CN113792655A

CN113792655A - Intention identification method and device, electronic equipment and computer readable medium

Info

Publication number: CN113792655A
Application number: CN202111076054.2A
Authority: CN
Inventors: 徐鑫; 张亮亮
Original assignee: Jingdong Kunpeng Jiangsu Technology Co Ltd
Current assignee: Jingdong Kunpeng Jiangsu Technology Co Ltd
Priority date: 2021-09-14
Filing date: 2021-09-14
Publication date: 2021-12-14

Abstract

The application discloses an intention identification method, an intention identification device, electronic equipment and a computer readable medium, which relate to the technical field of computers, and the intention identification method comprises the following steps: receiving an intention identification request, and acquiring barrier behavior information carried in the intention identification request; calling the dynamic Bayesian network, and determining the intention recognition posterior probability corresponding to the intention node in the dynamic Bayesian network at the previous moment and the state transition probability of each node in the dynamic Bayesian network; determining an intention identification prior probability corresponding to an intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node; updating the prior probability of the intention recognition according to the behavior information of the obstacle to obtain the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment; and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention. The intention of obstacles at the intersection is accurately identified.

Description

Intention identification method and device, electronic equipment and computer readable medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to an intention recognition method, an intention recognition apparatus, an electronic device, and a computer-readable medium.

Background

At present, traffic accidents between vehicles and Vulnerable Road Users (VRUs) are frequent. Most moving target object protection systems focus on detection and tracking of moving target objects, and there is less research on reducing the risk of collision by further motion prediction of moving target objects.

In the process of implementing the present application, the inventor finds that at least the following problems exist in the prior art:

the intent prediction for moving target objects is not accurate.

Disclosure of Invention

In view of the above, embodiments of the present application provide an intention identification method, an intention identification apparatus, an electronic device, and a computer-readable medium, which can solve the existing problem of inaccurate intention prediction for a moving target object.

To achieve the above object, according to an aspect of embodiments of the present application, there is provided an intention identifying method including:

receiving an intention identification request, and acquiring obstacle behavior information in the intention identification request;

calling the dynamic Bayesian network, determining an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment and determining a state transition probability of each node in the dynamic Bayesian network;

determining an intention identification prior probability corresponding to an intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node;

updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment;

and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention.

Optionally, before invoking the dynamic bayesian network, the intent recognition method further comprises:

acquiring an observation variable set corresponding to the barrier behavior information at the historical moment;

determining a condition variable set corresponding to the obstacle behavior information based on the corresponding relation between the preset obstacle behavior information and the condition variable;

acquiring an intention corresponding to the obstacle behavior information at the historical moment;

and taking each condition variable in the condition variable set as a hidden node in the network, taking each observation variable in the observation variable set as an observation variable node in the network, taking an intention corresponding to the barrier behavior information at the historical moment as an intention node in the network, updating the corresponding hidden node according to the observation variable corresponding to the barrier behavior information acquired at each moment, and updating the corresponding intention node based on the updated hidden node to construct and obtain the dynamic Bayesian network.

Optionally, before receiving the intention identification request, the intention identification method further comprises:

in response to determining that the current position is detected to be located in a preset intersection area, acquiring obstacle behavior information located in the preset intersection area;

based on the obstacle behavior information, an intent recognition request is generated.

Optionally, obtaining the obstacle behavior information carried in the intention identification request includes:

the minimum distance between the obstacle and the vehicle, the relative longitudinal distance between the obstacle and the intersection, the obstacle orientation angle, the obstacle limb action occurrence probability and whether the obstacle has ever acted on limbs at the current moment or the previous moment are obtained.

Optionally, the intention identification method further comprises:

and determining the state transition probability corresponding to the discontinuous object association factor node in the hidden node according to whether the barrier has limb movement at the current time and the previous time.

Optionally, determining an intention identification prior probability corresponding to an intention node in the dynamic bayesian network at the current time according to the intention identification posterior probability and the state transition probability of each node, including:

determining the transition probability of all nodes according to the product of the state transition probability of each node;

and determining the prior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment according to the transition probabilities and the posterior probability of the intention recognition of all the nodes.

Optionally, the updating the prior probability of the intention recognition according to the obstacle behavior information includes:

calculating to obtain state posterior probability according to the behavior information and the conditional probability of the barrier;

and summing the state posterior probabilities, and further updating the prior probability of the intention recognition so as to obtain the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment.

In addition, the present application also provides an intention identifying apparatus including:

the receiving unit is configured to receive the intention identification request and acquire the obstacle behavior information carried in the intention identification request;

a state transition probability determination unit configured to invoke the dynamic bayesian network to determine an intention identification posterior probability corresponding to an intention node in the dynamic bayesian network at a previous time and to determine a state transition probability of each node in the dynamic bayesian network;

the prior probability determining unit is configured to determine the prior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment according to the posterior probability of the intention recognition and the state transition probability of each node;

the posterior probability determining unit is configured to update the intention identification prior probability according to the obstacle behavior information so as to obtain the intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the current moment;

and the intention identification unit is configured to determine the intention corresponding to the intention identification posterior probability according to the preset corresponding relation between the intention identification probability and the intention.

Optionally, the intention recognition apparatus further comprises a dynamic bayesian network building unit configured to:

Optionally, the intention-identifying apparatus further comprises a request-generating unit configured to:

Optionally, the receiving unit is further configured to:

Optionally, the state transition probability determination unit is further configured to:

Optionally, the prior probability determination unit is further configured to:

Optionally, the posterior probability determination unit is further configured to:

In addition, the present application also provides an intention recognition electronic device, including: one or more processors; a storage device for storing one or more programs which, when executed by one or more processors, cause the one or more processors to implement the intent recognition method as described above.

In addition, the present application also provides a computer-readable medium, on which a computer program is stored, which when executed by a processor implements the intent recognition method as described above.

One embodiment of the above invention has the following advantages or benefits: the method comprises the steps of receiving an intention identification request, and obtaining obstacle behavior information carried in the intention identification request; calling the dynamic Bayesian network, determining an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment and a state transition probability of each node in the dynamic Bayesian network, and determining an intention identification prior probability corresponding to the intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node; updating the prior probability of the intention recognition according to the behavior information of the obstacle to obtain the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment; and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention. Thus, the intention of the obstacle at the intersection can be accurately identified.

Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.

Drawings

The drawings are included to provide a further understanding of the application and are not to be construed as limiting the application. Wherein:

fig. 1 is a schematic diagram of a main flow of an intention identifying method according to a first embodiment of the present application;

fig. 2 is a schematic diagram of a main flow of an intention identifying method according to a second embodiment of the present application;

fig. 3 is a schematic view of an application scenario of an intention recognition method according to a third embodiment of the present application;

FIG. 4 is a schematic diagram of a dynamic Bayesian network of an embodiment of the present application;

FIG. 5 is a flow diagram of a process for deliberate inference at a time in accordance with an embodiment of the present application;

FIG. 6 is a schematic diagram of the main elements of an intent recognition device, according to an embodiment of the present application;

FIG. 7 is an exemplary system architecture diagram to which embodiments of the present application may be applied;

fig. 8 is a schematic structural diagram of a computer system suitable for implementing the terminal device or the server according to the embodiment of the present application.

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

Fig. 1 is a schematic diagram of a main flow of an intention identifying method according to a first embodiment of the present application, and as shown in fig. 1, the intention identifying method includes:

and step S101, receiving the intention identification request, and acquiring the obstacle behavior information carried in the intention identification request.

In this embodiment, the execution subject (for example, a server) of the intention identification method may receive the intention identification request by means of wired connection or wireless connection. Specifically, the intention identifying request may be a request for identifying the intention of an obstacle located at the intersection. Specifically, in the embodiment of the present application, the obstacle may be a moving target object, and may be a rider or a pedestrian, for example. The obstacle behavior information may be, for example, information such as a gesture of a rider, a lifting motion of the rider, a distance between the rider and an intersection, a minimum distance between the rider and a bicycle, and a trunk orientation angle of the rider. The intersection can be a crossroad, a three-way intersection, an intersection at the door of a school and the like.

In the embodiment of the present application, taking an obstacle as a moving target object as an example: specifically, before receiving the intention identification request, the intention identification method further includes:

the execution main body responds to the fact that the current position is detected to be located in a preset intersection area, and barrier behavior information located in the preset intersection area is obtained; based on the obstacle behavior information, an intent recognition request is generated. For example, when the smart car detects that an obstacle such as a moving target object is present near the intersection area, behavior information of the obstacle such as the moving target object may be acquired, and a request for intention recognition of the obstacle such as the moving target object may be generated based on the acquired behavior information. The behavior information may include information on whether an obstacle such as a moving target object lifts an arm, whether a head is swung, and other actions.

Specifically, obtaining obstacle behavior information in the intention identification request includes:

the minimum distance between the obstacle and the vehicle, the relative longitudinal distance between the obstacle and the intersection, the obstacle orientation angle, the obstacle limb action occurrence probability and whether the obstacle has ever suffered limb action at the current moment or the previous moment in the intention identification request are obtained.

For example, the minimum distance between the moving target object and the host vehicle, the relative longitudinal distance between the moving target object and the intersection, the orientation angle of the moving target object, the limb action occurrence probability of the moving target object, and whether the moving target object has ever occurred in the current time or the previous time may be obtained.

Wherein, the minimum distance between the moving target object and the self-vehicle is as follows: the minimum distance between the moving target object and the vehicle at risk of collision when the vehicle travels forward at the current speed is taken as the dynamic environment factor H^DYNIs a continuous variable. This feature assumes that both the moving target object and the own vehicle advance at a constant speed. Although the speed of the moving target object is not constant during actual steering, the characteristic can still provide valuable information for judging the danger level of the situation, and the accuracy of the intention judgment of the moving target object is improved. In particular, dynamic environmental factor H^DYN: the variable defining whether the potential collision risk exists between the moving target object and the vehicle is a hidden variable and a discrete variable, and the value set is {0,1}, which respectively represents no collision risk and collision risk.

Relative longitudinal distance of moving target object from intersection: the intersection area is relatively fixed, and the moving target object can only perform steering action when entering the intersection area, but can keep the original motion state before entering the intersection area, and is used as a static environment factor H^STATThe observed quantity of (2). Static environmental factor H^STAT: the variable defining whether the moving target object is in the intersection region is a hidden variable, a discrete variable, and a continuous variable. The value set is {0,1}, which respectively represents that the moving target object is not in the intersection region and the moving target object is in the intersection region.

Moving target object orientation angle: defining a moving target object torso orientation angle of the normalized multi-task detector output, representing a moving target object in a fixed coordinate system of the target object established with a first frame orientation angle of the target object at which the movement is detected being 0 degreesThe orientation angle of the moving target object. Specifically, by performing rotational translation on the coordinates of the vehicle, when the orientation angle of the first frame of the moving target object relative to the vehicle is detected to be 0, a fixed coordinate system of the moving target object is established, and the executive body can take the angle value in the coordinate system (i.e. the orientation angle of the moving target object) as the continuous object factor H^ACTCThe observed quantity of (2). The torso orientation angle of the moving target object may characterize the destination of the moving target object. Continuous object factor H^ACTC: the variable defining the destination of the moving target object is a hidden variable and is a discrete variable. The set of values is {0,1,2}, indicating that the destination of the moving target object is directly in front, to the left, and to the right, respectively.

Probability of occurrence of limb movement of moving target object: defining a moving target object up-arm probability output by the multi-tasking detector, the closer the value is to 1, the greater the target object up-arm probability characterizing the movement, as a non-continuous object factor H^ACTDIs a continuous variable. The arm raising motion of the moving target object may characterize whether the moving target object indicates its intent with a gesture. Discontinuous object factor H^ACTD: the variable defining whether the moving target object raises the arm is a hidden variable and a discrete variable. The values are set to {0,1}, which respectively represent the arm not lifted and the arm lifted.

Whether the moving target object has limb motion at the current time or the previous time: the variable defining whether the moving target object has been raised at the current time or the previous time is a hidden variable, a discrete variable, and a non-continuous object association factor H^ACTDEDThe observed quantity of (2). Non-continuous object association factor H^ACTDED: the variable defining whether the moving target object has been raised at the current time or the previous time is a hidden variable and is a discrete variable. The values are set to {0,1}, which respectively represent the arm not lifted (arm not lifted at the previous moment or arm not lifted at the current moment or arm lifted at both the previous moment and the current moment) and the arm lifted (arm lifted at the current moment and arm lifted at the previous moment). The discontinuous object associated factor variable is used as the record of the discontinuous object factor, and the transition probability relation is the discontinuous pair quantity associated factor at the previous momentAnd the logical or relationship between the element and the Boolean value of the discontinuous object factor node at the current moment.

The above observations need the execution body to update at every moment according to the sensor data, which is a key factor for explaining the intention of the moving target object.

Specifically, the intention identifying method further includes:

Specifically, the hidden node corresponds to each node corresponding to a conditional variable set in the dynamic bayesian network. Wherein the condition variable set H represents factors related to the intention of the moving target object, e.g. dynamic environment factors H^DYNStatic environmental factor H^STATContinuous object factor H^ACTCDiscontinuous object factor H^ACTDAnd a non-continuous object association factor H^ACTDED。

The state transition probability corresponding to the discontinuous object association factor node in the hidden node can be specifically obtained by the following formula: p (H)_t ^ACTDED|H_t-1 ^ACTDED,H_t ^ACTD)。P(H_t ^ACTDED|H_t-1 ^ACTDED,H_t ^ACTD) Represents H_t ^ACTDAnd H_t ^ACTDED|H_t-1 ^ACTDEDProbability of co-occurrence. Wherein H_t ^ACTDED|H_t-1 ^ACTDEDThe probability of occurrence is at H_t-1 ^ACTDEDUnder the conditions that H occurs_t ^ACTDEDThe probability of occurrence.

P(H_t ^ACTDED|H_t-1 ^ACTDED,H_t ^ACTD) In (H)_t-1 ^ACTDEDRepresenting a non-continuous object association factor at time t-1 (which may also be understood as a time immediately preceding the current time), H_t ^ACTDEDRepresenting a non-continuous object association factor at time t (which may also be understood as the current time), H_t ^ACTDRepresents at time t: (Also understood as the current time) of the non-continuous object factor, H_t-1 ^ACTDRepresenting a non-continuous object factor at time t-1 (which may also be understood as a time immediately preceding the current time).

And step S102, calling the dynamic Bayesian network to determine an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment and determine a state transition probability of each node in the dynamic Bayesian network.

And step S103, determining the prior probability of the intention identification corresponding to the intention node in the dynamic Bayesian network at the current moment according to the posterior probability of the intention identification and the state transition probability of each node.

Specifically, prior to invoking the dynamic bayesian network, the intent recognition method further comprises constructing the dynamic bayesian network. The specific construction steps are as follows:

acquiring an observation variable set O corresponding to the obstacle behavior information at the historical time, wherein the observation variable set O is { O ═ O^DYN，O^STAT，O^ACTD，O^ACTCWherein the minimum distance O between the moving target object and the self-vehicle^DYN: defining the minimum distance D between the moving target object and the vehicle when the vehicle drives forwards according to the current speed^minIs a dynamic environmental factor H^DYNThe corresponding observed variable is a continuous variable. The actual value range is [ -10, 20 [)]And m is selected. Longitudinal distance O between moving target object and intersection^STAT: defining a relative longitudinal distance D of a moving target object from an intersection^tiIs a static environmental factor H^STATThe corresponding observed variable is a continuous variable. The actual value range is [ -10, 20 [)]And m is selected. Moving target object arm raising probability O^ACTD: defining a moving target object lift arm probability P output by a multi-tasking detector_arIs a discontinuous object factor H^ACTDThe corresponding observed variable is a continuous variable. The value interval is [0,1 ]]The score value of the arm raising action of the moving target object output by the detector is indicated, and the closer the score is to 1, the higher the probability of arm raising is. Moving target torso orientation angle O^ACTC: defining a moving target object torso orientation for the normalized multi-tasking detector outputAngle, is a continuous object factor H^ACTCThe corresponding observed variable is a continuous variable. The value interval is [0 °, 360 °), and represents the orientation angle of the moving target object in the moving target object fixed coordinate system established with the first frame orientation angle at which the moving target object is detected being 0 degrees.

Determining a condition variable set H corresponding to the obstacle behavior information based on the corresponding relation between the preset obstacle behavior information and the condition variable, wherein the condition variable set H is { H }^DYN，H^STAT，H^ACTD，H^ACTDED，H^ACTC}。

Wherein, the dynamic environment factor H^DYN: variables defining whether potential collision risks exist between the moving target object and the own vehicle, namely hidden variables and discrete variables. The value sets are {0,1}, which respectively represent no collision risk and collision risk. Static environmental factor H^STAT: variables, hidden variables and discrete variables, which define whether the moving target object is in the intersection area. The value sets are {0,1}, which respectively represent that the intersection region is not located and the intersection region is located. Discontinuous object factor H^ACTD: variables for defining whether the moving target object lifts the arm, hidden variables and discrete variables. The values are set to {0,1}, which respectively represent the arm not lifted and the arm lifted. Non-continuous object association factor H^ACTDED: variables defining whether the moving target object has been raised at the current time or the previous time, hidden variables and discrete variables. The values are set to {0,1}, which respectively represent the arm that was not lifted and the arm that was lifted. The variable is used as a record of discontinuous object factors, and the transition probability relation is H_t-1 ^ACTDEDAnd H_t ^ACTDEDLogical or relationships between node boolean values. Continuous object factor H^ACTC: variables defining the destination of the moving target object, hidden variables, discrete variables. The set of values is {0,1,2}, indicating that the destination of the moving target object is directly in front, to the left, and to the right, respectively.

Obtaining intention G corresponding to obstacle behavior information at historical time (e.g. t-1 time)_t-1. By way of example, the moving target object intent G: variables defining the intention of a moving target object, hidden variablesDiscrete variables. The value sets are {0,1,2}, which respectively indicate that the intention of the moving target object at the intersection is straight, left-turning and right-turning.

Taking each condition variable in the condition variable set as a hidden node H in the network_t-1 ^ACTD、H_t-1 ^ACTDED、H_t-1 ^ACTC、H_t-1 ^DYN、H_t-1 ^STAT、H_t ^ACTD、H_t ^ACTDED、H_t ^ACTC、H_t ^DYN、H_t ^STATTaking each observation variable in the observation variable set as an observation variable node O in the network_t-1 ^ACTD、O_t-1 ^ACTC、O_t-1 ^DYN、O_t-1 ^STAT、O_t ^ACTD、O_t ^ACTC、O_t ^DYN、O_t ^STATCorresponding the historical obstacle behavior information to the intention G_t-1And intention G of the current time_tAs the intention nodes in the network, the intention nodes G respectively corresponding hidden nodes at the historical time (for example, t-1 time) to the obstacle behavior information at the historical time_t-1Connected and pointed to the intention node G by hidden nodes_t-1The hidden nodes at the historical moment are respectively connected with the corresponding observation variable nodes, the hidden nodes at the historical moment point to the corresponding observation variable nodes, the hidden nodes at the current moment, the observation variable nodes and the intention nodes are also set according to the setting of the historical moment, the detailed description is omitted, and the hidden nodes at the historical moment point to the same hidden node at the current moment (for example, the moment t) and are connected (for example, H)_t-1 ^ACTDPoint of direction H_t ^ACTD、H_t-1 ^ACTDEDPoint of direction H_t ^ACTDED、H_t-1 ^ACTCPoint of direction H_t ^ACTC、H_t-1 ^DYNPoint of direction H_t ^DYN、H_t-1 ^STATPoint of direction H_t ^STATAnd connected) to point the intent node at the historical time to the intent node at the current time and connected (e.g., G)_t-1Direction G_tAnd are connected to each other to represent G_tAlso dependent on G_t-1) So as to construct a dynamic bayesian network as shown in fig. 4, where t-1 represents the historical time and t represents the current time. And then updating the hidden node corresponding to the current moment according to the observation variable corresponding to the acquired obstacle behavior information at each moment, and further updating the intention node corresponding to the current moment based on the updated hidden node of the current moment and each hidden node, observation variable node and intention node of the historical moment, so as to determine the intention of the obstacle at the current moment, for example, the intention of a moving target object (which may be a rider).

The embodiment takes the intention influence factor as a hidden node in the dynamic bayesian network and takes the observed quantity corresponding to the intention influence factor as an observed variable in the dynamic bayesian network through analyzing the behavior related factors of the moving target object (which can be a cyclist) in the prior art. According to the observation quantity received at each moment, the corresponding hidden variable is updated, the hidden variable determines the probability distribution of the intention node of the moving target object (which can be a cyclist), a dynamic Bayesian network is constructed, and the intention of the moving target object can be quickly and accurately identified based on the constructed dynamic Bayesian network.

In the dynamic bayesian network shown in fig. 4, the transition probability relationship corresponding to all hidden nodes in the conditional variable set is:

wherein t represents the time t and can be used for representing the current time; t-1 represents the time t-1, i.e. the time immediately preceding the time t, and can be used to characterize the time immediately preceding the current time, i.e. the historical time.

In the above formula, P (H)_t|H_t-1) Represented by conditional probabilities, i.e. represented at H_t-1Under the conditions that H occurs_tThe probability of occurrence. P (H)_t ^DYN|H_t-1 ^DYN) Is represented by H_t-1 ^DYNUnder the conditions which occur in the process of the present invention,H_t ^DYNthe probability of occurrence. P (H)_t ^STAT|H_t-1 ^STAT) Is represented by H_t-1 ^STATUnder the conditions that H occurs_t ^STATThe probability of occurrence. P (H)_t ^ACTDED|H_t-1 ^ACTDED,H_t ^ACTD) Represents H_t ^ACTDAnd H_t ^ACTDED|H_t-1 ^ACTDEDProbability of co-occurrence. Wherein H_t ^ACTDED|H_t-1 ^ACTDEDThe probability of occurrence is at H_t-1 ^ACTDEDUnder the conditions that H occurs_t ^ACTDEDThe probability of occurrence. P (H)_t ^ACTD|H_t-1 ^ACTD) Is represented by H_t-1 ^ACTDUnder the conditions that H occurs_t ^ACTDThe probability of occurrence. P (H)_t ^ACTC|H_t-1 ^ACTC) Is represented by H_t-1 ^ACTCUnder the conditions that H occurs_t ^ACTCThe probability of occurrence. Wherein H_t-1 ^DYNRepresenting the dynamic environmental factors at time t-1 (which may also be understood as the time immediately preceding the current time); h_t ^DYNRepresenting the dynamic environmental factors at time t (which may also be understood as the current time). H_t-1 ^STATRepresenting static environmental factors at time t-1 (which may also be understood as a time immediately preceding the current time); h_t ^STATRepresenting the static environmental factors at time t (which may also be understood as the current time). H_t-1 ^ACTDEDRepresenting a non-continuous object association factor at time t-1 (which may also be understood as a time immediately preceding the current time), H_t ^ACTDEDRepresenting a non-continuous object association factor at time t (which may also be understood as the current time), H_t ^ACTDRepresenting a non-continuous object factor at time t (which may also be understood as the current time), H_t-1 ^ACTDRepresenting a non-continuous object factor at time t-1 (which may also be understood as a time immediately preceding the current time). H_t-1 ^ACTCRepresenting a continuous object factor at time t-1 (which may also be understood as a time immediately preceding the current time); h_t ^ACTCRepresenting time t (as will also be understood)Current time of day).

The posterior probability of the intention identification corresponding to the intention node in the dynamic Bayesian network at the previous moment can be directly obtained. The execution subject can identify the posterior probability according to the intentions corresponding to the intention nodes in the dynamic Bayesian network at the previous moment

And the calculated state transition probability P (H) of each node_t|H_t-1) Calculating and obtaining the combined prior distribution of the intention identification corresponding to the intention node in the dynamic Bayesian network at the current moment

That is, the prior probability of the identification of the intention corresponding to the intention node in the dynamic bayesian network at the current time is obtained by the following calculation:

wherein G is_tMay indicate the intention of time t, i.e. the current time, G_t-1May indicate the intention of the moment t-1, i.e. the moment preceding the current moment, H_tSet of conditional variables, H, which can represent time t, i.e. the current time_t-1A set of conditional variables may be represented at time t-1, i.e., a time prior to the current time. P (G)_t|G_t-1,H_t) Represents H_tOccurs at G_t-1Conditions of occurrence of G_tGeneration of H_tAnd G_t|G_t-1The probability of the two occurring together. P (H)_t|H_t-1) Representing the state transition probability of each node. P (G)_t-1，H_t-1) And identifying posterior probabilities of intentions corresponding to the intention nodes in the dynamic Bayesian network at the previous moment.

And step S104, updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment.

In this embodiment, the dynamic bayesian network can be regarded as a forward filtering process, and when the observed quantity at each time is received, the probability distribution of each variable can be updated, so as to complete the intention inference process. The intent inference process employs an Assumed Density Filtering (ADF) method as the inference tool. The intent inference process can be divided into two processes, prediction and update. The intent inference process at a time (e.g., the current time, i.e., time t) is based on the joint posterior distribution at the previous time (i.e., time t-1) as shown in FIG. 5

According to a fixed transition probability P (H)_t|H_t-1) The state prior distribution at the moment can be obtained

And then obtaining the observed quantity and the conditional probability P (O) according to the observed quantity obtained at the moment (for example, the observed quantity and the conditional probability P (O) are obtained at the moment t shown in FIG. 5)_t|H_t) Obtaining the state posterior distribution at the moment

And further based on the state posterior distribution at that time

To obtain the intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the moment (for example, the current moment, for example, t moment)

The prediction process comprises the following steps:

specifically, the updating of the prior probability of the intention recognition according to the behavior information of the obstacle includes:

In this embodiment, the prediction process is to predict the prior distribution at the previous time according to the posterior distribution and the fixed transition probability at the previous time. Based on last moment combined posterior distribution

According to transition probability P (H)_t|H_t-1) The joint prior distribution of the moment can be calculated

And adding to obtain edge distribution

I.e. the state prior distribution at that instant, which may be the current instant, i.e. the instant t, for example.

In particular, the edge distribution at the current time

The calculation of (a) is shown as follows:

wherein the content of the first and second substances,

joint prior distribution probability, G, representing intent recognition_tMay indicate the intention of time t, i.e. the current time, G_t-1May indicate the intention of the moment t-1, i.e. the moment preceding the current moment, H_tSet of conditional variables, H, which can represent time t, i.e. the current time_t-1Can represent the condition change of t-1 time, i.e. the time before the current timeAnd (6) collecting the quantity. P (H)_t|H_t-1) Which represents the probability of the transition,

representing the combined posterior distribution at the last instant.

Then, the execution subject can calculate the state posterior probability of the moment (for example, the current moment, namely the t moment) according to the observation variable and the conditional probability relation

Wherein, the conditional probability can be obtained by the following formula (t in the formula represents t time):

wherein, P (O)_t ^DYN|H_t ^DYN) Is shown in H_t ^DYNUnder the conditions that O occurs_t ^DYNThe probability of occurrence. H_t ^DYNThe dynamic environment factor at time t (i.e., the current time) is shown. O is_t ^DYNThe observation variable corresponding to the dynamic environment factor at the time t (namely, the current time) is represented, namely, the minimum distance between the moving target object at the time t (namely, the current time) and the vehicle is represented. P (O)_t ^STAT|H_t ^STAT) Is shown at H_t ^STATUnder the conditions that O occurs_t ^STATThe probability of occurrence. P (O)_t ^ACTD|H_t ^ACTD) Is shown at H_t ^ACTDUnder the conditions that O occurs_t ^ACTDThe probability of occurrence. P (O)_t ^ACTC|H_t ^ACTC) Is shown at H_t ^ACTCUnder the conditions that O occurs_t ^ACTCThe probability of occurrence. H_t ^STATIndicating time t (alsoThat is the current time). O is_t ^STATThe observation variable corresponding to the static environment factor at the time t (namely the current time) is represented, namely the longitudinal distance between the moving target object at the time t (namely the current time) and the intersection. H_t ^ACTDA non-continuous object factor at time t (i.e., the current time) is indicated. O is_t ^ACTDThe observation variable corresponding to the discontinuous object factor at the time t (namely the current time) is represented, namely the arm raising probability of the moving target object at the time t (namely the current time). H_t ^ACTCThe continuous object factor at time t (i.e., the current time) is represented. O is_t ^ACTCThe observation variable corresponding to the continuous object factor at the time t (namely, the current time) is represented, namely the moving target object body orientation angle at the time t (namely, the current time).

Specifically, the state posterior probability at this time (which may be the current time, i.e., time t) is:

the meaning of each letter in the formula has been explained in the foregoing steps, and is not described herein again.

And summing the state posterior probabilities to execute an updating process, and further updating the intention identification prior probability to obtain the intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the current moment. The updating process is to update the posterior distribution at this time according to the observed variable at this time (which may be the current time, i.e., time t). Based on joint prior distribution obtained in prediction process

Calculating the combined posterior distribution of the moment according to the observation variable and the conditional probability relation

And add to obtain the posterior distribution of the node G

Specifically, the calculation can be obtained by the following formula:

intention identification posterior probability corresponding to intention node in dynamic Bayesian network at current moment

Comprises the following steps:

wherein G is_tCan indicate the intention of the time t, i.e. the current time, H_tA set of conditional variables at time t, i.e. the current time, can be represented.

And step S105, determining the intention corresponding to the posterior probability of intention recognition according to the preset corresponding relation between the intention recognition probability and the intention.

Specifically, the intent may be that the intent of the moving target object at the intersection is to go straight, turn left, and turn right.

In the embodiment of the application, a probability relation exists between the moving target object intention variable G node and each condition variable and the G node at the previous moment. I.e. the existence of the conditional probability relation P (G)_t|G_t-1，H_t ^DYN，H_t ^STAT，H_t ^ACTD，H_t ^ACTDED，H_t ^ACTC) Abbreviated as P (G)_t|G_t-1，H_t)。

In summary, there are three types of probability relationships in a dynamic bayesian network, the state transition conditional probability P (H) of a conditional variable_t|H_t-1) Conditional probability relationship P (G) between intention variable and condition variable_t|G_t-1，H_t) Conditional probability relationship P (O) between conditional variable and observed variable as an interpretation model of the intent of the moving target object_t|H_t) As a descriptive model of the intent of the moving target object. And then based on intents in the dynamic Bayesian network at the current momentThe intention corresponding to the node identifies the posterior probability and the description model of the intention of the moving target object to accurately determine the intention of the moving target object.

The embodiment obtains the obstacle behavior information in the intention identification request by receiving the intention identification request; calling the dynamic Bayesian network to determine an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment and a state transition probability of each node, and further determining an intention identification prior probability corresponding to the intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node; updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment; and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention. Therefore, the intention of the moving target object at the intersection can be accurately identified.

Fig. 2 is a schematic main flow diagram of an intention identification method according to a second embodiment of the present application, and as shown in fig. 2, the intention identification method includes:

step S201, receiving the intention identification request, and acquiring the obstacle behavior information carried in the intention identification request.

Step S202, calling the dynamic Bayesian network, determining an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment, and determining a state transition probability of each node in the dynamic Bayesian network.

Step S203, according to the intention identification posterior probability and the state transition probability of each node, the intention identification prior probability corresponding to the intention node in the dynamic Bayesian network at the current moment is determined.

The principle of step S201 to step S203 is similar to that of step S101 to step S103, and is not described here again.

Specifically, step S203 can also be realized by step S2031 to step S2032:

step S2031, determining transition probabilities of all nodes according to a product of the state transition probabilities of the nodes.

Specifically, in the dynamic bayesian network, the transition probability relationship corresponding to all hidden nodes in the conditional variable set is as follows:

wherein t represents the time t and can be used for representing the current time; t-1 represents time t-1, i.e., the time immediately preceding time t, and may be used to characterize the time immediately preceding the current time.

In the above formula, P (H)_t|H_t-1) Represented by conditional probabilities, i.e. represented at H_t-1Under the conditions that H occurs_tThe probability of occurrence. P (H)_t ^DYN|H_t-1 ^DYN) Is represented by H_t-1 ^DYNUnder the conditions that H occurs_t ^DYNThe probability of occurrence. P (H)_t ^STAT|H_t-1 ^STAT) Is represented by H_t-1 ^STATUnder the conditions that H occurs_t ^STATThe probability of occurrence. P (H)_t ^ACTDED|H_t-1 ^ACTDED,H_t ^ACTD) Represents H_t ^ACTDAnd H_t ^ACTDED|H_t-1 ^ACTDEDProbability of co-occurrence. Wherein H_t ^ACTDED|H_t-1 ^ACTDEDThe probability of occurrence is at H_t-1 ^ACTDEDUnder the conditions that H occurs_t ^ACTDEDThe probability of occurrence. P (H)_t ^ACTD|H_t-1 ^ACTD) Is represented by H_t-1 ^ACTDUnder the conditions that H occurs_t ^ACTDThe probability of occurrence. P (H)_t ^ACTC|H_t-1 ^ACTC) Is represented by H_t-1 ^ACTCUnder the conditions that H occurs_t ^ACTCThe probability of occurrence. Wherein H_t-1 ^DYNRepresenting the dynamic environmental factors at time t-1 (which may also be understood as the time immediately preceding the current time); h_t ^DYNRepresenting motion at time t (also understood to be the current time)And (4) ecological environment factors. H_t-1 ^STATRepresenting static environmental factors at time t-1 (which may also be understood as a time immediately preceding the current time); h_t ^STATRepresenting the static environmental factors at time t (which may also be understood as the current time). H_t-1 ^ACTDEDRepresenting a non-continuous object association factor at time t-1 (which may also be understood as a time immediately preceding the current time), H_t ^ACTDEDRepresenting a non-continuous object association factor at time t (which may also be understood as the current time), H_t ^ACTDRepresenting a non-continuous object factor at time t (which may also be understood as the current time), H_t-1 ^ACTDRepresenting a non-continuous object factor at time t-1 (which may also be understood as a time immediately preceding the current time). H_t-1 ^ACTCRepresenting a continuous object factor at time t-1 (which may also be understood as a time immediately preceding the current time); h_t ^ACTCRepresenting the continuous object factor at time t (which may also be understood as the current time).

Step S2032, according to the transition probability P (H) of all nodes_t|H_t-1) Intention recognition posterior probability corresponding to historical time (e.g., t-1 time)

And determining the prior probability of the intention identification corresponding to the intention node in the dynamic Bayesian network at the current moment.

Specifically, the posterior probability of the intention identification corresponding to the intention node in the dynamic bayesian network at the previous moment of the current moment can be directly obtained. The execution subject can identify the posterior probability according to the intention corresponding to the intention node in the dynamic Bayesian network at the previous moment (namely t-1 moment) of the current moment (namely t moment) acquired

Specifically, it can be calculated by the following formula:

In the embodiment, the prior probability of the intention recognition corresponding to the intention node in the dynamic bayesian network at the current moment is preliminarily determined according to the transition probability of each node and the posterior probability of the intention recognition corresponding to the historical moment, so that preparation is made for the finally obtained posterior probability of the intention recognition at the current moment, and the posterior probability of the intention recognition at the current moment is accurately determined.

And step S204, updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment.

Step S205, determining the intention corresponding to the posterior probability of intention recognition according to the corresponding relation between the preset intention recognition probability and the intention.

The principle of step S204 to step S205 is similar to that of step S104 to step S105, and is not described here again.

In the embodiment of the application, the distribution function of the observation variable is a corresponding empirical distribution function determined according to the distribution condition of the training data marked in the moving target object database, parameters of all the distribution functions are obtained through a Maximum Likelihood Estimation (MLE) algorithm, and parameters of the mixed gaussian distribution are determined through an Expectation Maximization (EM) algorithm. As an example, the specific choices are as follows:

1) under the condition of the dynamic environment state, the probability Distribution of the minimum distance between the moving target object and the self-vehicle conforms to Gamma Distribution (Gamma Distribution), and the parameters of the Distribution function comprise a shape parameter a and a scale parameter b.

In particular, the amount of the solvent to be used,

2) under the condition of the static environment state, the probability Distribution of the longitudinal distance between the moving target object and the intersection conforms to the Mixture of gaussian distributions (Mixture of m Gaussians Distribution). According to the sample data, D^tiThere are multiple modes, choose when H^STATWhen H is false, m is 2^STATWhen tu, m is 1, the parameters of the Gaussian mixture include the mean μ^(k)Variance σ^(k)And mixing weights

In particular, the amount of the solvent to be used,

3) in a discontinuous object state H^ACTDUnder the condition, the probability distribution P (O) of the arm raising probability Par of the moving target object^ACTD|H^ACTD) Compliant Beta Distribution (Beta Distribution), the parameters of the Distribution function include α and β:

in particular, the amount of the solvent to be used,

4) in a continuous object state H^ACTDConditional on moving target objectTorso orientation angle θ_bodyProbability distribution P (O)^ACTC|H^ACTC) In compliance with the Weibull Distribution (Weibull Distribution), the parameters include λ and k.

In particular, the amount of the solvent to be used,

the moving target object intention variable G node has a probability relation with each condition variable and the G node at the previous moment. I.e. the existence of the conditional probability relation P (G)_t|G_t-1，H_t ^DYN，H_t ^STAT，H_t ^ACTD，H_t ^ACTDED，H_t ^ACTC) Abbreviated as P (G)_t|G_t-1，H_t)。

In summary, there are three types of probability relationships in a dynamic bayesian network, the state transition conditional probability P (H) of a conditional variable_t|H_t-1) Conditional probability relationship P (G) between intention variable and condition variable_t|G_t-1，H_t) Conditional probability relationship P (O) between conditional variable and observed variable as an interpretation model of the intent of the moving target object_t|H_t) As a descriptive model of the intent of the moving target object.

Fig. 3 is a schematic view of an application scenario of an intention identifying method according to a third embodiment of the present application. The intention identification method can be applied to a scene for predicting the riding intention of the moving target object when the moving target object meets the automatic driving vehicle at the intersection. As shown in fig. 3, the server 303 receives the intention identification request 301, and acquires the obstacle behavior information 302 carried in the intention identification request 301. Server 303 invokes dynamic bayesian network 304 to determine the intended identification posterior probabilities 305 corresponding to the intended nodes in dynamic bayesian network 304 at the previous time and to determine the state transition probabilities 306 of the nodes in the dynamic bayesian network. According to the intention identification posterior probability 305 and the state transition probability 306 of each node, the intention identification prior probability 307 corresponding to the intention node in the dynamic Bayesian network 304 at the current moment is determined. The server 303 updates the prior probability 307 of the intention recognition according to the obstacle behavior information 302, and further obtains a posterior probability 308 of the intention recognition corresponding to the intention node in the dynamic bayesian network 304 at the current time. The server 303 determines an intention 310 corresponding to the intention recognition posterior probability 308 according to a preset correspondence 309 between the intention recognition probability and the intention.

The embodiment of the application provides an obstacle intention identification method based on a dynamic Bayesian network based on intention identification related obstacle (for example, a moving target object) multi-element characteristic, and the obstacle intention identification method is used for deducing the passing intention of the obstacle in an intersection scene. The embodiment of the application provides an obstacle intention identification method based on a dynamic Bayesian network. The method comprises the steps of selecting factors influencing the intention of an obstacle from three angles of dynamic environment, static environment and object factors, and designing a dynamic Bayesian network which takes the three factors as hidden variables and takes the characteristics of the minimum distance between the obstacle and a vehicle, the orientation of the trunk of the obstacle and the like as corresponding observation variables. On the basis of the dynamic Bayesian network, a conditional probability relation between hidden variables and observed variables and an inference method of an obstacle intention are provided, and the inference method is used for intention identification of obstacles in an intersection scene.

Fig. 6 is a schematic diagram of main units of an intention identifying apparatus according to an embodiment of the present application. As shown in fig. 6, the intention identifying apparatus includes a receiving unit 601, a prior probability determining unit 602, a posterior probability determining unit 603, and an intention identifying unit 604.

The receiving unit 601 is configured to receive the intention identification request, and acquire the obstacle behavior information in the intention identification request.

A state transition probability determination unit 602 configured to invoke the dynamic bayesian network to determine an intention identification posterior probability corresponding to an intention node in the dynamic bayesian network at a previous time and a state transition probability of each node.

A prior probability determining unit 603 configured to determine an intention identification prior probability corresponding to an intention node in the dynamic bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node.

The posterior probability determining unit 604 is configured to update the prior probability of the intention identification according to the behavior information of the obstacle, so as to obtain the posterior probability of the intention identification corresponding to the intention node in the dynamic bayesian network at the current moment.

An intention identifying unit 605 configured to determine an intention corresponding to the intention identification posterior probability according to a preset correspondence between the intention identification probability and the intention.

In some embodiments, the intent recognition apparatus further comprises a dynamic bayesian network building unit, not shown in fig. 6, configured to: acquiring an observation variable set corresponding to the barrier behavior information at the historical moment; determining a condition variable set corresponding to the obstacle behavior information based on the corresponding relation between the preset obstacle behavior information and the condition variable; acquiring an intention corresponding to the obstacle behavior information at the historical moment; and taking each condition variable in the condition variable set as a hidden node in the network, taking each observation variable in the observation variable set as an observation variable node in the network, taking an intention corresponding to the barrier behavior information at the historical moment as an intention node in the network, updating the corresponding hidden node according to the observation variable corresponding to the barrier behavior information acquired at each moment, and updating the corresponding intention node based on the updated hidden node to construct and obtain the dynamic Bayesian network.

In some embodiments, the intent recognition apparatus further comprises a request generation unit, not shown in fig. 6, configured to: in response to determining that the current position is detected to be located in a preset intersection area, acquiring obstacle behavior information located in the preset intersection area; based on the obstacle behavior information, an intent recognition request is generated.

In some embodiments, the receiving unit 601 is further configured to: the minimum distance between the obstacle and the vehicle, the relative longitudinal distance between the obstacle and the intersection, the obstacle orientation angle, the obstacle limb action occurrence probability and whether the obstacle has ever acted on limbs at the current moment or the previous moment are obtained.

In some embodiments, the state transition probability determination unit 602 is further configured to: and determining the state transition probability corresponding to the discontinuous object association factor node in the hidden node according to whether the barrier has limb movement at the current time and the previous time.

In some embodiments, the prior probability determination unit 603 is further configured to: determining the transition probability of all nodes according to the product of the state transition probability of each node; and determining the prior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment according to the transition probabilities and the posterior probability of the intention recognition of all the nodes.

In some embodiments, the a posteriori probability determination unit 604 is further configured to: calculating to obtain state posterior probability according to the behavior information and the conditional probability of the barrier; and summing the state posterior probabilities, and further updating the prior probability of the intention recognition so as to obtain the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment.

In the present application, the intention identifying method and the intention identifying device have corresponding relations in the concrete implementation contents, and therefore, the description of the repeated contents is omitted.

Fig. 7 illustrates an exemplary system architecture 700 to which the intent recognition method or intent recognition apparatus of embodiments of the present application may be applied.

As shown in fig. 7, the system architecture 700 may include

terminal devices

701, 702, 703, a network 704, and a server 705. The network 704 serves to provide a medium for communication links between the

terminal devices

701, 702, 703 and the server 705. Network 704 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.

A user may use the

terminal devices

701, 702, 703 to interact with a server 705 over a network 704, to receive or send messages or the like. The

terminal devices

701, 702, 703 may have installed thereon various communication client applications, such as a shopping-like application, a web browser application, a search-like application, an instant messaging tool, a mailbox client, social platform software, etc. (by way of example only).

The

terminal devices

701, 702, 703 may be various electronic devices having an intention recognition processing screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.

The server 705 may be a server providing various services, such as a background management server (for example only) providing support for intention-identifying requests submitted by users using the

terminal devices

701, 702, 703. The background management server can receive the intention identification request and acquire the obstacle behavior information carried in the intention identification request; and calling the dynamic Bayesian network to determine an intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the previous moment and determine the state transition probability of each node in the dynamic Bayesian network. Determining an intention identification prior probability corresponding to an intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node; updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment; and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention. Thus, the intention of the obstacle at the intersection can be accurately identified.

It should be noted that the intention identification method provided in the embodiment of the present application is generally executed by the server 705, and accordingly, the intention identification device is generally disposed in the server 705.

It should be understood that the number of terminal devices, networks, and servers in fig. 7 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

Referring now to FIG. 8, shown is a block diagram of a computer system 800 suitable for use in implementing a terminal device of an embodiment of the present application. The terminal device shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

As shown in fig. 8, the computer system 800 includes a Central Processing Unit (CPU)801 that can perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)802 or a program loaded from a storage section 808 into a Random Access Memory (RAM) 803. In the RAM803, various programs and data necessary for the operation of the computer system 800 are also stored. The CPU801, ROM802, and RAM803 are connected to each other via a bus 804. An input | output (I | O) interface 805 is also connected to bus 804.

The following components are connected to the I | O interface 805: an input portion 806 including a keyboard, a mouse, and the like; an output section 807 including a signal such as a Cathode Ray Tube (CRT), a liquid crystal credit authorization query processor (LCD), and the like, and a speaker; a storage portion 808 including a hard disk and the like; and a communication section 809 including a network interface card such as a LAN card, a modem, or the like. The communication section 809 performs communication processing via a network such as the internet. A drive 810 is also connected to the I | O interface 805 as necessary. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 810 as necessary, so that a computer program read out therefrom is mounted on the storage section 808 as necessary.

In particular, according to embodiments disclosed herein, the processes described above with reference to the flow diagrams may be implemented as computer software programs. For example, embodiments disclosed herein include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 809, and | or installed from the removable medium 811. The computer program executes the above-described functions defined in the system of the present application when executed by the Central Processing Unit (CPU) 801.

It should be noted that the computer readable medium shown in the present application may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In this application, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units described in the embodiments of the present application may be implemented by software or hardware. The described units may also be provided in a processor, and may be described as: a processor includes a receiving unit, a state transition probability determining unit, a prior probability determining unit, a posterior probability determining unit, and an intent identifying unit. Wherein the names of the elements do not in some way constitute a limitation on the elements themselves.

As another aspect, the present application also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs, and when the one or more programs are executed by the equipment, the equipment receives the intention identification request and acquires the obstacle behavior information carried in the intention identification request; and calling the dynamic Bayesian network to determine an intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the previous moment and determine the state transition probability of each node in the dynamic Bayesian network. Determining an intention identification prior probability corresponding to an intention node in the dynamic Bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node; updating the prior probability of the intention recognition according to the behavior information of the obstacle, and further obtaining the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment; and determining the intention corresponding to the posterior probability of the intention recognition according to the corresponding relation between the preset intention recognition probability and the intention.

According to the technical scheme of the embodiment of the application, the intention of the barrier at the intersection can be accurately identified.

The above-described embodiments should not be construed as limiting the scope of the present application. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An intent recognition method, comprising:

receiving an intention identification request, and acquiring barrier behavior information carried in the intention identification request;

calling a dynamic Bayesian network, determining an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment, and determining a state transition probability of each node in the dynamic Bayesian network;

determining an intention identification prior probability corresponding to an intention node in the dynamic Bayesian network at the current moment according to the intention identification prior probability and the state transition probability of each node;

updating the prior probability of the intention recognition according to the behavior information of the obstacle so as to obtain the posterior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment;

2. The method of claim 1, wherein prior to said invoking the dynamic bayesian network, the method further comprises:

and taking each condition variable in the condition variable set as a hidden node in the network, taking each observation variable in the observation variable set as an observation variable node in the network, taking the intention corresponding to the barrier behavior information at the historical moment as an intention node in the network, updating the corresponding hidden node according to the observation variable corresponding to the barrier behavior information acquired at each moment, and updating the corresponding intention node based on the updated hidden node to construct the dynamic Bayesian network.

3. The method of claim 1, wherein prior to said receiving an intent recognition request, the method further comprises:

in response to the fact that the current position is detected to be located in a preset intersection area, acquiring obstacle behavior information located in the preset intersection area;

4. The method according to claim 2, wherein the obtaining of the obstacle behavior information carried in the intention identification request comprises:

and acquiring the minimum distance between the obstacle and the vehicle, the relative longitudinal distance between the obstacle and the intersection, the obstacle orientation angle, the obstacle limb action occurrence probability and whether the obstacle has limb action at the current moment or the previous moment, wherein the obstacle is carried in the intention identification request.

5. The method of claim 4, further comprising:

and determining the state transition probability corresponding to the discontinuous object association factor node in the hidden nodes according to whether the barrier has a limb action at the current time and the previous time.

6. The method according to claim 1, wherein the determining an intention identification prior probability corresponding to an intention node in the dynamic bayesian network at the current moment according to the intention identification posterior probability and the state transition probability of each node comprises:

determining the transition probability of all nodes according to the product of the state transition probabilities of all the nodes;

and determining the prior probability of the intention recognition corresponding to the intention node in the dynamic Bayesian network at the current moment according to the transition probabilities of all the nodes and the posterior probability of the intention recognition.

7. The method of claim 1, wherein said updating the prior probability of intent recognition based on the obstacle behavior information comprises:

calculating to obtain a state posterior probability according to the barrier behavior information and the conditional probability;

and summing the state posterior probabilities, and further updating the intention identification prior probability to obtain the intention identification posterior probability corresponding to the intention node in the dynamic Bayesian network at the current moment.

8. An intention recognition apparatus, comprising:

the system comprises a receiving unit, a judging unit and a judging unit, wherein the receiving unit is configured to receive an intention identification request and acquire obstacle behavior information carried in the intention identification request;

the state transition probability determining unit is configured to call a dynamic Bayesian network, determine an intention identification posterior probability corresponding to an intention node in the dynamic Bayesian network at the previous moment and determine a state transition probability of each node in the dynamic Bayesian network;

a prior probability determination unit configured to determine an intention identification prior probability corresponding to an intention node in the dynamic bayesian network at a current moment according to the intention identification posterior probability and the state transition probability of each node;

a posterior probability determination unit configured to update the intention identification prior probability according to the obstacle behavior information, thereby obtaining an intention identification posterior probability corresponding to an intention node in the dynamic bayesian network at the current moment;

and the intention identification unit is configured to determine the intention corresponding to the intention identification posterior probability according to the corresponding relation between the preset intention identification probability and the intention.

9. The apparatus of claim 8, wherein the intent recognition apparatus further comprises a dynamic bayesian network building unit configured to:

10. The apparatus of claim 8, wherein the intent recognition apparatus further comprises a request generation unit configured to:

11. An intent recognition electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.

12. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-7.