辆HDV和自动网联车辆CAV的结合可以模拟出逼近真实的混合交通流特性。
当运行场景为道路交通流时，连续元胞自动机将车辆作为元胞，利用连续元胞自动机将人工驾驶车辆HDV元胞和自动网联车辆CAV元胞按照预设的比例进行混合，通过迭代过程生成人工驾驶车辆和自动网联车辆。
将所述连续元胞自动机的状态信息σi(t)，表示为：
其中，xn(t)、νn(t)和an(t)分别表示本车n在当前时刻t的位置、速度和加速度，En表示本车n的车辆理想状态向量，AFFn表示本车n的驾驶员情绪状态向量，分别表示本车n的车辆期望速度、本车n的车辆期望跟驰距离和本车n的车辆安全时间间隔，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员的情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
其中，将本车n在当前时刻t加速度an(t)表示为：
其中，A0表示车辆最大加速度，λ表示车辆加速度指数，sn(t)和Δν(t)分别表示前车n+1和本车n的相对距离和相对速度，B表示车辆舒适减速度绝对值，S0表示静止安全距离，vn+1(t)表示前车n+1在当前时刻t的速度。
其中，将本车n是指正在混在混合交通流中运行的人工驾驶车辆HDV或自动联网车辆CAV。
在所述人工驾驶车辆HDV体中嵌入情绪认知模块，根据模糊理论，获取不同粒度的情绪线索；具体步骤包括：
(1)利用所述情绪感知层分别通过人物通道和场景通道捕捉并提取运行场景中的人物情绪线索和场景情绪线索，形成初步特征向量。
(2)利用模糊理论对初步特征向量中连续维度的情绪值进行量化处理，获得情绪影响因子，如图2所示，具体步骤包括：
基于Mamdani模型，将情绪愉悦度、情绪唤醒度和情绪支配度作为输入变量，情绪影响因子作为输出变量，构建三值输入单值输出情绪模糊推理模型；
按照情绪愉悦度、情绪唤醒度和情绪支配度的情绪度量值分别构建输入变量模糊集合，包括情绪愉悦度模糊集合一、情绪愉悦度模糊集合二和情绪愉悦度模糊集合三，情绪唤醒度模糊集合一、情绪唤醒度模糊集合二和情绪唤醒度模糊集合三以及情绪支配度模糊集合一、情绪支配度模糊集合二和情绪支配度模糊集合三；
构建输出变量模糊集合，包括情绪因子模糊集合一、情绪因子模糊集合二和情绪因子模糊集合三；
利用高斯型隶属度函数，对输入变量模糊集合和输出变量模糊集合进行模糊化；
分析Emotic数据集中情绪愉悦度、情绪唤醒度和情绪支配度的数据样本，获得能够反映输入变量与输出变量之间关系的人工经验规则库；
根据人工经验规则库，构建情绪表征模糊规则矩阵，将构建的情绪表征模糊规则矩阵表示为：
其中，情绪表征模糊规则矩阵前三列分别表示输入变量情绪愉悦度、情绪唤醒度和情绪支配度对应的模糊集合索引，第四列表示输出变量情绪影响因子对应的模糊集合索引，第五列表示情绪表征模糊规则矩阵的规则权重，第六列表示情绪表征模糊规则矩阵规则的连接词，当规则权重为1时，逻辑连接词为“与”，当规则权重为2时，逻辑连接词为“或”；
其中，情绪表征模糊规则矩阵第一行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合三且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合一，第一行的规则权重为1，逻辑连接词为“与”；
模糊规则矩阵第二行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合二且输入情绪唤醒度属于情绪唤醒度模糊集合一且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合二，第二行的规则权重为1，逻辑连接词为“与”；
模糊规则矩阵第三行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合二且输入情绪支配度属于情绪支配度模糊集合一时，输出情绪因子模糊集合三，第三行的规则权重为1，逻辑连接词为“与”。
将初步特征向量中的连续情绪值通过模糊化转换为输入集合模糊集合的隶属度；
根据情绪表征模糊规则矩阵，使用模糊推理器将输入模糊集合映射到输出模糊集合，进行计算，获得最终的输出模糊集合；
使用质心法对最终的输出模糊集合进行去模糊化处理，获得情绪影响因子，将获得的情绪影响因子表达为：
k＝f(AFFn_V(t),AFFn_A(t),AFFn_D(t))；
其中，k表示情绪影响因子，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
(3)将所述情绪影响因子输入所述特征融合网络，进行融合处理，获取不同粒度的情绪线索。
根据所述不同粒度的情绪线索，构建人工移动智能体的演化规则，更新人工移动智能体的状态。
所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则。
将构建的人工移动智能体的加速规则表示为：
vn(t+1)＝min(vn(t)+an(t),vmax)；
其中，vn(t+1)表示本车n在t时刻的速度，νn(t)表示本车n在t时刻的速度，an(t)表示本车n在t时刻的加速度，vmax表示道路的最大限速。
将构建的人工移动智能体的减速规则表示为：
dsafe＝sn(t)-s0；
其中，dsafe表示安全跟驰距离，sn(t)表示前车n+1和本车n的相对距离，s0表示静止安全距离。
由于考虑驾驶员的情绪因素会影响人工驾驶车辆HDV的运行，基于本车n驾驶员的情绪影响因子以及随机慢化概率函数，本车n驾驶员有pn(t+1)的概率进行随机减速，其中，pn(t+1)表示本车n的随机慢化概率，将引入随机慢化概率函数的构建的随机慢化规则表示为：
vn(t+1)＝max(vn(t)-c,0)；
其中，vn(t+1)表示本车n在t时刻的速度，νn(t)表示本车n在t时刻的速度，c表示减速度数值；
本车n在当前时刻t的位置xn(t)和速度vn(t)共同决定本车n下一时刻t+1的位置xn(t+1)，将构建的人工移动智能体的位置更新规则的表达式为：
xn(t+1)＝xn(t)+vn(t+1)；
其中，xn(t+1)表示在本车n下一时刻(t+1)的位置，xn(t)表示本车n在当前时刻t的位置，vn(t)表示本车n在当前时刻t的速度。
获取自动移动智能体的位置、速度和加速度，基于改进的相互作用势能场函数，构建自动移动智能体演化规则，更新自动移动智能体的状态，具体步骤包括：
定义一个虚拟势能场，通过场力规划方法控制自动移动智能体运动以及运动路径，虚拟势能场包含势能场虚拟力，所述势能场虚拟力是由势能场中的势能梯度产生的。
虚拟势能场通常包括两部分：目标点产生的吸引力场和障碍物产生的排斥力场。
如图3所示，所述势能场虚拟力按产生机理可分为位置虚拟力、速度虚拟力和加速度虚拟力，所述势能场虚拟力按作用关系可分为位置虚拟引力与位置虚拟斥力，速度虚拟引力与速度虚拟斥力以及加速度虚拟引力与加速度虚拟斥力。
将虚拟位置斥力场Urep-d表示为：
将速度斥力场斥力场Urep-v表示为：
将加速度斥力场Urep-a表示为：
其中，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，Δv表示前车与本车的速度差值，Δa表示前车与本车的加速度差值，krep-d表示位置虚拟斥力势能系数，krep-v表示速度虚拟斥力势能系数，krep-a表示加速度虚拟斥力势能系数。
由于考虑车辆总是期望以最大速度行驶，若未到达最大期望速度则进行加速，直到达速进行匀速运动。
所以将虚拟位置的引力场Uatt-d表示为：
所以将虚拟速度引力场Uatt-v表示为：
所以将虚拟加速度的引力场Uatt-a表示为：
其中，rg表示自动联网车辆到目标点的距离，Δve表示期望速度和当前速度的速度差，Δae表示期望加速度和当前加速度的差值，katt-d表示位置虚拟引力势能系数，katt-v表示速度虚拟引力势能系数，katt-a表示加速度虚拟引力势能系数。
由于自动联网车辆在势能场中的运动规划与势能虚拟力相关，将势能场虚拟力与势能场之间的关系表示为：
其中，Fi表示势能场Ui产生的索引为i的虚拟力，负号表示势能场虚拟力的方向是由高势场能量处向低势场能量处延伸；表示梯度函数。
所述位置虚拟力包括位置虚拟引力和位置虚拟斥力，所述速度虚拟力包括速度虚拟引力和速度虚拟斥力，所述加速度虚拟力包括加速度虚拟引力和加速度虚拟斥力。
根据势能场虚拟力与势能场之间的关系，将所述位置虚拟斥力Frep-d表示为：
其中，krep-d表示位置虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示自动联网车辆到目标点的距离向量，表示自动联网车辆到目标点的距离向量的模长。
将所述位置虚拟引力Fatt-d表示为：
Fatt-d＝katt-drg (11)；
其中，katt-d表示位置虚拟引力势能系数，rg表示自动联网车辆到目标点的距离。
将所述速度虚拟斥力Frep-v表示为：
其中，krep-v表示速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的速度差向量，表示前车与本车的速度差向量的模长。
将所述速度虚拟引力Fatt-v表示为：
Fatt-v＝tanh(katt-v||Δve||)rg  (13)；
其中，katt-v表示速度虚拟引力势能系数，Δve表示期望速度和当前速度的速度差，rg表示自动联网车辆到目标点的距离，tanh表示双曲正切函数，||Δve||表示期望速度和当前速度的速度差向量模长。
将所述加速度虚拟斥力Frep-a表示为：
其中，krep-a表示加速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的加速度差值向量，表示前车与本车的加速度差向量的模长。
将所述加速度虚拟引力Fatt-a表示为：
Fatt-a＝tanh(katt-a||Δae||)rg；  (15)；
其中，katt-a表示加速度虚拟引力势能系数，Δae表示期望加速度和当前加速度差值，rg表示自动联网车辆到目标点的距离，||Δae||表示期望加速度和当前加速度差值向量的模长。
通过自动移动智能体的传感器实时获取自动移动智能体的位置、速度和加速度，定义相互作用势能场函数；
引入位置虚拟力、速度虚拟力和加速度虚拟力，改进相互作用势能场函数。
根据改进相互作用势能场函数，构建自动移动智能体演化规则，更新自动移动智能体的状态。
所述自动移动智能体的演化规则包括加速规则、减速规则和位置更新规则；
其中，将自动移动智能体的加速规则表示为：
其中，vn(t+1)表示本车n在t时刻的速度，νn(t)表示本车n在t时刻的速度，，vmax表示道路的最大限速，a表示根据双积分运动方程进行计算得到的本车n的加速度，m表示本车n的质量，dsafe表示安全跟驰距离，F表示总虚拟力。
自动移动智能体的减速规则为：当前车n+1虚拟势场产生的斥力大于引力时，本车n所受的总虚拟力表现为斥力；
自动移动智能体的的位置更新规则与人工移动智能体的位置更新规则一致。
通过场力规划方法控制自动移动智能体运动以及运动路径，并利用自动移动智能体的感知信息共享机制，结合自动移动智能体的运动和运动路径以及人工移动智能体运行场景下不同粒度的情绪线索，规划人工移动智能体的运动路径。
其中，通过场力规划方法控制自动移动智能体运动以及运动路径的方法包括：
根据改进的相互作用势能场函数，计算每个自动移动智能体受到的来自其他智能体的虚拟力，包括位置虚拟力、速度虚拟力和加速度虚拟力；
将每个自动移动智能体受到的所有虚拟力进行矢量合成，得到每个自动移动智能体的总虚拟力；
利用总虚拟力作为控制信号，通过控制算法调整自动移动智能体的行驶速度和方向；
利用势能场虚拟力规划自动移动智能体的行驶路径。
综上实施例，本发明通过引入情绪影响因子，精准量化连续维度情绪值并集成至随机慢化概率函数中，提升了异构移动智能体协同控制的真实性和精度，使人工移动智能体的行为更贴近实际。同时，基于势能场虚拟力的方法将自动移动智能体的位置、速度和加速度信息转化为虚拟场力，通过场力规划有效控制其运动，增强了自动移动智能体的自主决策和路径规划能力。本发明针对复杂环境协同控制难题，通过情绪认知、减速规则及势能场优化，显著提升了异构移动智能体协同控制的适应性和整体智能性。
本领域内的技术人员应明白，本发明的实施例可提供为方法、系统、或计算机程序产品。因此，本发明可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本发明可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明是参照根据本发明实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
以上结合附图对本发明的实施例进行了描述，但是本发明并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本发明的启示下，在不脱离本发明宗旨和权利要求所保护的范围情况下，还可做出很多形式，这些均属于本发明的保护之内。

Claims (10)

  1. 一种异构移动智能体协同控制与演化方法，其特征在于，包括：
    利用连续元胞自动机将人工移动智能体元胞和自动移动智能体元胞按照预设的比例进行混合，并通过迭代过程生成人工移动智能体和自动移动智能体；
    在所述人工移动智能体中嵌入情绪认知模块，根据模糊理论，获取不同粒度的情绪线索；
    根据所述不同粒度的情绪线索，构建人工移动智能体的演化规则，更新人工移动智能体的状态；
    所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则；
    获取自动移动智能体的位置、速度和加速度，建模为虚拟斥力场和虚拟引力场，获取自动移动智能体的位置、速度和加速度，基于改进的相互作用势能场函数，构建自动移动智能体演化规则，更新自动移动智能体的状态；
    所述自动移动智能体演化规则包括加速规则、减速规则和位置更新规则；
    所述加速规则，用于调整自动移动智能体或人工移动智能体的速度；
    所述减速规则，用于避免自动移动智能体和人工移动智能体之间发生碰撞；
    所述引入随机慢化概率函数的随机慢化规则，用于基于情绪影响因子以及随机慢化概率函数，对人工移动智能体进行随机减速；
    所述位置更新规则，用于决定自动移动智能体或人工移动智能体的位置；
    通过场力规划方法控制自动移动智能体运动以及运动路径，并利用自动移动智能体的感知信息共享机制，结合自动移动智能体的运动和运动路径以及人工移动智能体运行场景下不同粒度的情绪线索，规划人工移动智能体的运动路径。
  2. 根据权利要求1所述的异构移动智能体协同控制与演化方法，其特征在于，所述情绪认知模块包括情绪感知层和特征融合网络，所述情绪感知层采用双通道融合注意力机制网络结构；根据模糊理论，获取不同粒度的情绪线索的方法包括：
    利用所述情绪感知层分别通过人物通道和场景通道捕捉并提取运行场景中的人物情绪线索和场景情绪线索，形成初步特征向量；
    利用模糊理论对初步特征向量中连续维度的情绪值进行量化处理，获得情绪影响因子；
    将所述情绪影响因子输入所述特征融合网络，进行融合处理，获取不同粒度的情绪线索。
  3. 根据权利要求2所述的异构移动智能体协同控制与演化方法，其特征在于，利用模糊理论对初步特征向量中连续维度的情绪值进行量化处理，获得情绪影响因子的方法，包括：
    基于Mamdani模型，将情绪愉悦度、情绪唤醒度和情绪支配度作为输入变量，情绪影响因子作为输出变量，构建三值输入单值输出情绪模糊推理模型；
    按照情绪愉悦度、情绪唤醒度和情绪支配度的情绪度量值分别构建输入变量模糊集合，包括情绪愉悦度模糊集合一、情绪愉悦度模糊集合二和情绪愉悦度模糊集合三、情绪唤醒度模糊集合一、情绪唤醒度模糊集合二和情绪唤醒度模糊集合三以及情绪支配度模糊集合一、情绪支配度模糊集合二和情绪支配度模糊集合三；
    构建输出变量模糊集合，包括情绪因子模糊集合一、情绪因子模糊集合二和情绪因子模糊集合三；
    利用高斯型隶属度函数，对输入变量模糊集合和输出变量模糊集合进行模糊化；
    分析Emotic数据集中情绪愉悦度、情绪唤醒度和情绪支配度的数据样本，获得能够反映输入变量与输出变量之间关系的人工经验规则库；
    根据人工经验规则库，构建情绪表征模糊规则矩阵；
    将初步特征向量中的连续情绪值通过模糊化转换为输入集合模糊集合的隶属度；
    根据情绪表征模糊规则矩阵，使用模糊推理器将输入模糊集合映射到输出模糊集合，进行计算，获得最终的输出模糊集合；
    使用质心法对最终的输出模糊集合进行去模糊化处理，获得情绪影响因子。
  4. 根据权利要求3所述的异构移动智能体协同控制与演化方法，其特征在于，根据人工经验规则库，构建情绪表征模糊规则矩阵，将构建的情绪表征模糊规则矩阵表示为：
    其中，情绪表征模糊规则矩阵前三列分别表示输入变量情绪愉悦度、情绪唤醒度和情绪支配度对应的模糊集合索引，第四列表示输出变量情绪影响因子对应的模糊集合索引，第五列表示情绪表征模糊规则矩阵的规则权重，第六列表示情绪表征模糊规则矩阵规则的连接词，当规则权重为1时，逻辑连接词为“与”，当规则权重为2时，逻辑连接词为“或”；
    其中，情绪表征模糊规则矩阵第一行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合三且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合一，第一行的规则权重为1，逻辑连接词为“与”；
    模糊规则矩阵第二行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合二且输入情绪唤醒度属于情绪唤醒度模糊集合一且输入情绪支配度属于情绪支配度模糊集合三时，输出情绪因子模糊集合二，第二行的规则权重为1，逻辑连接词为“与”；
    模糊规则矩阵第三行的规则为：当输入情绪愉悦度属于情绪愉悦度模糊集合一且输入情绪唤醒度属于情绪唤醒度模糊集合二且输入情绪支配度属于情绪支配度模糊集合一时，输出情绪因子模糊集合三，第三行的规则权重为1，逻辑连接词为“与”。
  5. 根据权利要求1所述的异构移动智能体协同控制与演化方法，其特征在于，当所述运行场景为道路交通流时，所述连续元胞自动机将车辆作为元胞，通过迭代过程生成人工驾驶车辆和自动网联车辆，所述连续元胞自动机的状态信息σi(t)，表示为：
    其中，xn(t)、νn(t)和an(t)分别表示本车n在当前时刻t的位置、速度和加速度，En表示本车n的车辆理想状态向量，AFFn表示本车n的驾驶员情绪状态向量，分别表示本车n的车辆期望速度、本车n的车辆期望跟驰距离和本车n的车辆安全时间间隔，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员的情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
    所述连续元胞自动机将车辆作为元胞，通过迭代过程生成人工驾驶车辆和自动网联车辆，将人工驾驶车辆an(t)表示为：
    其中，A0表示车辆最大加速度，λ表示车辆加速度指数，sn(t)和Δν(t)分别表示前车n+1和本车n的相对距离和相对速度，B表示车辆舒适减速度绝对值，S0表示静止安全距离，vn+1(t)表示前车n+1在当前时刻t的速度。
  6. 根据权利要求5所述的异构移动智能体协同控制与演化方法，其特征在于，使用质心法对最终的输出模糊集合进行去模糊化处理，获得情绪影响因子，将获得的情绪影响因子k表达为：
    k＝f(AFFn_V(t),AFFn_A(t),AFFn_D(t))   (4)；
    其中，AFFn_V(t)、AFFn_A(t)和AFFn_D(t)分别表示在当前时刻t本车n驾驶员情绪愉悦度、情绪唤醒度和情绪支配度的度量值。
  7. 根据权利要求6所述的异构移动智能体协同控制与演化方法，其特征在于，所述人工移动智能体的演化规则包括加速规则、减速规则、引入随机慢化概率函数的随机慢化规则和位置更新规则；其中，将人工移动智能体的加速规则表示为：
    vn(t+1)＝min(vn(t)+an(t),vmax)  (5)；
    其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，an(t)表示本车n在当前时刻t的加速度，vmax表示道路的最大限速。
    将人工移动智能体的减速规则表示为：
    dsafe＝sn(t)-s0  (6)；
    其中，dsafe表示安全跟驰距离，sn(t)表示前车n+1和本车n的相对距离，s0表示静止安全距离。
    基于本车n驾驶员的情绪影响因子以及随机慢化概率函数，本车n驾驶员有pn(t+1)的概率进行随机减速，其中，pn(t+1)表示本车n的随机慢化概率，将引入随机慢化概率函数的随机慢化规则表示为：
    vn(t+1)＝max(vn(t)-c,0)  (7)；
    其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，c表示减速度数值；
    本车n在当前时刻t的位置xn(t)和速度vn(t)共同决定本车n下一时刻t+1的位置xn(t+1)；
    将人工移动智能体的位置更新规则的表达式为：
    xn(t+1)＝xn(t)+vn(t+1)   (8)；
    其中，xn(t+1)表示在本车n下一时刻(t+1)的位置，xn(t)表示本车n在当前时刻t的位置，vn(t)表示本车n在当前时刻t的速度。
  8. 根据权利要求5所述的异构移动智能体协同控制与演化方法，其特征在于，所述自动移动智能体的演化规则包括加速规则、减速规则和位置更新规则；
    其中，将自动移动智能体的加速规则表示为：
    其中，vn(t+1)表示本车n在当前时刻t的速度，νn(t)表示本车n在当前时刻t的速度，，vmax表示道路的最大限速，a表示根据双积分运动方程进行计算得到的本车n的加速度，m表示本车n的质量，dsafe表示安全跟驰距离，F表示总虚拟力。
    自动移动智能体的减速规则为：当前车n+1虚拟势场产生的斥力大于引力时，本车n所受的总虚拟力表现为斥力；
    自动移动智能体的的位置更新规则与人工移动智能体的位置更新规则一致。
  9. 根据权利要求1所述的异构移动智能体协同控制与演化方法，其特征在于，通过场力规划方法控制自动移动智能体运动以及运动路径的方法，包括：
    根据改进的相互作用势能场函数，计算每个自动移动智能体受到的来自其他智能体的虚拟力，包括位置虚拟力、速度虚拟力和加速度虚拟力；
    将每个自动移动智能体受到的所有虚拟力进行矢量合成，得到每个自动移动智能体的总虚拟力；
    利用总虚拟力作为控制信号，通过控制算法调整自动移动智能体的行驶速度和方向；
    利用势能场虚拟力规划自动移动智能体的行驶路径。
  10. 根据权利要求9所述的异构移动智能体协同控制与演化方法，其特征在于，所述位置虚拟力包括位置虚拟引力和位置虚拟斥力，所述速度虚拟力包括速度虚拟引力和速度虚拟斥力，所述加速度虚拟力包括加速度虚拟引力和加速度虚拟斥力；
    将所述位置虚拟斥力Frep-d表示为：
    其中，krep-d表示位置虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示自动联网车辆到目标点的距离向量，表示自动联网车辆到目标点的距离向量的模长。
    将所述位置虚拟引力Fatt-d表示为：
    Fatt-d＝katt-drg   (11)；
    其中，katt-d表示位置虚拟引力势能系数，rg表示自动联网车辆到目标点的距离。
    将所述速度虚拟斥力Frep-v表示为：
    其中，krep-v表示速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的速度差向量，表示前车与本车的速度差向量的模长。
    将所述速度虚拟引力Fatt-v表示为：
    Fatt-v＝tanh(katt-v||Δve||)rg (13)；
    其中，katt-v表示速度虚拟引力势能系数，Δve表示期望速度和当前速度的速度差，rg表示自动联网车辆到目标点的距离，tanh表示双曲正切函数，||Δve||表示期望速度和当前速度的速度差向量模长。
    将所述加速度虚拟斥力Frep-a表示为：
    其中，krep-a表示加速度虚拟斥力势能系数，r表示自动联网车辆到目标点的距离，r0表示排斥势场影响范围，表示前车与本车的加速度差值向量，表示前车与本车的加速度差向量的模长。
    将所述加速度虚拟引力Fatt-a表示为：
    Fatt-a＝tanh(katt-a||Δae||)rg；   (15)；
    其中，katt-a表示加速度虚拟引力势能系数，Δae表示期望加速度和当前加速度差值，rg表示自动联网车辆到目标点的距离，||Δae||表示期望加速度和当前加速度差值向量的模长。
PCT/CN2024/136661 2024-07-31 2024-12-04 一种异构移动智能体协同控制与演化方法一种异构移动智能体协同控制与演化方法 Pending WO2026025751A2 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202411040032.4 2024-07-31
CN202411040032.4A CN119165795B (zh) 2024-07-31 2024-07-31 一种异构移动智能体协同控制与演化方法

Publications (1)

Publication Number Publication Date
WO2026025751A2 true WO2026025751A2 (zh) 2026-02-05

Family

ID=93877559

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/136661 Pending WO2026025751A2 (zh) 2024-07-31 2024-12-04 一种异构移动智能体协同控制与演化方法一种异构移动智能体协同控制与演化方法

Country Status (2)

Country Link
CN (1) CN119165795B (zh)
WO (1) WO2026025751A2 (zh)

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090306946A1 (en) * 2008-04-08 2009-12-10 Norman I Badler Methods and systems for simulation and representation of agents in a high-density autonomous crowd
CN102693550A (zh) * 2011-03-25 2012-09-26 上海日浦信息技术有限公司 虚拟场景中的智能群体运动模拟方法
CN110991120A (zh) * 2019-11-18 2020-04-10 郑州大学 一种排队场景下人群运动建模仿真方法
CN114613179B (zh) * 2021-12-22 2023-03-21 江苏大学 网联自动驾驶车辆混行交叉口集聚通行方法及其控制系统
CN114564567B (zh) * 2022-01-20 2026-03-17 浙江工业大学 基于用户行为的人工智能体的设计方法及系统

Also Published As

Publication number Publication date
CN119165795B (zh) 2025-10-21
CN119165795A (zh) 2024-12-20

Similar Documents

Publication Publication Date Title
Chen et al. Stabilization approaches for reinforcement learning-based end-to-end autonomous driving
CN113110592B (zh) 一种无人机避障与路径规划方法
Sun et al. Crowd navigation in an unknown and dynamic environment based on deep reinforcement learning
CN109543285B (zh) 一种融合数据驱动与强化学习的人群疏散仿真方法和系统
CN108491598A (zh) 一种基于路径规划的人群疏散仿真方法及系统
Yuan et al. Multi-reward architecture based reinforcement learning for highway driving policies
CN112183288A (zh) 一种基于模型的多智能体强化学习方法
CN112348285A (zh) 一种基于深度强化学习的动态环境下人群疏散模拟方法
CN116027788A (zh) 融合复杂网络理论和部分可观察马尔可夫决策过程的智能驾驶行为决策方法及设备
CN111339668B (zh) 基于情绪认知的人群疏散可视化方法
CN118560530B (zh) 一种基于生成对抗模仿学习的多智能体驾驶行为建模方法
CN119975359A (zh) 基于多智能体的多车道匝道合流区车辆控制方法及系统
CN116859931A (zh) 轨迹规划模型的训练方法、车辆控制方式及装置
Wu et al. Emotion contagion model for dynamical crowd path planning
Jiang et al. A deep-learning-based approach for simulating pedestrian turning flow
Wang et al. Intelligent vehicle driving decision-making model based on variational AutoEncoder network and deep reinforcement learning
CN117193355B (zh) 一种基于深度强化学习的无人机自主导航及避障方法
CN118349031A (zh) 融合先验策略的改进td3算法的无人机航迹规划方法
CN118627662A (zh) 驾驶场景的预测方法和装置、设备、存储介质及程序产品
CN120927321B (zh) 基于世界模型的端到端自动驾驶系统闭环对抗测试方法
CN113353102B (zh) 一种基于深度强化学习的无保护左转弯驾驶控制方法
Choi et al. Enhancing autonomous driving with pre-trained imitation and reinforcement learning
Li et al. Lane-change decision of automatic driving based on reinforcement learning framework
CN119165795B (zh) 一种异构移动智能体协同控制与演化方法
Youssef et al. Deep reinforcement learning with external control: Self-driving car application

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24949469

Country of ref document: EP

Kind code of ref document: A2