JP7519337B2 - Information processing device, information processing method, and information processing program - Google Patents

Description

The present disclosure relates to an information processing device, an information processing method, and an information processing program.

There is a traffic flow simulation device that simulates the traffic flow of vehicles traveling on a road. The simulation device simulates traffic flow taking into account the characteristics of the vehicle and the individual identity of the driver of the vehicle. The driver's individual characteristics indicate the driver's tendency regarding recognition and judgment, and have parameters such as stop sign recognition degree, stop sign passing degree, judgment accuracy, desired speed, accelerator opening, and stopping position relative to the stop line. In other words, the simulation device adjusts the driver's susceptibility to human error and changes the vehicle's movement by setting each parameter of the driver's individual characteristics (see Patent Document 1).

JP 2007-080216 A

The simulation device described in Patent Document 1 allows each parameter of the driver's personal characteristics to be set arbitrarily. In this case, it is not possible to set what kind of human characteristics each model parameter will have as a whole, and there is a possibility that the simulated person's image will be an image of a person that is not normally possible. As an example, it is thought that there is a correlation between the degree of compliance with traffic lights and the degree of compliance with stop signs in the real world. However, because the simulation device described in Patent Document 1 allows each parameter of the driver's personal characteristics to be set arbitrarily, there is a possibility that the two example parameters mentioned above will be set to values that have a low correlation with each other.

The present disclosure provides an information processing device, an information processing method, and an information processing program capable of estimating parameters related to a person's behavior.

An information processing device according to one embodiment includes an acquisition unit that acquires behavioral information on a person's behavior and attribute information on the person's attributes, a classification unit that performs classification according to the tendencies of the person's behavior based on the behavioral information acquired by the acquisition unit, a first learning unit that learns the association between the person's attributes and behavioral tendencies based on the attribute information acquired by the acquisition unit and the behavioral tendencies classified by the classification unit, and generates a first trained model, and a second learning unit that learns the association between the behavioral tendencies classified by the classification unit and parameters used in a behavior estimation model that estimates the person's behavior according to the tendencies, and generates a second trained model.

According to one aspect, behavioral information on a person's behavior and attribute information on the person's attributes are acquired, the behavior of the person is classified according to the tendency of the person's behavior based on the behavioral information, the association between the person's attributes and the tendency of the behavior based on the attribute information and the tendency of the behavior according to the classification is learned to generate a first trained model, and the association between the tendency of the behavior according to the classification and parameters used in a behavior estimation model that estimates the behavior of the person according to that tendency is learned to generate a second trained model, so that parameters related to the behavior of the person can be estimated.

FIG. 1 is a diagram illustrating an information processing device according to an embodiment. FIG. 1 is a block diagram illustrating an information processing device according to an embodiment. FIG. 1 is a diagram for explaining the behavior of a person. FIG. 11 is a diagram for explaining an example of selected attribute items. FIG. 1 is a diagram for explaining an information processing method according to an embodiment. 1 is a flowchart illustrating an information processing method according to an embodiment, which is a method of advance preparation when estimating a person's behavior. 1 is a flowchart illustrating an information processing method according to an embodiment, which is a method for estimating a person's behavior.

One embodiment is described below.

[Overview of information processing device 100]
First, an overview of an information processing device 100 according to an embodiment will be described.
FIG. 1 is a diagram illustrating an information processing device 100 according to an embodiment.

The information processing device 100 may be configured, for example, as a learning device that generates a trained model used to estimate a person's behavior. The information processing device 100 may be configured, for example, as a parameter specifying device that specifies parameters of a model (behavior estimation model) that estimates a person's behavior using the trained model. The information processing device 100 may be configured, for example, as a human behavior estimation device that estimates a person's behavior using the specified parameters.
The information processing device 100 may be configured as, for example, a setting device that sets various parameters related to a person (traffic participant) when simulating the movement of the person (traffic participant) on a road, an intersection, etc. The traffic participant may be, for example, a moving object that appears when simulating a vehicle driven by a person, a motorcycle, a pedestrian, an autonomous vehicle, etc.

The information processing device 100 may be configured as a computer such as a server, desktop, laptop, or tablet.

First, the information processing device 100 performs advance preparations for estimating a person's behavior (preparation before executing a simulation).
The information processing device 100 acquires behavioral information related to a person's behavior and attribute information related to the person's attributes. The information processing device 100 performs classification according to the person's behavioral tendency based on the behavioral information. That is, the information processing device 100 classifies the person's behavior based on the behavioral information into which of a plurality of behavioral tendencies it corresponds. The information processing device 100 learns the association between the person's attributes based on the attribute information and the behavioral tendencies classified as described above, and generates a first trained model. The information processing device 100 learns the association between the person's behavioral tendency classified as described above and parameters of a behavior estimation model that estimates the person's behavior, and generates a second trained model.

Next, the information processing device 100 performs person behavior estimation (executes a simulation).
For example, when estimating a behavioral tendency of a person using the first learned model, the information processing device 100 selects one of items (multiple attribute items) related to the attributes of the person used in the first learned model. The information processing device 100 estimates the behavioral tendency of the person according to the attribute item based on the attribute item selected as described above and the first learned model. The information processing device 100 determines parameters (parameters used in the behavior estimation model) according to the behavioral tendency of the person based on the estimated behavioral tendency and the second learned model. The information processing device 100 inputs (sets) the parameters determined as described above into the behavior estimation model to estimate the behavior of the person.

[Details of information processing device 100]
Next, details of the information processing device 100 according to an embodiment will be described.
FIG. 2 is a block diagram for explaining the information processing device 100 according to an embodiment.

The information processing device 100 includes a communication unit 121, a storage unit 122, a display unit 123, and a control unit 110. The control unit 110 includes, for example, an acquisition unit 111, a classification unit 112, a first learning unit 113, a second learning unit 114, a first identification unit 115, a second identification unit 116, and a behavior estimation unit 117. The control unit 110 may be configured, for example, by a calculation processing unit of the information processing device 100. The control unit 110 (for example, a calculation processing unit, etc.) may realize the functions of each unit (for example, the acquisition unit 111, the classification unit 112, the first learning unit 113, the second learning unit 114, the first identification unit 115, the second identification unit 116, and the behavior estimation unit 117, etc.) by, for example, appropriately reading and executing various programs, etc. stored in the storage unit 122, etc.

The communication unit 121 can, for example, send and receive various information to and from devices (external devices) external to the information processing device 100.

The storage unit 122 may store, for example, various information and programs. Examples of the storage unit 122 may be a memory, a solid state drive, a hard disk drive, etc.

The display unit 123 is capable of displaying, for example, various characters, symbols, images, etc.

Next, we will explain the functional block that performs the advance preparations for estimating a person's behavior (preparation before running a simulation).

The acquisition unit 111 acquires behavioral information related to the behavior of a person and attribute information related to the attributes of the person. The acquisition unit 111 may acquire behavioral information and attribute information from an external device (such as an external server) (not shown) outside the information processing device 100 via the communication unit 121, for example. Alternatively, when an external memory (not shown) in which behavioral information and attribute information generated by an external device is recorded is connected to the information processing device 100, the acquisition unit 111 may acquire behavioral information and attribute information from the external memory. Alternatively, when a person's behavior and attributes are input by operating an input unit (such as a keyboard and mouse) (not shown) arranged in the information processing device 100, the acquisition unit 111 may acquire behavioral information and attribute information based on the input.

The acquisition unit 111 acquires behavior (behavior information) and attributes (attribute information) for the same person. The acquisition unit 111 may acquire behavior (behavior information) and attributes (attribute information) for each of multiple people.

Figure 3 is a diagram to explain a person's actions.

The behavior of a person may be, for example, the behavior of the person depending on the situation around the person. As an example, as illustrated in FIG. 3, the behavior of the person 201 may be, for example, the behavior of the person 201 depending on the driving conditions of the vehicle 202 traveling on the road 203 and an intersection. As a more specific example, the behavior of the person 201 may be, when the person 201 crosses the road 203, whether or not the person 201 crosses the road 203 depending on the position (driving conditions) of the vehicle 202 traveling on the road 203. The behavior of the person 201 may be estimated, for example, by reproducing the road 203 on which the vehicle 202 travels using a vehicle simulator or the like and testing the behavior of the person 201 depending on the driving conditions of the vehicle 202.

In other words, the acquisition unit 111 may acquire, as behavior information, information on the behavior of a plurality of people regarding the behavior of a person in a situation in which there is a person and a vehicle traveling around the person. As an example of this case, the acquisition unit 111 may acquire behavior information including an elapsed time (measurement time), the position of the person relative to a road (intersection), etc., the speed at which the person moves, the position of the vehicle traveling on the road (intersection), and the speed and acceleration of the vehicle. In other words, the acquisition unit 111 may acquire, for example, behavior information for a plurality of people that corresponds the behavior of the person to be estimated with the situation around the person in a situation in which a simulation is to be performed in the behavior estimation model.

A person's attributes indicate, for example, the nature and characteristics of a person. As an example, a person's attributes are various natures and characteristics including the person's age (age group), the frequency of driving a vehicle, the person's personality (personality test), and the person's athletic ability. A person's attributes may be estimated, for example, based on the results of a questionnaire or test conducted on the person. In other words, the acquisition unit 111 may acquire, as attribute information, information on at least one of the person's age, the frequency of driving a vehicle, the person's personality, and the person's athletic ability, which are the results of a questionnaire or test conducted in advance on the person.

The classification unit 112 classifies the behavior of the person according to the tendency of the person's behavior based on the behavior information acquired by the acquisition unit 111. The classification unit 112 classifies the behaviors of the multiple people into one of multiple behavioral tendencies based on multiple pieces of behavioral information. The behavioral tendency is, for example, the tendency of the behavior that the person takes in response to the traffic conditions around the person. As a specific example, when a person crosses a road at an intersection, etc., the behavioral tendency may be the tendency of the person to cross the road or to wait without crossing the road depending on the driving conditions of the vehicles driving around the person (vehicle positions, speeds, accelerations, etc.).

Here, the classification unit 112 may, for example, express the behavioral tendency as a numerical value and perform feature quantification. As an example, the classification unit 112 may use the reward function value obtained by using inverse reinforcement learning for each of the multiple people as a feature. In this case, the classification unit 112 may, for example, calculate the reward function (e.g., the intention of the person) based on the behavioral information of the multiple people. It is considered that if the behavior of a person differs, the tendency (behavioral tendency) of the person when acting differs, and the reward function (the intention of the person) differs. As a specific example, when a person crosses a road at an intersection, the classification unit 112 estimates the behavior of the person, that is, the person crossing the road (the tendency to approach the intersection and go ahead of the vehicle even if a vehicle is coming) or stopping in front of the road (the tendency to give way to the vehicle without approaching the intersection at the same time as the vehicle), as a reward function, according to the situation of the person (e.g., the distance of the person from the intersection) and the driving situation of the vehicle (the driving speed of the vehicle, the acceleration of the vehicle, and the distance of the vehicle from the intersection).

The classification unit 112 performs clustering on the features of the reward function estimated as described above, for example. In this case, the classification unit 112 performs classification based on the similarity between the features of the person by using various clustering methods, including, for example, K-means and hierarchical clustering. The classification unit 112 may classify the similarity between the features into three or more classes. That is, the classification unit 112 may classify based on the similarity of the person's behavioral tendencies according to at least one of the positional relationship between the vehicle and the person based on the behavioral information, the vehicle speed, and the vehicle acceleration.

The first learning unit 113 learns the association between the person's attributes and behavioral tendencies based on the attribute information acquired by the acquisition unit 111 and the behavioral tendencies classified by the classification unit 112, and generates a first learned model. In this case, the first learning unit 113 may generate, as the first learned model, an item that specifies an attribute related to the person's behavioral tendencies when estimating the person's behavior, and a learned model for outputting the estimation result of the person's behavior according to the item. That is, the first learning unit 113 analyzes the association between the result of clustering the person's behavioral tendencies performed as in the above example and the person's attributes, and selects an attribute item (attribute item) that has a relatively strong association. In this case, the first learning unit 113 may select an attribute item that has a relatively strong association with the clustering result by using various testing methods including, for example, correlation analysis, t-test, and analysis of variance.

Figure 4 is a diagram explaining an example of selected attribute items.

As a specific example, the first learning unit 113 may perform a correlation analysis between various attributes as the results of the questionnaire, such as "impatient driving tendency," "preparatory driving at traffic lights," "falsehood detection scale," and "understanding of road environment (stress sensitivity)," and the clustering of a person's behavioral tendencies, and select an attribute item using a P value as a result of the analysis. The P value may be, for example, the probability that the current questionnaire result (attribute information) will be observed in a test. Here, the first learning unit 113 may select the questionnaire illustrated in FIG. 4 as an attribute item when, for example, the P value is <0.1 when determining whether or not there is a significant tendency.

The first learning unit 113 also learns a model that estimates a person's behavioral tendency based on the attribute information acquired by the acquisition unit 111. In this case, the first learning unit 113 may use, for example, various models used in classification problems, including decision trees, SVMs, and neural networks, as the model. The first learning unit 113 may, for example, perform learning to solve a classification problem using data that has attribute items selected as described above as input and the clustering results as output, thereby generating a first trained model.

The second learning unit 114 learns the association between the tendency of behavior classified by the classification unit 112 and parameters used in a behavior estimation model that estimates the behavior of a person according to the tendency, and generates a second learned model. The second learning unit 114 learns parameters of a behavior estimation model (traffic participant model) that estimates the behavior of a person, for example, by using behavior information of a person in each class of behavior tendency as described above. That is, the second learning unit 114 may generate the second learned model by learning parameters of traffic participants that are used in the behavior estimation model according to each of the multiple behavior tendencies classified by the classification unit 112. Here, the traffic participant model may be, for example, a logistic regression model that determines whether, when a person crosses a road at an intersection, the person crosses before the vehicles traveling around the intersection, or whether the person crosses after the vehicles (waits at the intersection). Let p be the probability that a person will cross the street before a vehicle, x _* be a variable indicating the situation such as the pedestrian-vehicle distance, vehicle speed, and vehicle acceleration, and a _* and b be model parameters. Then, the probability p is calculated using the following equation (1).

…（１）

…(1)

The second learning unit 114 may, for example, learn to use a variable value indicating a situation as input and output whether a person crossed the road before the vehicle in that situation (went ahead) (for example, 1 if the person was ahead and 0 if the person waited). That is, the second learning unit 114 may, for example, learn behavioral information of multiple people (the behavior of those people and the traffic conditions around the people), the results of classifying behavioral tendencies based on the behavioral information of multiple people, and parameters for each class to generate the second trained model.

Next, we will explain the functional blocks involved in estimating people's behavior (simulating traffic participants).

The first identification unit 115 identifies one or more attributes from the attribute items of the multiple people based on the first learned model generated by the first learning unit 113, and identifies the behavior of the person according to the identified attribute. When estimating the behavioral tendency of the person performing the simulation, the first identification unit 115 first identifies one or more of the multiple attribute items. Here, the multiple attribute items may be attribute items that are relatively strongly related to the result of the clustering performed by the classification unit 112 described above. The first identification unit 115 estimates the behavioral tendency of the person performing the simulation (the traffic participant to be simulated) based on the identified attribute items (input items) and the first learned model. That is, the first identification unit 115 identifies (selects) the attribute items of the traffic participant to be simulated from the multiple attribute items, and estimates the behavioral tendency of the person according to the identified attribute items using the first learned model. More specifically, the first identification unit 115 inputs the identified attribute items into the first learned model and outputs the behavioral tendency of the person.

The second identification unit 116 identifies parameters to be used in the behavior estimation model based on the person's behavior identified by the first identification unit 115 and the second learned model generated by the second learning unit 114. That is, the second identification unit 116 acquires parameters for estimating the person's behavior (simulating a traffic participant) based on the person's estimated behavioral tendency and the second learned model. In other words, the second identification unit 116 inputs the person's behavioral tendency acquired by the first identification unit 115 into the second learned model and outputs parameters for estimating the person's behavior.

The behavior estimation unit 117 estimates the behavior of the person based on a behavior estimation model to which the parameters identified by the second identification unit 116 are input. The behavior estimation unit 117 sets the parameters identified by the second identification unit 116 in a simulation model (a behavior estimation model that simulates traffic participants) and estimates the behavior of the person (performs a simulation of the behavior of traffic participants). The behavior estimation unit 117 can use, for example, various models (such as publicly known models) as the behavior estimation model (simulation model).

The behavior estimation unit 117 may transmit information about the result of estimating the behavior of the person (simulation result) to an external device (for example, an external server (not shown)) via the communication unit 121, for example.
The behavior estimation unit 117 may store, for example, information on the results of estimation of the person's behavior (simulation results) in the storage unit 122 .
The behavior estimation unit 117 may display, for example, the results of estimation of the person's behavior (simulation results) on the display unit 123 .

[Information processing method]
Next, an information processing method according to an embodiment will be described.
FIG. 5 is a diagram for explaining an information processing method according to an embodiment.

In step ST1, the information processing device 100 performs advance preparations for estimating a person's behavior (preparation before executing a simulation).
That is, the information processing device 100 acquires behavioral information related to a person's behavior and attribute information related to the person's attributes and performs learning. Then, the information processing device 100 identifies a plurality of attribute items that are options when estimating a person's behavior, and generates an estimation model (1) (first trained model) for estimating the behavioral tendency of the person according to each attribute item. Furthermore, the information processing device 100 generates an estimation model (2) (second trained model) for identifying parameters when estimating (simulating) the behavior of the person according to the behavioral tendency of the person.

In step ST2, the information processing device 100 estimates the behavior of a person (executes a simulation).
That is, the information processing device 100 determines the attributes of the person for which behavior estimation is to be performed from among the multiple attribute items identified in step ST1. The information processing device 100 inputs the determined attributes into an estimation model (1) (first trained model) generated in step ST1, and determines the behavioral tendency of the person whose behavior is to be estimated. The information processing device 100 inputs the determined behavioral tendency into an estimation model (2) (second trained model) generated in step ST1, and determines parameters of the behavior estimation model (simulation model). The information processing device 100 sets the determined parameters in the behavior estimation model, and estimates the person's behavior (executes a simulation).

Next, the information processing method according to the embodiment described above will be described in detail.

First, the above-mentioned step ST1 will be described in detail.
FIG. 6 is a flowchart illustrating an information processing method according to an embodiment, which is a method of advance preparation when estimating a person's behavior.

In step ST101, the acquisition unit 111 acquires behavioral information regarding the behavior of a person and attribute information regarding the attributes of the person. For example, the acquisition unit 111 may acquire, as the behavioral information, information regarding the behavior of a plurality of people regarding the behavior of the person in a situation in which there is a person and a vehicle traveling around the person. In addition, the acquisition unit 111 may acquire, as the attribute information, information regarding at least one of the person's age, the frequency with which the person drives a vehicle, the person's personality, and the results of a questionnaire or test conducted in advance on the person regarding the person's athletic ability.

In step ST102, the classification unit 112 performs classification according to the behavioral tendencies of the person based on the behavioral information acquired in step ST101. Here, the classification unit 112 may classify according to the similarity of the behavioral tendencies of the person based on, for example, at least one of the positional relationship between the vehicle and the person based on the behavioral information, the speed of the vehicle, and the acceleration of the vehicle.

In step ST103, the first learning unit 113 learns the association between a person's attributes and behavioral tendencies based on the attribute information acquired in step ST101 and the behavioral tendencies classified in step ST102, and generates a first learned model (estimation model (1)). In this case, the first learning unit 113 may generate, for example, as the first learned model, an item that identifies an attribute related to a person's behavioral tendencies when estimating the person's behavior, and a learned model for outputting an estimation result of the person's behavior according to the item.

In step S104, the second learning unit 114 learns the association between the behavioral tendencies classified in step ST102 and the parameters used in the behavior estimation model that estimates the behavior of a person according to the tendencies, and generates a second learned model (estimation model (2)). In this case, the second learning unit 114 may generate the second learned model, for example, by learning parameters of traffic participants used in the behavior estimation model according to each of the multiple behavioral tendencies classified by the classification unit 112.

Next, the above-mentioned step ST2 will be described in detail.
FIG. 7 is a flowchart illustrating an information processing method according to an embodiment, which is a method for estimating a person's behavior.

In step ST201, the first identification unit 115 identifies attribute items to be input into the first learned model generated in step ST103. Furthermore, the first identification unit 115 identifies a behavioral tendency of a person according to the attribute items based on the identified attribute items and the first learned model generated in step ST103. That is, the first identification unit 115 identifies a behavioral tendency of a person who is the subject of behavior estimation.

In step ST202, the second identification unit 116 identifies parameters to be used in the behavior estimation model based on the behavior (behavioral tendency) of the person identified in step ST201 and the second trained model generated in step ST104.

In step ST203, the behavior estimation unit 117 estimates the person's behavior using the behavior estimation model to which the parameters identified in step ST202 have been input. The behavior estimation unit 117 can use, for example, various models as the behavior estimation model (simulation model). In this case, in the above-mentioned step ST104, a second trained model capable of identifying parameters according to the behavior estimation model (simulation model) used in step ST203 is generated.

[Modification]
Next, a modification of the above-described embodiment will be described.

The above-mentioned classification unit 112 has been described as performing clustering according to the behavioral tendencies of people, for example. However, the present disclosure is not limited to this example, and the classification unit 112 may perform feature quantification using some numerical value for the behavioral tendencies. In this case, the classification unit 112 may perform feature quantification using, for example, parameters of a traffic participant model such as equation (1) used in the simulation.

The first learning unit 113 may, for example, select attribute information during learning to generate the first learned model so as to improve the accuracy of the learned model.

The second learning unit 114 has been described as estimating a traffic participant model using, for example, a logistic regression model. However, the present disclosure is not limited to this example, and the second learning unit 114 may use various models (regardless of the shape of the traffic participant model). For example, the second learning unit 114 may be able to use data in which behavioral tendencies are classified (all data within a behavioral tendency class) to calculate average traffic participant model parameters, and then adjust some of the parameters of the traffic participant model to match the behavioral information. The second learning unit 114 may use attribute information as information (input information) to be used for the adjustment.

Each of the above-mentioned units of the information processing device 100 may be realized as a function of a computer's arithmetic processing device, etc. That is, the acquisition unit 111, the classification unit 112, the first learning unit 113, the second learning unit 114, the first identification unit 115, the second identification unit 116, and the behavior estimation unit 117 (control unit 110) of the information processing device 100 may be realized as an acquisition function, a classification function, a first learning function, a second learning function, a first identification function, a second identification function, and a behavior estimation function (control function), respectively, by the computer's arithmetic processing device, etc.
The information processing program can cause a computer to realize each of the above-mentioned functions. The information processing program may be recorded in a non-transitory computer-readable recording medium, such as a memory, a solid-state drive, a hard disk drive, or an optical disk.
As described above, each unit of the information processing device 100 may be realized by a computer arithmetic processing device or the like. The arithmetic processing device or the like is configured by, for example, an integrated circuit or the like. Therefore, each unit of the information processing device 100 may be realized as a circuit constituting the arithmetic processing device or the like. That is, the acquisition unit 111, the classification unit 112, the first learning unit 113, the second learning unit 114, the first identification unit 115, the second identification unit 116, and the behavior estimation unit 117 (control unit 110) of the information processing device 100 may be realized as an acquisition circuit, a classification circuit, a first learning circuit, a second learning circuit, a first identification circuit, a second identification circuit, and a behavior estimation circuit (control circuit) constituting the arithmetic processing device or the like of a computer.
The communication unit 121, the storage unit 122, and the display unit 123 of the information processing device 100 may be realized as a communication function, a storage function, and a display function including the functions of an arithmetic processing device, for example. The communication unit 121, the storage unit 122, and the display unit 123 of the information processing device 100 may be realized as a communication circuit, a storage circuit, and a display circuit by being configured by, for example, an integrated circuit, etc. The communication unit 121, the storage unit 122, and the display unit 123 of the information processing device 100 may be realized as a communication device, a storage device, and a display device by being configured by, for example, a plurality of devices.

The information processing device 100 can combine one or any two or more of the above-mentioned units.
In this disclosure, the term "information" is used, but the term "information" can be replaced with "data" and the term "data" can be replaced with "information."

[Aspects and Effects of the Present Embodiment]
Next, one aspect of this embodiment and the effects of each aspect will be described. Note that this embodiment is not limited to each aspect described below, and may be realized by appropriately combining each of the above-mentioned parts. Also, the effects described below are examples, and the effects of each aspect are not limited to those described below.

(Aspect 1)
An information processing device of one embodiment includes an acquisition unit that acquires behavioral information regarding a person's behavior and attribute information regarding the person's attributes, a classification unit that performs classification according to the person's behavioral tendencies based on the behavioral information acquired by the acquisition unit, a first learning unit that learns the association between the person's attributes and behavioral tendencies based on the attribute information acquired by the acquisition unit and the behavioral tendencies classified by the classification unit, and generates a first learned model, and a second learning unit that learns the association between the behavioral tendencies classified by the classification unit and parameters used in a behavior estimation model that estimates the person's behavior according to the tendencies, and generates a second learned model.
As a result, the information processing device generates the first learned model and the second learned model based on behavioral information and attribute information of the same person (corresponding behavioral information and attribute information of multiple people), thereby preventing multiple parameters from being set independently and enabling correlated behavioral estimation (simulation) of various behaviors of a person.
As a specific example, when estimating a person's behavior, the information processing device can prevent the occurrence of settings that would be unnatural in the real world, such as a person not obeying traffic lights but obeying stop signs.
The information processing device can make it easy for a user of the information processing device to understand what attributes of a person the parameters as a whole represent.

(Aspect 2)
In the information processing device of one aspect, the acquisition unit may acquire, as the behavior information, information on the behavior of a plurality of people regarding the behavior of the person in a situation in which there is a person and a vehicle traveling around the person.
This enables the information processing device to generate a trained model suitable for estimating a person's behavior according to the vehicle's driving conditions.

(Aspect 3)
In one aspect of the information processing device, the classification unit may classify according to the similarity of the person's behavioral tendencies depending on at least one of the positional relationship between the vehicle and the person based on the behavioral information, the vehicle's speed, and the vehicle's acceleration.
This enables the information processing device to obtain behavior information regarding how a person behaves depending on the driving conditions of the vehicle.

(Aspect 4)
In one embodiment of the information processing device, the first learning unit may generate, as the first learned model, an item that identifies an attribute related to a person's behavioral tendency when estimating the person's behavior, and a learned model for outputting an estimation result of the person's behavior based on the item.
This enables the information processing device to estimate a behavioral tendency according to a person's attributes (attribute items).

(Aspect 5)
In one embodiment of the information processing device, the acquisition unit may acquire, as attribute information, information regarding at least one of the following: the person's age, the person's years of driving experience and driving frequency of a vehicle, the person's personality, and the results of a questionnaire or test conducted in advance on the person regarding the person's athletic ability.
This allows the information processing device to generate a first trained model according to various attributes of a person. That is, when the information processing device acquires attributes (such as the above-mentioned age and driving frequency) that are easy for a user of the device to understand as parameters (attribute information), it can generate a trained model that outputs parameters related to the attributes. In addition, the information processing device can make it easy for a user to understand what attributes the above-mentioned attribute information represents as a whole.

(Aspect 6)
In one embodiment of the information processing device, the second learning unit may generate the second learned model by learning parameters of traffic participants to be used in the behavior estimation model according to each of the multiple behavioral tendencies classified by the classification unit.
This allows the information processing device to estimate the behavior of a person according to the person's behavioral tendencies.

(Aspect 7)
An information processing device of one embodiment includes a first identification unit that identifies one or more attributes from items of attributes of a plurality of persons based on a first learned model generated by a first learning unit, and identifies behavior of the person according to the identified attributes, a second identification unit that identifies parameters to be used in a behavior estimation model based on the behavior of the person identified by the first identification unit and a second learned model generated by a second learning unit, and a behavior estimation unit that estimates behavior of the person based on the behavior estimation model into which the parameters identified by the second identification unit are input.
This enables the information processing device to perform behavior estimation that is in line with the real world and that has correlations between various behaviors of a person.

(Aspect 8)
An information processing device of one embodiment includes a first identification unit that identifies one or more attributes from items of attributes of a plurality of persons based on a first learned model generated by learning a classification result classified according to a tendency of a person's behavior based on behavior information regarding the person's behavior and attribute information regarding the person's attributes, and identifies the person's behavior according to the identified attribute; a second learned model generated by learning an association between the behavioral tendencies classified according to the tendency of the person's behavior and parameters used in a behavior estimation model that estimates the person's behavior according to the tendency; a second identification unit that identifies parameters to be used in the behavior estimation model based on the person's behavior identified by the first identification unit; and a behavior estimation unit that estimates the person's behavior based on the behavior estimation model into which the parameters identified by the second identification unit have been input.
As a result, when the information processing device specifies an attribute (attribute item) of a person whose behavior is to be estimated, parameters related to that attribute are set all at once, thereby preventing parameters that would be unnatural in the real world from being set.
The information processing device can eliminate the need for a user of the device to set parameters of a person whose behavior is to be estimated, and can estimate the behavior of a person who has the behavioral tendency of that attribute simply by specifying the attribute of the person.

(Aspect 9)
In one aspect of the information processing method, a computer executes the following steps: an acquisition step in which behavioral information regarding a person's behavior and attribute information regarding the person's attributes are acquired; a classification step in which the computer classifies the person's behavior according to its tendencies based on the behavioral information acquired by the acquisition step; a first learning step in which the computer learns the association between the person's attributes and behavioral tendencies based on the attribute information acquired by the acquisition step and the behavioral tendencies classified by the classification step, thereby generating a first trained model; and a second learning step in which the computer learns the association between the behavioral tendencies classified by the classification step and parameters used in a behavior estimation model that estimates the person's behavior according to those tendencies, thereby generating a second trained model.
As a result, the information processing method can achieve the same effects as the information processing device according to the above-described aspect.

(Aspect 10)
An information processing program of one embodiment causes a computer to realize an acquisition function that acquires behavioral information regarding a person's behavior and attribute information regarding the person's attributes, a classification function that performs classification according to the person's behavioral tendencies based on the behavioral information acquired by the acquisition function, a first learning function that learns the association between the person's attributes and behavioral tendencies based on the attribute information acquired by the acquisition function and the behavioral tendencies classified by the classification function, and generates a first learned model, and a second learning function that learns the association between the behavioral tendencies classified by the classification function and parameters used in a behavior estimation model that estimates a person's behavior according to those tendencies, and generates a second learned model.
As a result, the information processing program can achieve the same effects as the information processing device according to the above-described aspect.

(Aspect 11)
In one aspect of the information processing method, a computer executes the following steps: a first identification step in which a computer identifies one or more attributes from among a plurality of attribute items of a person based on a first trained model generated by learning a classification result that is classified according to a person's behavioral tendency and attribute information related to the person's attributes based on behavioral information related to the person's behavior, and identifies the person's behavior according to the identified attribute; a second identification step in which a computer learns an association between the behavioral tendency classified according to the person's behavioral tendency and parameters used in a behavior estimation model that estimates the person's behavior according to the tendency, and identifies parameters to be used in the behavior estimation model based on a second trained model and the person's behavior identified by the first identification step; and a behavior estimation step in which a computer estimates the person's behavior based on the behavior estimation model into which the parameters identified by the second identification step have been input.
As a result, the information processing method can achieve the same effects as the information processing device according to the above-described aspect.

(Aspect 12)
An information processing program of one embodiment causes a computer to realize: a first identification function that identifies one or more attributes from among multiple attribute items of a person based on a first learned model generated by learning a classification result that is classified according to a person's behavioral tendency and attribute information related to the person's attributes based on behavioral information related to the person's behavior, and identifies the person's behavior according to the identified attribute; a second identification function that learns an association between the behavioral tendency classified according to the person's behavioral tendency and parameters used in a behavior estimation model that estimates the person's behavior according to the tendency, and identifies parameters to be used in the behavior estimation model based on a second learned model and the person's behavior identified by the first identification function; and a behavior estimation function that estimates the person's behavior based on the behavior estimation model into which the parameters identified by the second identification function have been input.
As a result, the information processing program can achieve the same effects as the information processing device according to the above-described aspect.

REFERENCE SIGNS LIST 100 Information processing device 110 Control unit 111 Acquisition unit 112 Classification unit 113 First learning unit 114 Second learning unit 115 First identification unit 116 Second identification unit 117 Behavior estimation unit 121 Communication unit 122 Storage unit 123 Display unit

Claims (10)

an acquisition unit that acquires behavior information related to a person's behavior and attribute information related to the person's attributes;
A classification unit that classifies a person's behavior according to a tendency of the person's behavior based on the behavior information acquired by the acquisition unit;
a first learning unit that learns associations between attributes and behavioral tendencies of a person based on the attribute information acquired by the acquisition unit and the behavioral tendencies classified by the classification unit, and generates a first trained model;
a second learning unit that learns an association between a tendency of the behavior classified by the classification unit and a parameter used in a behavior estimation model that estimates a person's behavior according to the tendency, and generates a second learned model;
An information processing device comprising: The information processing apparatus according to claim 1 , wherein the acquisition unit acquires, as the behavior information, information on the behavior of a plurality of people regarding the behavior of the person in a situation in which the person and a vehicle traveling around the person are present. The information processing device according to claim 2 , wherein the classification unit classifies the person according to a similarity of tendencies of the person's behavior according to at least one of a positional relationship between the vehicle and the person based on the behavior information, a vehicle speed, and a vehicle acceleration. The information processing device according to any one of claims 1 to 3, wherein the first learning unit generates, as a first learned model, an item that identifies an attribute related to a tendency of a person's behavior when estimating the person's behavior, and a learned model for outputting an estimation result of the person's behavior according to the item. The information processing device according to any one of claims 1 to 4, wherein the acquisition unit acquires, as attribute information, at least one of information regarding the person's age, the person's years of driving experience and driving frequency of a vehicle, the person's personality, and the results of a questionnaire or test conducted in advance on the person regarding the person's athletic ability. The information processing device according to any one of claims 1 to 5, wherein the second learning unit generates the second learned model by learning parameters of traffic participants to be used in a behavior estimation model corresponding to each of a plurality of behavioral tendencies classified by the classification unit. A first identification unit that identifies one or more attributes from items of attributes of a plurality of persons based on a first trained model generated by the first learning unit, and identifies a behavior of the person according to the identified attributes;
a second identification unit that identifies parameters to be used in a behavior estimation model based on the behavior of the person identified by the first identification unit and a second trained model generated by the second learning unit;
a behavior estimation unit that estimates a person's behavior based on a behavior estimation model to which the parameters identified by the second identification unit are input;
The information processing device according to any one of claims 1 to 6, comprising: a first identification unit that identifies one or more attributes from among a plurality of attribute items of a person based on a first trained model generated by learning a classification result in which the person's behavior is classified according to the tendency of the person's behavior based on behavior information related to the person's behavior and attribute information related to the person's attributes, and identifies the person's behavior according to the identified attribute;
a second identification unit that identifies parameters to be used in the behavior estimation model based on a second trained model generated by learning an association between a behavioral tendency classified according to a person's behavioral tendency and a parameter to be used in the behavior estimation model that estimates the person's behavior according to the tendency and the person's behavior identified by the first identification unit;
a behavior estimation unit that estimates a person's behavior based on a behavior estimation model to which the parameters identified by the second identification unit are input;
An information processing device comprising: The computer
An acquisition step of acquiring behavior information related to a person's behavior and attribute information related to the person's attributes;
a classification step of classifying the person's behavior according to the tendency of the person's behavior based on the behavior information acquired by the acquisition step;
a first learning step of learning associations between attributes and behavioral tendencies of a person based on the attribute information acquired by the acquiring step and the behavioral tendencies classified by the classifying step, to generate a first trained model;
a second learning step of learning an association between the tendency of the behavior classified by the classification step and a parameter used in a behavior estimation model that estimates a person's behavior according to the tendency, thereby generating a second learned model;
An information processing method for performing the above. On the computer,
An acquisition function for acquiring behavior information related to a person's behavior and attribute information related to the person's attributes;
A classification function that classifies a person's behavior according to a tendency of the person's behavior based on the behavior information acquired by the acquisition function;
A first learning function that learns associations between attributes and behavioral tendencies of a person based on attribute information acquired by the acquisition function and behavioral tendencies classified by the classification function, and generates a first learned model;
A second learning function that learns an association between a tendency of behavior classified by the classification function and a parameter used in a behavior estimation model that estimates a person's behavior according to the tendency, and generates a second learned model;
An information processing program that realizes this.

Images