Artificial emotional potential field path planning algorithm

Abstract

The artificial emotion potential field path planning algorithm relates to the application field of path planning of mobile service robots, and the planned path can consider the real experience and emotional feeling of the user in the service process of the old-age service robot. The application proposes an algorithm based on a mathematical model, which can solve the problem that the path planning of the mobile service robot cannot take care of the user's emotion in a specific situation, and the planned path has more advantages in the emotional experience of the user than other methods. The application comprises the following steps: step one: design a human-computer interaction experiment to obtain a distance emotion model; step two: establish an artificial emotion potential field model; obtain the position of the robot at t-1 time and the emotional state (valence, arousal) of the person, and the position of the robot at t time and the emotional state (valence, arousal) of the user; step three: robot path planning process.

Description

TECHNICAL FIELD

[0001] The present application relates to the path planning application field of a mobile service robot, and the planned path can consider the real experience and emotional feeling of a user in a service process of a pension service robot. BACKGROUND

[0002] Under the background of an increasing number of elderly population, a shortage of nursing staff for the elderly, and insufficient supply of medical service resources, with the continuous development of artificial intelligence technology and service robot technology, it is an inevitable trend to use a service robot to replace a human being to take care of the elderly. Existing mobile service robots, such as an intelligent wheelchair, can only complete simple escort work, and the working mode is stiff and mechanical. The user feels like being served as goods, completely lacks basic interaction and communication, and does not have the perception and processing of the emotional state of the user. Therefore, it is becoming more and more important to strengthen the perception function of the mobile service robot to the user's emotion and psychology, and it is very meaningful to study a path planning method based on the emotional state of the user.

[0003] The core of a mobile robot navigation system is a path planning algorithm. Existing path planning algorithms focus on improving and strengthening the planning and adaptation ability of a mobile robot in a more complex environment.

[0004] An artificial potential field (APF) is an excellent algorithm combining computer technology and physical principles, which was first proposed by Khatib in 1994. See the document Khatib O. Real-time obstacle avoidance for manipulators and mobile robots [C] / / Proceedings. 1985 IEEE International Conference on Robotics and Automation. IEEE, 1985, 2: 500-505.

[0005] The artificial potential field method is mainly used to solve the obstacle avoidance problem of a mechanical arm in the early stage, and is increasingly applied to various path planning problems. In this algorithm, a virtual attractive field existing in the overall environment is added to the target point to be reached in path planning, and a virtual repulsive field within a certain range is added to all obstacles in the environment. When the mobile robot moves in the environment, it will be affected by the comprehensive effect of the attractive field of the target point and the repulsive field of various obstacles in the environment. The motion direction of the robot is the gradient direction under the influence of the resultant field generated by the interaction of the two force fields. When the robot is closer to the target point, the influence of the attractive field is greater and greater, and finally reaches the target point.

[0006] For the mobile robot in the service robot category that needs to be manned to complete the service task, the service experience and emotional feeling of the user in the service process are also important. On the basis of the obstacle and target point emotional distance model, by improving the artificial potential field method, the artificial emotional potential field path planning algorithm (AEPF) is proposed to solve the defect that the path planned by the artificial potential field method cannot consider the user's emotion.

[0007] The expression of the user's emotional state (valence, arousal) is based on the two-dimensional valence-arousal emotional model proposed by Russell in 1980. The model divides emotions into two dimensions, namely valence dimension (Valance) and arousal dimension (Arousal). The negative half of the valence dimension represents negative emotions, and the positive half represents positive emotions. The negative half of the arousal dimension represents calm emotions, and the positive half represents intense emotions. For details, please refer to Russell J A.A circumplex model of affect[J].Journal of personality and social psychology,1980,39(6):1161. In order to use (v t ,a t ) to represent the emotional state of the user and quantify the emotional state of the user, we use a SAM self-assessment scale based on a two-dimensional valence-arousal model to quantify it. The model uses numbers 1-9 to represent valence and arousal, and each emotional state can be represented by a (valence, arousal) numerical pair. For details, please refer to Bradley M M, Lang P J.Measuring emotion: the self-assessment manikin and the semantic differential[J].Journal of behavior therapy and experimental psychiatry,1994,25(1):49-59. SUMMARY

[0008] To solve the above problems, the present application proposes an algorithm based on a mathematical model, which can solve the problem that the path planning of the mobile service robot in a specific situation cannot consider the user's emotion, and the planned path has more advantages in user emotional experience than other methods.

[0009] To achieve the above purpose, the technical scheme adopted by the present application is as follows: the present application comprises the following steps:

[0010] Step one: design human-computer interaction experiment to obtain distance emotion model;

[0011] Step two: establish artificial emotion potential field model; obtain the position of the robot at time t-1 and the emotional state (valence, arousal) of the person, as well as the position of the robot at time t and the emotional state (valence, arousal) of the user;

[0012] Step three: robot path planning process; set the robot motion step l, the step l is greater than 0 and less than the Euclidean distance from the starting point to the target point, the step affects the distance of the robot advancing each time; at time t, the robot advances a step l in the direction of the resultant force F. t

[0013] As a preferred solution, the step one of the present application is based on an immersive virtual reality emotion induction platform, which builds an interactive and immersive 3D scene to induce the emotions of the subjects in a single external factor; the immersive virtual reality emotion induction platform includes a head-mounted VR display device and a smart wheelchair with serial communication function, and the display content of the head-mounted VR display device uses Unity3D integrated game development engine for dynamic visualization of the scene; during the experiment, the subject wears VR glasses, and the smart wheelchair carries the passenger to the end point;

[0014] Combining the two-dimensional valence-arousal emotion model, the SAM self-assessment scale results of the subjects under the influence of the single external emotion factor are obtained while inducing the emotions of the subjects; the distance is used as a variable to perform experiments, and the rules of the emotional state of the person affected by the obstacle at different distances and the rules of the emotional state of the person affected by the target point at different distances are obtained; the experimental variable of the obstacle distance experiment is to control the change of the closest distance d obs of the smart wheelchair and the obstacle, and the obstacle distance emotion model is obtained after curve fitting of the experimental results:

[0015]

[0016]

[0017] The experimental variable of the target point distance experiment is to control the change of the straight line distance d goal of the smart wheelchair and the target point, and the target point distance emotion model is obtained after curve fitting of the experimental results:

[0018] v goal = 6.65-0.98 ln d goal

[0019] a goal = 4.01-1.55 ln d goal .

[0020] ​As another preferred solution, in step two, the position information of the robot is the physical position coordinates of the robot at the current time, which is represented by (x t ,y t ) coordinates in a two-dimensional plane;

[0021] At time t-1, the robot carrying the user moves from x t-1 to x t , the distance between the robot and the obstacle changes from d obs,t-1 to d obs,t , and the valence of the user's emotional state changes from v t-1 to v t ;

[0022] The expression of the emotional potential field around the obstacle is defined as:

[0023]

[0024] where k r1 and k r2 are positive proportional gain coefficients, d goal,t and d goal,t-1 are the Euclidean distances between the robot and the obstacle at time t and t-1, respectively, d goal,t is the Euclidean distance between the robot and the target point at time t, v t and a t are the valence and arousal of the user at time t, and v t-1 and a t-1 are the valence and arousal of the user at time t-1;

[0025] The expression of the emotional potential field of the target point is:

[0026]

[0027] where k a1 and k a2 are positive proportional gain coefficients, v t and a t are the valence and arousal at time t, respectively, v t-1 is the valence at time t-1, and d goal,t is the Euclidean distance between the robot and the target point at time t;

[0028] The total potential field of the artificial emotional potential field method is the superposition of the obstacle emotional potential field and the target point emotional potential field, and the calculated resultant force determines the next movement direction of the robot;

[0029] The size of the resultant force is represented as:

[0030]

[0031] where F tF represents the size of the resultant force suffered by the robot at time t rep,t F represents the size of the obstacle potential field force suffered by the robot at time t att,t U represents the size of the target point potential field force suffered by the robot at time t rep,t U represents the obstacle potential field around the robot at time t att,t U represents the target point potential field around the robot at time t.

[0032] Advantages of the present application.

[0033] The path planning algorithm plans a path, which not only needs to safely and reasonably carry the user to reach the target point, but also needs to consider the emotional experience of the user. In the process of the robot from the starting point to the target point, the path corresponding to the user emotional valence planned by the artificial emotional potential field method is higher than that of the artificial potential field method, and the fluctuation is more gentle; the user emotional arousal is lower than that of the artificial potential field method, and the fluctuation is more gentle. From the perspective of the change of the emotional state of the user, it is explained that the result planned by the artificial emotional potential field method is more in line with the emotional and psychological feelings of the user than that of the artificial potential field method. BRIEF DESCRIPTION OF DRAWINGS

[0034] The present application will be further described below in combination with the drawings and specific embodiments. The protection scope of the present application is not limited to the following content.

[0035] Figure 1 The structure diagram of the immersive virtual reality experiment platform of the present application.

[0036] Figure 2 The display scene schematic diagram of the head-mounted VR display device of the present application.

[0037] Figure 3 The obstacle distance emotional model experiment design diagram of the present application.

[0038] Figure 4 The SAM emotional self-assessment scale of the present application.

[0039] Figure 5 The target point distance emotional model experiment design diagram of the present application.

[0040] Figure 6 The basic information required by the present application.

[0041] Figure 7 The flowchart of step three of the present application.

[0042] Figure 8 The robot path planning task schematic diagram in the example of the present application.

[0043] Figure 9Results of path planning of the artificial emotional potential field algorithm and the artificial potential field algorithm for the robot in the examples of the present application.

[0044] Figure 10 Comparison of results of path planning of the artificial emotional potential field algorithm and the artificial potential field algorithm for the valence change of the emotional state of the user in the examples of the present application.

[0045] Figure 11 Comparison of results of path planning of the artificial emotional potential field algorithm and the artificial potential field algorithm for the arousal change of the emotional state of the user in the examples of the present application. DETAILED DESCRIPTION

[0046] The artificial potential field algorithm introduces target point potential field and obstacle potential field on the basis of physical principles, and the robot changes its movement direction under the joint action of the two potential fields. The human service robot needs to avoid obstacles while safely reaching the target point in the process of completing the task requirements of the user, so the influencing factors of the emotional state of the user mainly include the influence of the target point and the obstacle. In the artificial emotional potential field algorithm, not only the virtual physical potential field for ensuring the safe movement of the robot without collision is introduced, but also the virtual emotional potential field for ensuring the comfort and peace of mind of the user is introduced. The movement direction and behavior of the robot are the result of the joint action and influence of the physical field and the emotional field of the obstacle and the target point.

[0047] To achieve the above object, the present application uses the following technical scheme, which comprises the following steps:

[0048] Step 1: Design a human-computer interaction experiment to obtain a distance emotion model

[0049] Emotion modeling refers to describing the emotional state by establishing a mathematical model, and classifying or even quantifying the emotional state from a scientific point of view.

[0050] Based on the immersive virtual reality emotion induction platform, an interactive and immersive 3D scene is built to induce the emotions of the subjects in a single external factor. As shown in Figure 1 , the immersive virtual reality emotion induction platform mainly consists of a head-mounted VR display device with a type of Oculus rift cv1 and a smart wheelchair with serial communication function. As Figure 2 , the display content of the head-mounted VR display device uses the Unity3D integrated game development engine to realize dynamic visualization of the scene. During the experiment, the subject wears VR glasses, and the smart wheelchair carries the passenger to the destination.

[0051] Combined with the two-dimensional valence-arousal emotional model, while inducing the subjects' emotions, the results of the SAM self-assessment scale under the influence of this single external emotional factor were obtained. Experiments were conducted with distance as a variable to obtain the patterns of how obstacles affect people's emotional state at different distances, and how the target point affects people's emotional state at different distances. Figure 3 The experimental variable of the obstacle distance experiment is to control the shortest distance d between the smart wheelchair and the obstacle. obs The experiment set up 6 groups with a distance of d obs The settings are: 0.2m, 0.4m, 0.6m, 0.8m, 1.0m, 1.2m. After each experiment, the subjects were asked to fill in the SAM self-assessment form. The filling process is as follows Figure 5 As shown. After curve fitting the experimental results, the obstacle distance emotion model is obtained:

[0052]

[0053]

[0054] like Figure 6 The experimental variable of the target point distance experiment is to control the straight-line distance d between the smart wheelchair and the target point. goal The experiment set up 4 groups with a distance of d goal The time intervals are set as follows: 1m, 2m, 3m, and 4m. After each experiment, the subjects were asked to fill in the SAM self-assessment form. The filling process is as follows: Figure 5 As shown. After curve fitting the experimental results, the target point distance emotion model is obtained:

[0055] v goal =6.65-0.98 ln d goal

[0056] a goal =4.01-1.55 ln d goal

[0057] Step 2: Establish an artificial emotion potential field model:

[0058] This solution is based on obtaining the robot's position at time t-1 and the user's emotional state (valence, arousal), as well as the robot's position at time t and the user's emotional state (valence, arousal). Therefore, before proceeding with the following steps, it is necessary to obtain the above information. Among the information obtained, the robot's position information is the robot's current physical position coordinates, which are used here on the two-dimensional plane (x t ,y t ) coordinate representation.

[0059] At time t-1, the robot carries the user from xt-1 Move to x t , the distance between the robot and the obstacle is d obs,t-1 Change to d obs,t， The valence of the user's emotional state changes from v t-1 Change to v t .

[0060] The expression for defining the emotional potential field around an obstacle is:

[0061]

[0062] Among them, k r1 With k r2 is the proportional gain coefficient, d goal,t with d goal,t-1 are the Euclidean distances between the robot and the obstacle at time t and time t-1, respectively, goal,t is the Euclidean distance between the robot and the target point at time t, v t with a t is the user valence and arousal at time t, v t-1 with a t-1 is the user valence and arousal at time t-1.

[0063] When a robot carries a user toward a target point, within the target point's emotional influence range, the user's emotions will change due to the motivation of the task progress. To comply with this objective law, an emotion-related potential field component with a certain range is superimposed on the target point's physical potential field, which exists throughout the entire environment. This guides the robot to more quickly move toward the target point based on the user's emotional feedback and complete the user's task.

[0064] The expression of the target point emotional potential field is:

[0065]

[0066] where k a1 With k a2 is the proportional gain coefficient, v t with a t are the valence and arousal at time t, v t-1 is the potency at time t-1, d goal,t is the Euclidean distance between the robot and the target point at time t.

[0067] The total potential field of the artificial emotional potential field method is the superposition of the obstacle's emotional potential field and the target point's emotional potential field. The calculated resultant force determines the robot's next movement direction.

[0068] The magnitude of the resultant force can be expressed as:

[0069]

[0070] wherein, F t represents the size of the force received by the robot at time t, F rep,t represents the size of the obstacle potential field force received by the robot at time t, F att,t represents the size of the target point potential field force received by the robot at time t, U rep,t represents the obstacle potential field around the robot at time t, U att,t represents the target point potential field around the robot at time t.

[0071] Step three: robot path planning process

[0072] Set the robot motion step length l, the step length l is greater than 0 and less than the Euclidean distance from the starting point to the target point, the step length affects the distance of the robot advancing each time. At time t, the robot advances a step length l in the direction of the resultant force F t .

[0073] In step one, we have obtained the obstacle distance emotion model and the target point distance emotion model, which can be brought into the model according to the distance d obs of the robot from the nearest obstacle and the distance d goal of the robot from the target point to obtain the emotional state of the user.

[0074] In step two, we have established an artificial emotional potential field model. In different navigation environments, we first obtain information such as Figure 6 , and use these basic information to perform the path planning process according to step two Figure 7 .

[0075] Figure 8 The path planning simulation environment diagram of the present application example is shown in the following figure. In a two-dimensional map, the artificial emotional potential field method and the artificial potential field method are simulated for path planning. The parameters common to different methods remain the same. The simulation environment is shown in the following figure. The starting point is represented by “●”, the obstacle is represented by “█”, and the target point is represented by “★”.

[0076] The present study first designs a simulation experiment to verify the basic functions of the artificial emotional potential field method, and observes the ability of the method to complete tasks and avoid obstacles in the simulation environment. In a 10x10 unit area two-dimensional plane, six obstacles are set to be irregularly distributed between the starting point and the target point. The starting point coordinates are (0, 0), the target point coordinates are (10, 10), and the distance of the robot advancing each time is set to 0.01 units.

[0077] The artificial potential field method and the artificial emotional potential field method are used to test the simulation environment.

[0078] When the robot is at position x tThe obstacle distance mood model established by step one:

[0079]

[0080]

[0081] The target point distance mood model established by step one:

[0082] v goal,t = 6.65-0.98 ln d goal,t

[0083] a goal,t = 4.01-1.55 ln d goal,t

[0084] Therefore, the corresponding mood state is v t , a t :

[0085] v t = v obs,t + v goal,t

[0086] a t = a obs,t + a goal,t

[0087] In the manner described in step three, Figure 2 The described process carries out the path planning task. The present application uses Python language-based simulation for illustration, and the software is Pycharm.

[0088] The path planning task starts, the initial coordinates of the robot is x0=(0,0), the target point coordinates is x goal =(10,10), the obstacle coordinates is x obs1 =(1,1), x obs2 =(2.5,2.8), x obs3 =(4.2,1.6), x obs4 =(5.5,4.3), x obs5 =(4,5.8), x obs6 =(8,2.1).

[0089] The parameters of the artificial mood potential field method are initialized as k r1 = 50, k r2 = 1, k a1 = 15, k a2 = 10, R obs = 1, R goal= 4, l = 0.01. The robot advances one step l in a random direction, and the position is x1 = (0.01, 0). It is calculated whether the Euclidean distance between the position x1 and the position of the target point (10, 10) is less than 0.2 units of length, and it is determined whether the target point is reached. If it is less than 0.2, it is determined that the robot has reached the target point, and the navigation is ended. If it is greater than 0.2, it is determined that the target point has not been reached, and the distance between the robot positions x0, x1 and the nearest obstacle and the distance to the target point are continued to be calculated, that is, The obstacle distance mood model and the target point mood model are brought in, that is,

[0090]

[0091]

[0092] v goal,0 = 6.65 - 0.98 ln d goal,0 = 4.050764262524225

[0093] a goal,0 = 4.01 - 1.55 ln d goal,0 = -0.10839307858328517

[0094] v0 = v obs,0 + v goal,0 = 9.739715429963313

[0095] a0 = a obs,0 + a goal,0 = 0.5867329259696277

[0096] Similarly, it can be calculated that

[0097] v1 = v obs,1 + v goal,1 = 9.73978568735224

[0098] a1 = a obs,1 + a goal,1 = 0.5868441802136902

[0099] The following function is defined in python to calculate the angle between the position vector and the positive direction of the x-axis.

[0100]

[0101] The angle between the position vector and the x-axis between the nearest obstacle position and the robot position x0, x1 is calculated and denoted as θ0bs,x1 ,

[0102] The repulsion, attraction and resultant force can be calculated by bringing the above information into the artificial emotional potential field model.

[0103] U rep,1 = 0, F1 = 1498.5007499999997.

[0104] The angle between the resultant force and the positive direction of the x-axis Calculate the next position

[0105] Similarly, the values of x3, x4… are calculated, and the navigation task is stopped when the robot reaches the target position.

[0106] Analysis shows that the artificial emotional potential field method can plan a reasonable path in a complex environment with multiple obstacles. When the parameters of this method are the same as those of the artificial potential field method, the path planned by this method tends to "avoid" obstacles due to the presence of emotional components in the potential field of obstacles and target points, and the total number of travel steps is more.

[0107] Path planning algorithms plan paths that not only need to safely and reasonably carry users to the target point, but also need to consider the emotional experience of users. Based on the distance-emotion model, the valence changes in the emotional change process of the paths obtained by different planning algorithms are as follows Figure 10 , and the arousal changes are as follows Figure 11 .

[0108] During the process of the robot from the starting point to the target point, the valence of the user's emotion corresponding to the path planned by the artificial emotional potential field method is higher than that of the artificial potential field method, and the fluctuation is more stable; the user's emotional arousal is lower than that of the artificial potential field method, and the fluctuation is more stable. From the perspective of the change of the user's emotional state, it is shown that the result planned by the artificial emotional potential field method is more in line with the user's emotions and psychological feelings than that planned by the artificial potential field method.

[0109] It can be understood that the above specific description of the present application is only used to illustrate the present application and is not limited to the technical solutions described in the embodiments of the present application. Those skilled in the art should understand that the present application can still be modified or replaced equivalently to achieve the same technical effect; as long as the use needs are met, it is within the protection scope of the present application.

Claims

1. An artificial emotional potential field path planning algorithm, characterized in that The method comprises the following steps: Step 1: design human-computer interaction experiment to obtain distance emotion model; Step 2: establish artificial emotion potential field model; obtain the position of the robot at t-1 moment and the emotion state of the person, and the position of the robot at t moment and the emotion state of the user; Step 3: robot path planning process; Setting a robot motion step length l, the step length l takes a value greater than 0 and less than the Euclidean distance from the starting point to the target point, the step length affects the distance of the robot advancing each time; at time t, the robot advances a step length l towards the direction of the resultant force F t . The step 1 is based on an immersive virtual reality emotion induction platform, builds an interactive and immersive 3D scene, and performs experiments in the manner of inducing emotions of subjects under a single external factor; the immersive virtual reality emotion induction platform comprises a head-mounted VR display device and an intelligent wheelchair with serial communication function; the display content of the head-mounted VR display device uses Unity3D integrated game development engine to dynamically visualize the scene; during the experiment, the subject wears VR glasses, and the intelligent wheelchair carries the passenger to the terminal point; Combined with the two-dimensional valence-arousal emotion model, while inducing the subjects' emotions, the results of the SAM self-assessment scale under the influence of this single external emotional factor were obtained; the experiment was conducted with distance as the variable to obtain the rules of how people's emotional state is affected by obstacles at different distances and how people's emotional state is affected by target points at different distances; the experimental variable of the obstacle distance experiment is to control the closest distance d between the smart wheelchair and the obstacle. obs The obstacle distance emotion model is obtained by curve fitting the experimental results: The experimental variable of the target point distance experiment is to control the straight-line distance d between the smart wheelchair and the target point. goal The target point distance emotion model is obtained by curve fitting the experimental results: v goal = 6.65 - 0.98ln d goal a goal = 4.01 - 1.55 lnd goal ; In the second step, the position information of the robot is the physical position coordinate of the robot at present, which is expressed by (x t ,y t ) coordinate in two-dimensional plane. At time t-1, the robot carrying the user moves from x t-1 to x t , the distance between the robot and the obstacle changes from d obs,t-1 to d obs,t , and the valence in the emotional state of the user changes from v t-1 to v t ; The expression of the emotion potential field around the obstacle is defined as: Among them, k r1 With k r2 is the proportional gain coefficient, d goal,t with d goal,t-1 are the Euclidean distances between the robot and the obstacle at time t and time t-1, respectively, goal,t is the Euclidean distance between the robot and the target point at time t, v t with a t is the user valence and arousal at time t, v t-1 with a t-1 is the user valence and arousal at time t-1; The expression of the emotion potential field of the target point is: where k a1 With k a2 is the proportional gain coefficient, v t with a t are the valence and arousal at time t, v t-1 is the potency at time t-1, d goal,t is the Euclidean distance between the robot and the target point at time t; The total potential field of the artificial emotion potential field method is the superposition of the emotion potential field of the obstacle and the emotion potential field of the target point, and the resultant force calculated determines the movement direction of the robot in the next step; The size of the resultant force is represented as: Wherein, F t represents the size of the force suffered by the robot at time t, F rep,t represents the size of the obstacle potential field force suffered by the robot at time t, F att,t represents the size of the target point potential field force suffered by the robot at time t, U rep,t represents the obstacle potential field around the robot at time t, U att,t represents the target point potential field around the robot at time t.

Images