JP4900967B2 - Driving support device and driving support method - Google Patents

Description

  The present invention relates to a driving support device and a driving support method that are mounted on a vehicle such as an automobile and a hybrid car and that create a driver model for each driver and that perform a driving support device for a driver based on the created driver model. .

  When the driver changes lanes while driving the vehicle, there is a blind spot that cannot be confirmed by the mirror. Further, when the driver looks directly, the amount of movement of the line of sight is larger than when the mirror is confirmed, and the time for viewing other than the front becomes longer, so the risk increases. Moreover, even if direct visual inspection is performed, in the case of a small vehicle such as a two-wheeled vehicle, there is a possibility that it may become a blind spot due to the shadow of a pillar.

Therefore, various driving assistance devices that perform driving assistance devices for the driver when the driver changes lanes have been proposed. For example, in Patent Document 1 (Japanese Patent Laid-Open No. 2008-15758), an imaging unit that is provided in a vehicle and can capture a rear side area including a lane different from the own lane, and an image captured by the imaging unit are displayed. A driving support device comprising a display means for displaying to a passenger so as to be visible, wherein the image indicates a mark indicating at least one of the degree to which the other vehicle on the rear side of the lane approaches or separates from the own vehicle. There is disclosed a driving support device characterized by comprising a mark display means for superimposing and displaying.
JP 2008-15758 A

  However, in the driving support device described in Patent Document 1, although a warning regarding the surrounding vehicle is given to the driver using the imaging unit, the vehicle is frequently present in the adjacent traveling lane during normal traveling, It is not good from the viewpoint of convenience to simply issue a warning by simply imaging the vehicle with the imaging means.

  In order to solve the above problems, behavior prediction based on the driver model is performed, and only when the driver tries to change lanes, for example, a camera image that covers the blind spot is automatically displayed, and there is a risk of collision. The present invention provides a driving support device and a driving support method for safely changing lanes by giving a warning when another vehicle or the like is present at a certain position.

For this purpose, the invention according to claim 1 acquires physiological state data means for acquiring data relating to the physiological state of the driver, driving operation data acquisition means for acquiring data relating to the driving operation, and acquisition of data relating to the driving situation. Driving status data acquisition means, table update means for updating a table that stochastically determines the causal relationship among physiological status data, driving status data, and driving operation data; and the table updated by the table updating means; Driving operation prediction means for predicting a driving operation from current driving situation data, and the network structure in the causal relationship is a node in which the physiological state data is affected by the driving situation data, driving operation data is the network structure as a node affected by the driving condition data and the physiological condition data A driving support apparatus according to claim and.

The invention according to claim 2 is the driving support device according to claim 1 , wherein the causal relationship of the network structure is determined by a Bayesian network.

Further, the invention according to claim 3 is the driving support device according to claim 1 or 2 , wherein warning notifying means for notifying a driver of warning based on the prediction of the driving operation predicting means; It is characterized by having.

According to a fourth aspect of the present invention, in the driving support device according to any one of the first to third aspects, the driving operation assist for assisting the driver in driving operation based on the prediction of the driving operation prediction means. And means.

Further, the invention according to claim 5 is the driving support device according to any one of claims 1 to 4 , wherein the driving status data acquisition means acquires a plurality of driving status data for acquiring data related to the driving status. It is characterized by that.

The invention according to claim 6 is a physiological state data acquisition step for acquiring data relating to a driver's physiological state, a driving operation data acquisition step for acquiring data relating to a driving operation, and data relating to a driving situation. A driving state data acquisition step , a physiological state data , a driving operation data, a step of updating a table that determines the causal relationship between the driving state data stochastically by using a Bayesian network, and the updated latest And predicting a driving operation from the current driving situation data and the current driving situation data, and in the network structure in the Bayesian network, the physiological state data is a node affected by the driving situation data, and the driving operation data is defined as a node affected by the driving condition data and the physiological condition data A driving support method comprising and.

  According to the driving support apparatus of the first aspect of the present invention, the driving operation is predicted by the table in which the causal relation is stochastically determined, so that efficient and accurate driving support can be performed.

  Further, according to the driving support device of the second aspect of the present invention, the driving operation is predicted and the driver is notified of a warning or the like by using the table in which the causal relationship is stochastically determined, so that it is efficient and accurate. Driving assistance can be performed.

  Further, according to the driving support device of the third aspect of the present invention, the driving operation is predicted by the table in which the causal relation is stochastically determined, and the driving operation is assisted to the driver. It becomes possible to perform a driving assistance.

  According to the driving support apparatus of the fourth aspect of the present invention, in addition to the causal relationship between the driving situation data and the driving operation data, the table update means has a physiological relationship that is not related to the driving operation or the driving situation. Since the table including the causal relationship with the state data is updated, the driving operation is predicted including the margin, so that efficient and accurate driving support can be performed.

  Further, according to the driving support apparatus of the fifth aspect of the present invention, the driving status data acquisition means acquires a plurality of driving status data for acquiring data related to the driving status, so that the driving operation prediction is performed more accurately. Thus, efficient and accurate driving support can be performed.

Further, according to the driving support apparatus of the sixth aspect of the present invention, the causal relationship between the driving situation data and the driving operation data is determined by the Bayesian network, and the driving operation is predicted based on the causal relationship. This makes it possible to provide accurate driving support.

  Moreover, according to the driving support method of the seventh aspect of the present invention, since the driving operation is predicted based on the table in which the causal relationship is determined stochastically, it is possible to perform efficient and accurate driving support. .

Hereinafter, preferred embodiments of the driving support apparatus of the present invention will be described with reference to the drawings.

(1) Outline of Embodiment FIG. 1 is a diagram showing a concept of a driving support apparatus according to an embodiment of the present invention. The downward arrow is a time-series image of driving behavior and driving environment. The left side of the arrow shows the configuration related to learning of the driver model, and the right side shows the configuration related to prediction of lane change.
In the driving support device of this embodiment, every time the driver drives a vehicle equipped with the device of this embodiment, a driver model for each driver is created, and the standard driver model stored in advance is The driver model created in the progressive form is taken into account in order and learned as a driver model for each individual. The driver model is created by driving data (for example, information on the vehicle, such as accelerator, brake, steering wheel operation amount, vehicle speed, acceleration, etc., distance between vehicles, road conditions, traffic volume, weather, etc.) This is done by collecting information around the vehicle). “Driver model learning data” (A in the figure) is data that is collected when such a driver model is created and is the basis for creation. In such a driver model, driving parameters for each driver and driving habits are reflected in the parameters. In the driving support apparatus of the present invention, in creating this driver model, causality between individual data of driving states is determined using an algorithm called a Bayesian network. When predicting driving behavior such as lane change based on the driver model, using this Bayesian network, even if there are missing parameters in the collected current `` driver model learning data '', Driving operation behavior can be easily estimated and output.

  Next, when storing the driver model in the driver model database, the “lane change driver model data” that records the lane change pattern for each individual and the pattern of the preceding action before the lane change was recorded. The “lane change preceding action driver model data” and the “normal driver model data” in which the pattern other than the time of the lane change including the preceding action is recorded and stored (B in the figure). In the present embodiment, the algorithm for distributing the three driver models also has a great feature.

  In the driving support device of the present embodiment, for each driver individual, a driver model at the time of “normal”, a driver model at the time of “lane change”, and a driver model at the time of “preceding action” are created and stored for each driver. (C in the figure).

  Next, a lane change is predicted using the driver model created as described above (D in the figure). When this lane change prediction is performed, both the collected current “driver model learning data” and the driver model that has been learned and accumulated are used. In this prediction, an algorithm called a Bayesian network is used.

Furthermore, in the driving assistance apparatus of this embodiment, when a lane change is predicted, driving assistance such as warning the driver or providing driver assistance as necessary is executed. (E in the figure). In the driving support device of this embodiment,
Since warning based on such lane change prediction is performed, efficient and accurate driving support can be performed.

  The driving support device according to the present embodiment collects driving state data (vehicle information such as accelerator, brake, steering wheel operation amount, vehicle speed, inter-vehicle distance, acceleration, etc.) while the driver is driving, From the driving state data, the part where the driver is changing the lane is extracted as “lane change preceding action driver model data”, and the part when the lane change is being executed is “lane change driver model data”. ", And the other part of normal operation is defined as" normal driver model data ". Accordingly, a driver model related to a driving pattern such as lane change can be automatically created without making the driver aware of it.

  In the driving support device of the present embodiment, for example, on a three-lane one-way national road, turn right at an intersection with a blue light from a right-turn exclusive lane, there is an oncoming vehicle, there are pedestrians on a pedestrian crossing, rain When it is falling and the driver is talking, the driver model is learned for each scene (situation) of the surrounding environment of the vehicle during traveling.

  Next, an outline of the Bayesian network used for the driver model of the driving support apparatus of the present invention will be described. FIG. 2 is a diagram showing a concept of a Bayesian network used in the driving support apparatus according to the embodiment of the present invention. A Bayesian network is a model that represents a probabilistic causal relationship in a graph structure. Each node X1, X2, X3, and X4 shown in FIG. 2 represents a random variable, the presence or absence of a causal relationship between them is a directed graph, and the causal relationship between nodes of the values that the random variable can take is a conditional probability Represented by The fact that there is a stochastic relationship between the node X1 and the node X2 is indicated by an arrow from the node X1 to the node X2. The node X2 is affected by the nodes X1 and X3, and the node X4 is affected by the node X3. The probability table taking into account such a causal relationship is a conditional probability table shown in FIG. It is. On the other hand, the nodes X1 and X3 are parent nodes that are not affected by any node, and the probability table at this time is like the prior probability table shown in the figure.

  In the case of what is shown in FIG. 2, from the structure of the graph, the joint probabilities of all random variables are P (X1, X2, X3, X4) = P (X1) P (X2 | X1, X3) P (X3) P (X4 | X3). Inference using a Bayesian network is performed by obtaining peripheral posterior probabilities by marginalizing this connection probability. How to use such a Bayesian network in the driver model of the driving support device of the present invention will be described later.

  In the driving support device according to the present invention, the lane change specific advance behavior is learned as a driver model as described above by automatically recognizing the lane change. Then, a lane change operation is predicted in advance using this driver model. Further, the state of other vehicles around the vehicle is sensed by an in-vehicle camera or the like, and the other vehicles existing in the proximity position are grasped. Moreover, the camera which image | photographs diagonally right and left back is provided, and the image of the part used as a blind spot of a driver is provided. As a result, even for vehicles that are in a position that is not dangerous in normal driving, such as driving in the lane next to your vehicle, you can change lanes only when you try to change lanes. It is possible to perform a warning in advance in consideration of the danger of a collision in advance.

(2) Details of Embodiment FIG. 3 shows a block configuration of a driving support device mounted on a vehicle. In FIG. 2, 100 is an ECU (electronic control unit), 200 is a host vehicle information acquisition unit, 201 is a steering wheel steering angle sensor, 202 is an accelerator pedal position sensor, 203 is a brake pedal position sensor, 204 is a speedometer, and 205 is acceleration. Sensor, 206, electric operation status acquisition unit, 207, blinker, 208, light, 210, driver information acquisition unit, 211, face camera, 212, facial expression recognition unit, 213, gaze recognition unit, 214, face direction recognition unit, 215 Is a skin potential sensor, 216 is a microphone, 220 is a vehicle surrounding environment information acquisition unit, 221 is a front monitoring camera, 222 is a rear monitoring camera, 223 is a side monitoring camera, 224 is an inter-vehicle distance radar, 225 is a corner sensor, 230 is Car navigation system, 231 is GPS unit, 232 is road information unit, 233 Route guidance unit, 234 is a map database, 240 is an external communication unit, 241 is a mobile phone communication unit, 242 is a VICS communication unit, 243 is a wireless LAN communication unit, 244 is an inter-vehicle communication unit, 245 is an inter-vehicle communication unit, 246 Is a data broadcasting unit, 300 is a driver model database unit, 301 is normal driver model data, 302 is lane change driver model data, 303 is lane change preceding behavior driver model data, 310 is a driver model learning data recording unit, and 311 is the own vehicle Information, 312 is vehicle surrounding environment information, 313 is position information, 400 is an information presentation unit, 401 is a display unit, 402 is a voice warning unit, 403 is a hazard lamp, 500 is a pattern matching data storage unit, and 501 is a normal face Facial expression template data, 502 is pedestrian template data, 5 3 shows a bicycle template data, respectively.

  Note that the configuration of the driving support device described with reference to FIG. 3 is not all necessary, but describes each part and device that can be used to create a driver model and perform driving support in the present embodiment. Therefore, the apparatus can be configured by appropriately selecting according to the driving support function to be adopted, and other devices and apparatuses having similar functions can be additionally used.

  The driving support device includes an ECU (electronic control unit) 100, a host vehicle information acquisition unit 200, a driver information acquisition unit 210, a vehicle surrounding environment information acquisition unit 220, a car navigation system 230, an external communication unit 240, a driver model database unit 300, A driver model learning data recording unit 310, an information presentation unit 400, and a pattern matching data storage unit 500 are provided.

  The ECU 100 is configured by a general-purpose computer system that includes various parts such as a CPU, a ROM, a RAM, and an interface. The ECU 100 cooperates and operates with each component connected to the illustrated ECU 100. In particular, in the present invention, the ECU 100 mainly monitors driver driving behavior based on the acquired information of the own vehicle information acquiring unit 200, and monitors changes in biological information that is physiological state data based on the acquired information of the driver information acquiring unit 210. The driver model learned and registered in the driver model database unit 300 is compared with the driving operation information obtained from the own vehicle information acquisition unit 200, the information indicating the lane change indication based on the driver model database unit 300, the information presentation Appropriate information and warnings for signs of lane changes are presented by the section 400. In addition, the ECU 100 is configured to execute data processing necessary for creating and outputting a driver model.

  The own vehicle information acquisition unit 200 includes a steering wheel steering angle sensor 201, an accelerator pedal position sensor 202, a brake pedal position sensor 203, a speedometer 204, an acceleration sensor 205, an electric operation state acquisition unit 206, a winker 207, a light 208, and other sensors. It has.

  FIG. 4 is a diagram showing host vehicle information acquired by the host vehicle information acquisition unit 200 in the driving support apparatus according to the embodiment of the present invention. The steering wheel angle sensor 201 is the steering wheel operation amount, the accelerator pedal position sensor 202 is the accelerator pedal operation amount, the brake pedal position sensor 203 is the brake pedal operation amount, the speedometer 204 is the vehicle speed, and the acceleration sensor 205 is the vehicle speed. Detect acceleration. In the electric operation status acquisition unit 206, the winker 207 detects the winker operation status, and the light 208 detects the light operation status.

  The driver information acquisition unit 210 includes a face camera 211, a facial expression recognition unit 212, a gaze recognition unit 213, a face orientation recognition unit 214, a skin potential sensor 215, and a microphone 216.

  FIG. 5 is a diagram showing driver information acquired by the driver information acquisition unit 210 in the driving support apparatus according to the embodiment of the present invention. The face camera 211 is an imaging device of a driver who drives the vehicle, and performs image recognition based on imaging data acquired by the imaging device. The facial expression recognition unit 212 performs image recognition of the driver's facial expression, the gaze recognition unit 213 performs image recognition of the driver's gaze, and the face direction recognition unit 214 performs image recognition of the driver's face direction. Further, it is determined whether or not the driver is talking based on the audio data acquired by the microphone 216. Skin potential sensor 215 is a sensor that is provided at the steering of the vehicle driver's seat or the like and detects the skin potential of the driver. According to the skin potential acquired by this skin potential sensor, there is a possibility that the tension state of the driver before the lane change can be detected.

  The vehicle surrounding environment information acquisition unit 220 includes a front monitoring camera 221, a rear monitoring camera 222, a side monitoring camera 223, an inter-vehicle distance radar 224, and a corner sensor 225. The ECU 100 grasps the situation around the vehicle using these cameras, radar, and sensors. In particular, by combining the position information by the GPS unit 231, the map information of the map database 234, and the lane information analyzed based on the images captured by the front monitoring camera 221 and the rear monitoring camera 222, the ECU 100 determines whether the current vehicle is Information such as which lane is running can be acquired.

  FIG. 6 is a diagram showing vehicle surrounding environment information acquired by the vehicle surrounding environment information acquiring unit 220 in the driving support apparatus according to the embodiment of the present invention. The front monitoring camera 221 captures and acquires a front image, the rear monitoring camera 222 captures and acquires a rear image, and the side monitoring camera 223 captures and acquires a side image. The inter-vehicle distance radar 224 measures the inter-vehicle distance from a vehicle traveling ahead, and the corner sensor 225 measures the distance from an obstacle existing around the vehicle.

  The car navigation system 230 includes a GPS unit 231 that detects current position information of a vehicle and a map database 234 that stores map information corresponding to the detected current position information. The route guide unit 233 derives a route from the current position of the vehicle according to the set destination and the like.

  FIG. 7 is a diagram showing car navigation system information acquired by the car navigation system 230 in the driving support apparatus according to the embodiment of the present invention. In the car navigation system 230, information on “latitude”, “longitude”, and “travel lane” is acquired as position information. Further, the car navigation system 230 acquires data classified as “road type”, “road shape”, “road width”, and “traffic regulation” as road information. Each small item data shown in the figure is acquired under each middle item.

  The external communication unit 240 includes a mobile phone communication unit 241, a VICS communication unit 242, a wireless LAN communication unit 243, an inter-vehicle communication unit 244, an inter-vehicle communication unit 245, and a data broadcasting unit 246. In each of these communication units, the vehicle-to-vehicle communication unit 244 performs communication with surrounding vehicles that run on the host vehicle, and the road-to-vehicle communication unit 245 performs communication between the infrastructure system such as ITS and the vehicle. It is. The road-to-vehicle communication unit 245 obtains information related to emergency vehicles traveling around.

FIG. 8 is a diagram showing information acquired by the external communication unit 240 in the driving support apparatus according to the embodiment of the present invention. Information acquired by the external communication unit 240 and classified into the middle items includes “traffic information”, “weather information”, “road surface information”, “signal information”, “peripheral vehicle information”, and other information. Each small item data shown in the figure is acquired under each middle item.

  The driver model database unit 300 includes lane change driver model data 302 that stores a driver model at the time of lane change, lane change preceding action driver model data 303 that stores a driver model of the preceding action, and drivers at other times And normal driver model data 301 for recording the model.

  FIG. 9 is a diagram showing a data structure stored in the driver model database unit 300 in the driving support apparatus according to the embodiment of the present invention. The driver model database unit 300 has the same data structure for the driver model at the time of lane change, the driver model at the time of lane change preceding action, and the driver model at times other than at the time of lane change, and FIG. 9 shows the common structure. Show. “Continuous node data” is data relating to the driver model itself, and has parameters “Discrete”, “Average”, and “Dispersion” in a Gaussian distribution.

  The information presentation unit 400 notifies and warns the driver based on the predicted driving behavior, and has a display unit 401, a voice warning unit 402, a hazard lamp 403, and other configurations.

Further, the pattern matching data storage unit 500 includes normal facial expression template data 501,
It has three pattern matching template data: pedestrian template data 502 and bicycle template data 503. These pattern matching data are for handling data that is difficult to quantify. FIG. 10 is a diagram showing template data stored in the pattern matching data storage unit 500 in the driving support apparatus according to the embodiment of the present invention. In FIG. 10, for example, “normal facial expression template data” is image data of a facial expression of a driver at normal times. Based on such basic image data, it is possible to derive whether or not the facial expression recognized by the facial expression recognition unit 212 is normal. By using template data, various situations are converted into discrete data and substituted into a Bayesian network algorithm.

  The driver model learning data recording unit 310 stores own vehicle information 311, vehicle surrounding environment information 312, position information 313, and the like. Information in the driver model learning data recording unit 310 is used for creating a driver model and predicting a lane change. The information in the driver model learning data recording unit 310 is recorded as a log for about 10 minutes from the latest one, for example.

  The “lane change driver model data” and the “lane change preceding behavior driver model data” must be created in a form including the preceding behavior traced back to the past from the point where the lane change has been completed. For this reason, when creating the “lane change driver model data”, as described above, the “lane change driver model data” is based on the data from the log data for about 10 minutes to the point where the lane change starts and ends. And “lane change preceding behavior driver model data” is created from data from a certain period before the starting point to the starting point.

For creating each driver model (more specifically, for creating a probability table to be described later), Arthur Demster, Nan Laird, and
Donald Rubin. “Maximum likelihood from incomplete
e data via the EM algorithm ". Journal of
th
e Royal Statistical Society, Series B, 39 (1): 1-38, 1977. The EM algorithm shown in FIG.

  FIG. 11 is a diagram showing a data structure stored in the driver model learning data recording unit 310 in the driving support apparatus according to the embodiment of the present invention. As shown in FIG. 11, the data in the driver model learning data recording unit 310 includes data relating to driving operations (accelerator, brake, steering wheel operation amounts) during driving of the driver, and data relating to the surrounding environment at that time (roads) Type, traffic volume, driver facial expression, etc.), a driver model is created based on these data, and the lane change based on the latest data in the driver model learning data recording unit 310 is created. A prediction is made.

  Next, a specific configuration of the driver model used in the driving support device of the present invention will be described. FIG. 12 is a diagram showing a network structure of a driver model used in the driving support apparatus according to the embodiment of the present invention. The driver model used in the driving support device of the present invention is composed of (a) the structure of the Bayesian network shown in FIG. 12, (b) a conditional probability table, a prior probability table, and (c) a Gaussian distribution parameter. . Of these three, (b) and (c) are always replaced by the latest by learning. That is, when learning the driver model, the conditional probability table and the parameters of the Gaussian distribution are updated. On the other hand, the structure of the Bayesian network is defined at the time of the initial design, and the same one is continuously used thereafter.

  In the structure of the Bayesian network in FIG. 12, □ indicates a discrete node, ◯ indicates a continuous node, and the number in the discrete node indicates the number of states that each node takes. For example, “facial expression Fc (2)” takes two states, Fc = 0 (normal) and Fc = 1 (abnormal). In the driving support device of the present invention, since it has a configuration having a plurality of such discrete nodes, it becomes possible to perform driving operation prediction more accurately, and to provide efficient and accurate driving support. It becomes possible.

  In the network structure of FIG. 12, the facial expression Fc (2) shown in the upper stage, the line of sight Gz (4),... Traffic volume Td (3), and the vehicle speed Sp (8 shown in the lower stage. ), Inter-vehicle distance Fd (8),... Each lane change Lc (3) node is a parent node, and these nodes affect the node of the physiological state Ph (2) and the node of the driving behavior variable DB. I think. In the network structure, it is considered that the node of the physiological state Ph (2) influences the node of the driving behavior variable DB. That is, the physiological state data is a node affected by the driving situation data, and the driving operation data has a network structure that is a node affected by the driving situation data and the physiological state data. The relationship is defined by a Bayesian network. The network structure of the physiological state data, the driving state data, and the driving operation data has been confirmed experimentally, and the reason for setting in this way is as follows. Although the physiological state data is a node that is influenced by the driving situation data, the physiological state of the driver may change even if the driver's physiological state may change depending on the situation of the surrounding vehicle. The change of the vehicle does not affect the situation of the surrounding vehicle. In addition, the driving operation data is a node that is influenced by the driving situation data and physiological state data, but even if the situation and physiological state of the surrounding vehicle that is running may affect the driving operation, This is because the driving operation does not affect the situation and physiological state of the surrounding vehicle. Therefore, if this network structure is changed to a different structure, the accuracy of the prediction is remarkably lowered, and the network structure of this embodiment is optimal for the driving support apparatus. This is particularly suitable when determined by a Bayesian network.

  Here, the facial expression Fc (2) shown in the upper part of the network structure, the line of sight Gz (4),... Traffic volume Td (3), and the vehicle speed Sp (8) shown in the lower part. Each node of the inter-vehicle distance Fd (8),... Lane change Lc (3) corresponds to the driver model learning data acquired by the driving support device as shown in FIG. This corresponds to the continuous data row of the driver model learning data in FIG. Continuous data is held in the form of “discrete”, “average”, and “variance” of a Gaussian distribution, and is replaced with a predetermined probability by a method described later.

  The physiological state Ph (2) node does not correspond to the driver model learning data shown in FIG. 11, but is an independent node. This physiological state Ph (2) node works as a kind of tolerance, and it is known that the driving support apparatus of the present invention may not operate effectively when this node does not exist. Even if the state of each node of the facial expression Fc (2), line of sight Gz (4),..., Lane change Lc (3) (state of driving situation) is uniquely determined, This is because the driving behavior variable DB (the state of the driving operation amount) is not uniquely determined. In the driving support device of the present invention, by introducing the physiological state Ph (2) node, it becomes possible to handle causality including uncertain elements. The physiological state Ph (2) node is very important when the driving support apparatus of the present invention is executed by a computer program.

  FIG. 13 is a diagram showing a state taken by the physiological state Ph (2) in the Bayesian network structure. As shown in FIG. 13, in the present embodiment, the physiological state Ph (2) assumes two states of Ph = 0 (normal) and Ph = 1 (abnormal). In the driving support device, the physiological state Ph can be set to an appropriate natural number of 2 or more.

  Next, in the driving support apparatus of the present invention, a method for predicting a lane change by performing probability calculation using a driver model based on a Bayesian network will be described. Since the network structure of the driver model shown in FIG. 12 has a large number of nodes and is complicated, it will be described using a simple example.

  FIG. 14 is a diagram illustrating a simplified driver model network structure. As a parent node as shown in the figure, a Bayesian network structure having two nodes of facial expression Fc (2) and lane change Lc (3) will be examined below.

  At this time, the facial expression Fc (2) node and the lane change Lc (3) node have a prior probability table as shown, and the physiological state Ph (2) node has a conditional probability table as shown. It shall be. These prior probability table and conditional probability table are determined by accumulation of learning of the driver model, and the above-described EM algorithm is used to obtain the probabilities in these tables.

  The lane change driver model data 302 in the previous driver model database unit 300 is a table with Lc = 2, the lane change preceding behavior driver model data 303 is a table with Lc = 1, and the normal driver model data 301 is a table with Lc = 0. Each can be thought of as holding.

A method for obtaining the prior probability table will be described by taking the prior probability calculation at the lane change (Lc) node as an example. When N _k is the occurrence time of the event (Lc = k) in the learning data, the prior probability of the lane change (Lc) can be calculated as the following equation (1).

ただし、

However,

これと同様に、顔の表情（Ｆｃ）の事前確率テーブルを求めることができる。

Similarly, a prior probability table of facial expression (Fc) can be obtained.

The calculation method of the conditional probability table will be described by taking the biological state Ph node as an example.
When N _k is the occurrence time of the event (Ph = k, Pa (Ph) = j) in the learning data, the prior probability of the biological state Ph can be calculated as the following equation (3).

ただし、

However,

また、運動行動変数ＤＢに係る確率は、ＥＭアルゴリズムで求められた離散、平均、分散によって、

In addition, the probability related to the motor behavior variable DB is determined by the discrete, average, and variance obtained by the EM algorithm.

ただし、Ｎは正規分布関数、μは平均、Σは分散を表している。

However, N represents a normal distribution function, μ represents an average, and Σ represents variance.

  The joint probability of the Bayesian network shown in FIG.

そこで、次にこの式を用いて、車線変更（Ｌｃ）の所定条件下における確率を求めることを考える。Ｌｃ以外の変数の状態が分かっており、そのときの、例えばＬｃ＝２である確率がどれだけあるかを求める。図１５は、Ｐｈ＝ｐｈ、ＤＢ＝ｄｂ、Ｆｃ＝ｆｃであるときの、Ｌｃ＝ｌｃの確率の計算の計算を示す図である。なお、ネットワークの同時確率式から、不要なＬｃを消去する作業を周辺化と称する。図に示すように、点線で囲まれている部分は確率テーブルの値を代入することができ、一点鎖線で囲まれている部分はガウス分布から計算することができる。このように、本発明の運転支援装置では、ドライバモデルによって、他のパラメータが分かっているときに、例えばＬｃ＝２である確率、すな
わち、車両が車線変更中である確率を求めることができるのである。

Therefore, it is next considered to obtain the probability under a predetermined condition of lane change (Lc) using this equation. The state of a variable other than Lc is known, and the probability of, for example, Lc = 2 at that time is obtained. FIG. 15 is a diagram illustrating the calculation of the probability of Lc = lc when Ph = ph, DB = db, and Fc = fc. The operation of deleting unnecessary Lc from the simultaneous probability formula of the network is referred to as peripheralization. As shown in the figure, the value enclosed in the dotted line can be substituted with the value in the probability table, and the part surrounded by the alternate long and short dash line can be calculated from the Gaussian distribution. Thus, in the driving support apparatus of the present invention, when other parameters are known from the driver model, for example, the probability that Lc = 2, that is, the probability that the vehicle is changing lanes can be obtained. is there.

In addition, although the method for deriving the predetermined probability by the Bayesian network has been described above in a simplified manner, in the driving support device of the present invention, in more detail,
○ Jensen, F.M. V. "Bayesian networks and decision graphics", Springer, 2001.
O David C.I. MacKay, “Information Theory, Information and Learning Algorithms”, Cambridge University Press, 2003.
It shall be incorporated with reference to the method described in the above.

  Next, “lane change driver model data” that records the lane change pattern, “lane change advance behavior driver model data” that records the preceding behavior pattern before the lane change, and lane change that includes the preceding behavior A description will be given of an algorithm for learning by assigning three patterns of “normal driver model data” in which patterns at other times are recorded.

  The information in the driver model learning data recording unit 310 is recorded, for example, as a log for about 10 minutes from the latest one. As described above, the preceding behavior of the lane change is traced back to the past when the lane change is completed. The data of etc. is cut out.

  FIG. 16 is a diagram showing a flowchart of the driver model learning process in the driving support apparatus according to the embodiment of the present invention. The flowchart of FIG. 16 is periodically executed by the ECU 100 at a sufficiently short time interval. Further, in execution, it is possible to perform other processing by exiting the loop as necessary.

  Note that points such as T and S in the flowchart of FIG. 16 conform to FIG. FIG. 20 is a diagram showing lane change timing and the like in time series.

  In FIG. 16, when the driver model learning process is started in step S100, the process proceeds to step S101, where the own vehicle information acquisition unit 200, the vehicle surrounding environment information acquisition unit 220, the car navigation system 230, the external communication unit 240, and the like. The vehicle information / peripheral information is acquired by the above.

  Next, in step S102, the driver information acquisition unit 210 acquires driver information. In step S103, driver model learning data is acquired from the driver model learning data recording unit 31.

  In step S104, a subroutine for estimating the end point T of the lane change action is executed. This subroutine will be described later.

  In step S105, it is determined whether or not the end point T exists. If YES is determined in the step S105, the process proceeds to a step S106, and if NO is determined in the step S105, the process proceeds to a step S111.

  In step S111, the driver model database unit 300 learns the information of the section A from the beginning of the learning data to a certain time U as a normal driving behavior model.

In step S106, a subroutine for estimating the start point S of the lane change operation is executed, and then the process proceeds to step S107. This subroutine will be described later. In step S107, a subroutine for estimating the start point R of the preceding operation of the lane change operation is executed. This subroutine will be described later.

  In step S108, the driver model database unit 300 learns the information of the section C from the start point R of the lane change preceding action to the start point S of the lane change action as a driver model of the lane change preceding action.

  In step S109, the driver model database unit 300 learns information on the section D from the start point S of the lane change action to the end point T of the lane change action as a driver model of the lane change action.

In step S110, the driver model database unit 300 learns the information of the section B from the beginning of the learning data to the start point R of the preceding action for lane change as a driver model for normal driving action.
In step S112, the process ends.

  Next, a subroutine for estimating the end point T of the lane change action in step S104 will be described. FIG. 17 is a flowchart of a subroutine for estimating the end point T of the lane change action in the driving support apparatus of the present invention. Θ appearing in the flowchart of FIG. 17 is shown in FIG. That is, θ is an angle formed by the direction of the host vehicle and the white line at the boundary between the lanes.

  In step S200, when a subroutine for estimating the end point T of the lane change action is started, the process proceeds to step S201, and the vehicle front image is acquired by the front monitoring camera 221.

  In subsequent step S202, the lane is recognized, and the vehicle position in the lane and the angle θ between the lane and the host vehicle are calculated. The definition of θ is as shown in FIG.

  In step S203, it is determined whether | θ |> threshold. When the determination result of step D203 is YES, the process proceeds to step S204, and when the determination result of step D203 is NO, the process proceeds to step S213. In step S213, “T does not exist” is obtained as a return value, and the process returns.

  In step S204, it is determined whether or not the host vehicle crosses the lane. When the determination result of step D204 is YES, the process proceeds to step S205, and when the determination result of step D204 is NO, the process proceeds to step S213, and the process returns.

  In step S <b> 205, the vehicle front image is acquired by the front monitoring camera 221. In step S206, the lane is recognized, and an angle θ formed between the lane and the host vehicle is calculated.

  In step S207, it is determined whether or not | θ | <threshold. When the determination result of step S207 is YES, the process proceeds to step S208, and when the determination result of step S207 is NO, the process returns to step S205 again.

  In step S208, the end point T of the lane change operation is temporarily marked. In step S209, the information of the intersection that can be turned right and left at a close distance and the road information of the right and left turn destination are acquired from the map information.

In step S210, it is determined whether the vehicle has traveled straight through the intersection. When the determination result of step S210 is YES, the process proceeds to step S212, and when the determination result of step S210 is NO, the process proceeds to step S214.

  In step S211, since the vehicle has not changed lanes, the assumed end point T of the lane change operation is discarded, and in step S214, “T does not exist” is returned as a return value, and the process returns. .

  In step S212, since the vehicle has changed lanes, the assumed end point T of the lane change operation is adopted. Then, the following step S215 obtains the position of T as a return value and returns.

  Next, a subroutine for estimating the start point S of the lane change operation in step SS106 will be described. FIG. 18 is a flowchart showing a subroutine for estimating the starting point S of the lane change operation in the driving support apparatus of the present invention. FIG. 22 is a diagram illustrating a driver's action for estimating the start point S.

  If the processing of the subroutine for estimating the start point S of the lane change operation is started in step S300, the process proceeds to step S301, and the vehicle peripheral ring information, own vehicle information, Get driver information.

  In step S302, it is determined whether the turn signal has been operated. When the determination result of step S302 is YES, the process proceeds to step S308, and when the determination result of step S302 is NO, the process proceeds to step S303. In step S308, the point where the turn signal is operated is recorded as the lane change search start position, and the process proceeds to step S306.

  In step S303, it is determined whether the driver has seen the side mirror. When the determination result of step S303 is YES, the process proceeds to step S307, and when the determination result of step S303 is NO, the process proceeds to step S304. In step S307, the point where the side mirror is viewed is recorded as the search start position for lane change, and then the process proceeds to step S305.

  In step S304, a predetermined time before is recorded as a search start position for lane change.

  In step S305, an angle θ formed by the lane from the recorded search start position to the present and the host vehicle is calculated. In step S306, the point where the angle θ formed becomes a certain value or more is recorded as the start position S of the lane change action. In step S309, the position S is obtained as a return value, and the process returns to the main routine.

  Next, a subroutine for estimating the start point R of the preceding operation of the lane change operation in step S107 will be described. FIG. 19 is a flowchart showing a subroutine for estimating the start point R of the preceding operation of the lane change operation in the driving support apparatus of the present invention.

When the subroutine for estimating the start point R of the preceding operation of the lane change operation is started in step S400, the process proceeds to step S401, and a predetermined time before the start position of the lane change is recorded as the start position R of the preceding action. In step S402, the position R is obtained as a return value, and the process returns.

Next, a description will be given of predicting the driving operation behavior related to the lane change based on the learned driver model and warning the driver or the like as necessary. FIG. 23 is a diagram showing a flowchart of the lane change support process in the driving support apparatus according to the embodiment of the present invention. 23, the probability that the current driving action is a lane change action is Pc, the probability that the current driving action is a lane change preceding action is Pp, and the probability that the current driving action is a normal driving action is Pn. It shall be shown. The flowchart of FIG. 23 is periodically executed by the ECU 100 at a sufficiently short time interval. Further, in execution, it is possible to perform other processing by exiting the loop as necessary.

  In the flowchart of FIG. 23, when the flow of the lane change support process is started in step S500, the process proceeds to step S501, and the probabilities Pn, Pc, and Pp are calculated in the manner described above by the Bayesian network.

  In step S502, it is determined whether Pp> Pn, Pp> Pc, and Pp> threshold. When the determination result in step S502 is YES, the process proceeds to step S503, and when the determination result is NO, the process returns to step S501.

In step S503, road information (including traffic sign information) is taken in from the car navigation system 230. In continuing step S504, it is determined whether the lane change is permitted.
When the determination result at step S504 is YES, the process proceeds to step S505, and when the determination result at step S504 is NO, the process proceeds to step S513. In step S513, the information presenting unit 400 gives a warning about lane departure or lane departure possibility. In place of such a warning operation, a driving operation assist operation that does not cause a lane departure can be performed. Alternatively, both warning notification and driving operation assistance may be executed.

  In step S505, lane change support information is presented. FIG. 24 is a diagram showing an example of a support information presentation method in the driving support apparatus according to the embodiment of the present invention. FIG. 24 shows a state when two subsequent vehicles of the own vehicle are confirmed from the driver's seat of the own vehicle. In this support information presentation method, an imaging device (not shown) is provided in the vicinity of the side mirror on the driver's seat side, and an image captured by the imaging device is presented to the driver on the imaging device captured image monitor. The succeeding vehicle 2 in the blind spot is displayed on the screen.

  Next, in step S506, probabilities Pn, Pc, and Pp are calculated as described above using a Bayesian network. In the next step S507, the host vehicle information acquisition unit 200 acquires host vehicle information.

  In step S508, it is determined whether or not the winker 207 is operating. When the result of determination in step S508 is YES, the process proceeds to step S509, and when the result of determination in step S508 is NO, the process proceeds to step S514. In step S514, attention is paid to the driver so that the information presenting unit 400 takes out the blinker. In addition, you may perform the driving operation assistance which operates a blinker automatically instead of performing such attention. Alternatively, caution notification and driving operation assistance may be executed.

In step S509, the vehicle surrounding environment information acquisition unit 220 acquires vehicle surrounding environment information. In step S510, it is determined whether or not there is a risk of collision at the lane change destination.
When the determination result of step S510 is YES, the process proceeds to step S515, and when the determination result of step S510 is NO, the process proceeds to step S511.

  In step S515, which proceeds when there is a danger of a collision, the information presenting unit 400 presents that there is an obstacle to the driver.

  In a succeeding step S516, it is determined whether or not Pc> Pp is satisfied. When the determination result of step S516 is YES, the process proceeds to step S517, and when the determination result of step S516 is NO, the process proceeds to step S511. In step S517, since it is predicted that the driver will make a lane change more intensely, the driver is warned of a risk of collision stronger than step S515. In addition, it can replace with such a warning operation | movement and can also be comprised so that the driving operation assistance operation | movement which avoids a collision may be performed. Alternatively, both warning notification and driving operation assistance may be executed.

  In step S511, it is determined whether Pn> Pc and Pn> Pp. When the determination result of step S511 is YES, the process proceeds to step S512, and when the determination result of step S511 is NO, the process returns to step S506.

  In step S512, the lane change support is terminated, and in the subsequent step S518, the flow process is terminated.

  According to the configuration of the present invention as described above, driving situation data (facial expression Fc (2), line of sight Gz (4) shown in the upper part of the network structure in FIG. 12, line of sight Tz (3) ), Vehicle speed Sp (8), inter-vehicle distance Fd (8),... Lane change Lc (3) nodes) and driving operation data (driving behavior variable DB) shown in the lower row, The causal relationship between the third data not related to (nodes of the physiological state Ph (2)) is determined by the Bayesian network, the driving operation is predicted based on this, and the driver is warned and assisted. Efficient and accurate driving support can be performed.

  Next, an algorithm for digitizing each node of “facial expression Fc (2)” in FIG. 13 will be described. The node of “Facial expression Fc (2)” cannot directly digitize the acquired data as it is, so by performing pattern matching, Fc = 0 (normal), Fc = 1 (abnormal) One of the two states is determined. For this purpose, pattern matching is performed based on a normal face image stored in the normal facial expression template data 501 in advance to determine whether Fc = 0 (normal) or Fc = 1 (abnormal).

  Next, an algorithm for digitizing each node of “obstruction Ob (2) by pedestrian / bicycle / stopped vehicle” in FIG. 13 will be described. In the node of “obstruction Ob (2) by pedestrian / bicycle / stopped vehicle”, the data acquired as it is cannot be directly converted into discrete numerical values, so this is performed by performing pattern matching. FIG. 25 is a diagram showing a flowchart of pattern matching processing in the driving support apparatus according to the embodiment of the present invention. This process is executed as a subroutine when the value of “obstruction Ob (2) by pedestrian / bicycle / stopped vehicle” is required in the main routine process in the ECU 100. The purpose of this process is to obtain a value of Ob = 0 (no interference) or Ob = 1 (with interference).

  In FIG. 25, when the flow of the pattern matching process is started in step S600, the process proceeds to step S601, where the vehicle surrounding environment information acquisition unit 220 acquires vehicle surrounding information.

  In the subsequent step S602, it is determined whether or not the distance between the vehicles <the threshold and | the relative speed−the vehicle speed | <the threshold. When the result of determination in step S602 is YES, the process proceeds to step S607, and when the result of determination in step S602 is NO, the process proceeds to step S603. In step S607, after obtaining Ob = 1 (with interference) as a return value, the process returns.

  In step S603, a front image is acquired. In step S604, pedestrian template data 502 and bicycle template data 503 (pattern matching template data) in the pattern matching data storage unit 500 are acquired. In step S605, pattern matching processing is performed. Conventionally known processing can be used for this pattern matching processing.

  In step S606, it is determined whether there is matching data. When the determination result at step S606 is YES, the process proceeds to step S609, and when the determination result at step S606 is NO, the process proceeds to step S608. In step S608, after obtaining Ob = 0 (no interference) as a return value, the process returns. In step S609, after obtaining Ob = 1 (with interference) as a return value, the process returns.

  In addition, about the structure which concerns on the driving assistance device of this embodiment, although mounting on vehicles, such as a motor vehicle, a hybrid vehicle, and an electric vehicle, is assumed for the time being, mounting in another moving means is also considered.

  Further, the “driving operation data acquisition means” in the claims includes driving status data (facial expression Fc (2), line of sight Gz (4) shown in the upper part of the network structure in FIG. 12, traffic Driver information acquisition unit 210 for acquiring each node of the amount Td (3), vehicle speed Sp (8), inter-vehicle distance Fd (8),... The vehicle surrounding environment information acquisition unit 220, the car navigation system 230, the external communication unit 240, the pattern matching data storage unit 500, and the ECU (electronic control unit) 100 that treats the acquired data are all included.

  The “driving condition data acquisition means” in the claims includes an own vehicle information acquisition unit 200 that acquires driving operation data (driving behavior variable DB), an ECU (electronic control unit) 100 that processes the acquired data, and the like. The whole thing.

Further, the “table updating means” in the claims is stored in the driver model database unit 300 such as a conditional probability table, a prior probability table, and a Gaussian distribution parameter in each node of the structure of the Bayesian network shown in FIG. A configuration of an ECU (electronic control unit) 100 or the like that executes a table update process.
Further, the “driving operation prediction means” in the claims is an ECU that executes arithmetic processing for driving operation prediction based on a prior probability table stored in the driver model database unit 300, parameters of a Gaussian distribution, and the like. (Electronic control device) 100 and the like.

  Although various embodiments have been described above, embodiments that are configured by arbitrarily combining the constituent elements of the respective embodiments also fall within the scope of the present invention.

It is a figure which shows the concept of the driving assistance apparatus which concerns on embodiment of this invention. It is a figure which shows the concept of the Bayesian network used with the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the block configuration of the driving assistance apparatus which concerns on embodiment of this invention. It is a figure which shows the own vehicle information acquired by the own vehicle information acquisition part 200 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the driver information acquired in the driver information acquisition part 210 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the vehicle surrounding environment information acquired in the vehicle surrounding environment information acquisition part 220 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the car navigation system information acquired with the car navigation system 230 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the information acquired in the external communication part 240 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the information memorize | stored in the driver model database part 300 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the template data memorize | stored in the data storage part 500 for pattern matching in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the data structure memorize | stored in the driver model learning data recording part 310 in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the network structure of the driver model used with the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the state which physiological state Ph (2) takes in a Bayesian network structure. It is a figure which shows the network structure of the simplified driver model. It is a figure which shows the calculation of the calculation of the probability of Lc = lc when Ph = ph, DB = db, and Fc = fc. It is a figure which shows the flowchart of the driver model learning process in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the flowchart of the subroutine which estimates the end point T of the lane change action in the driving assistance device of this invention. It is a figure which shows the flowchart of the subroutine which estimates the starting point S of the lane change operation | movement in the driving assistance device of this invention. It is a figure which shows the flowchart of the subroutine which estimates the starting point R of the preceding operation | movement of the lane change operation | movement in the driving assistance device of this invention. It is a figure which shows the timing of a lane change, etc. in time series. It is a figure which defines angle (theta). It is a figure which shows the action of the driver for estimating the starting point. It is a figure which shows the flowchart of the lane change assistance process in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows an example of the assistance information presentation method in the driving assistance device which concerns on embodiment of this invention. It is a figure which shows the flowchart of the pattern matching process in the driving assistance device which concerns on embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... ECU (electronic control apparatus), 200 ... Own vehicle information acquisition part, 201 ... Steering wheel steering angle sensor, 202 ... Accelerator pedal position sensor, 203 ... Brake pedal position sensor, 204 ..Speedometer, 205 ... acceleration sensor, 206 ... electric operation status acquisition unit, 207 ... winker, 208 ... light, 210 ... driver information acquisition unit, 211 ... face camera, 212 ... facial expression recognition unit, 213 ... gaze recognition unit, 214 ... face orientation recognition unit, 215 ... skin potential sensor, 216 ... microphone, 220 ... vehicle surrounding environment information acquisition unit, 221: Front monitoring camera, 222: Rear monitoring camera, 223 ... Side monitoring camera, 224 ... Inter-vehicle distance radar, 225 ... Corner sensor, 2 DESCRIPTION OF SYMBOLS 0 ... Car navigation system, 231 ... GPS part, 232 ... Road information part, 233 ... Route guidance part, 234 ... Map database, 240 ... External communication part, 241 ... Mobile phone communication unit, 242 ... VICS communication unit, 243 ... Wireless LAN communication unit, 244 ... Vehicle-to-vehicle communication unit, 245 ... Road-to-vehicle communication unit, 246 ... Data broadcasting unit, 300 .. Driver model database unit 301... Normal driver model data 302. Lane change driver model data 303. Lane change preceding action driver model data 310. Driver model learning data recording unit 311 ... Vehicle information, 312 ... Vehicle environment information, 313 ... Position information, 400 ... Information presentation unit, 401 ... Display 402 ... Voice warning unit, 403 ... Hazard lamp, 500 ... Pattern matching data storage unit, 501 ... Normal facial expression template data, 502 ... Pedestrian template data, 503 ...・ Bicycle template data

Claims (6)

Physiological state data means for acquiring data relating to the physiological state of the driver;
Driving operation data acquisition means for acquiring data relating to driving operation;
Driving status data acquisition means for acquiring data relating to the driving status;
A table updating means for updating a table that probabilistically determines the causal relationship between physiological state data, driving situation data, and driving operation data ;
Driving operation prediction means for predicting a driving operation from the table updated by the table update means and the current driving situation data ,
The network structure in the causal relationship is a network in which the physiological state data is a node affected by the driving situation data, and the driving operation data is a node affected by the driving situation data and the physiological state data. A driving support device characterized by a structure . The driving support apparatus according to claim 1 , wherein the causal relationship of the network structure is determined by a Bayesian network. The driving support device according to claim 1 , further comprising warning notifying means for notifying a driver of a warning or the like based on the prediction of the driving operation predicting means. The driving support device according to any one of claims 1 to 3 , further comprising driving operation assisting means for assisting the driver in driving operation based on the prediction of the driving operation prediction means. The driving support apparatus according to any one of claims 1 to 4 , wherein the driving status data acquisition unit acquires a plurality of driving status data for acquiring data related to the driving status. A physiological state data acquisition step of acquiring data relating to the physiological state of the driver ;
A driving operation data acquisition process for acquiring data relating to the driving operation ;
An operation status data acquisition step for acquiring data relating to the operation status ;
Updating a table in which the causal relationship between physiological state data , driving operation data, and driving situation data is determined stochastically by using a Bayesian network;
A process of predicting a driving operation from the updated updated table and current driving situation data ,
In the network structure in the Bayesian network, the physiological state data is a node affected by the driving situation data, and the driving operation data is a node affected by the driving situation data and the physiological state data. A driving support method characterized by being defined .

Images