System for detecting abnormal driving behavior
The system uses AR-HMM to analyze driving conditions and behaviors without abnormal behavior models, enhancing the accuracy and simplicity of detecting abnormal driving behaviors.
Patent Information
- Application Number
- DE102014204792
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2013-03-14
- Filing Date
- 2014-03-14
- Publication Date
- 2024-11-21
- Estimated Expiration
- 2034-03-14
AI Technical Summary
Existing methods for detecting abnormal driver behaviors face challenges in accurately determining whether driving behaviors are normal or abnormal due to the difficulty in collecting observed values during abnormal driving conditions and the variability of abnormal behaviors, making it hard to prepare comprehensive models.
A system that uses an autoregressive hidden Markov model (AR-HMM) to analyze driving conditions and behaviors without relying on pre-defined abnormal behavior models, by calculating mode probabilities and deviations from normal driving patterns to identify abnormal behaviors.
Enables accurate detection of abnormal driving behaviors with a simpler procedure, improving the accuracy and robustness of determining whether driving behaviors are normal or abnormal.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is based on and claims priority from Japanese patent application JP 2013 - 052 115 A, filed on March 14, 2013, the disclosure of which is incorporated by reference. TECHNICAL FIELD
[0002] The present disclosure relates to systems for detecting abnormal driver driving behaviors based on the driving conditions of a vehicle or the driver operation conditions of the vehicle. BACKGROUND
[0003] There is an urgent need to prevent vehicle accidents due to driver error in order to improve traffic safety. Regarding these needs, there are known methods for detecting abnormal driver behavior based on observed values indicating the driving conditions of a vehicle or the driver operation conditions of the vehicle. One of these methods, for example, is disclosed in Japanese Patent Application Publication No. JP 2009-154675 A.
[0004] The method disclosed in the patent publication uses models of normal behavior and models of abnormal behavior. Each of the normal behavior models represents a model of the driver's driving behavior when it is normal. Each of the abnormal behavior models represents a model obtained by modeling the driver's driving behavior when it is abnormal, which includes, for example, the driver's driving behavior when the driver falls asleep.
[0005] Specifically, the method collects cyclically observed values of the driving states of a vehicle or the driver operation states of the vehicle. Then, based on the previously obtained observed values and the models of normal behavior, the method estimates a current observed value as a first estimate, and based on the previously obtained observed values and the models of abnormal behavior, the method estimates a current observed value as a second estimate.
[0006] Then, the method determines whether a current observed value that is actually observed is closer to one of the first estimated value and the second estimated value than to the other thereof, and determines whether the driver driving behaviors are normal or abnormal.
[0007] Reference is also made to DE 60 2004 010 968 T2 and US 4 604 611 A, which have been identified as prior art. DE 60 2004 010 968 T2 describes a method for estimating a driver's alertness from averaged steering wheel deflection values. US 4 604 611 A describes a method for estimating a driver's fatigue from a ratio between steering movements and return movements. OVERVIEW OF THE INVENTION
[0008] In the above-described method disclosed in the patent publication, in order to prepare models of abnormal behavior, it is necessary to collect observed values of the driving conditions of a vehicle or the driver operation conditions of the vehicle while a driver is operating the vehicle abnormally. However, collecting these observed values may be difficult.
[0009] Furthermore, since there are different variations of abnormal driver behaviors, it may be difficult to prepare the models of abnormal behavior with respect to all variations of abnormal driver behaviors, which makes it difficult to determine with high accuracy whether the driver behaviors are normal or abnormal.
[0010] In view of the circumstances explained above, one aspect of the present disclosure seeks to provide abnormal driving behavior detection systems that can solve the problems explained above.
[0011] In particular, an alternative aspect of the present disclosure aims to provide such systems that can detect abnormal driver driving behaviors without using such models of abnormal behavior.
[0012] According to a first exemplary aspect of the present disclosure, a system for detecting abnormal driving behavior for a vehicle is provided. The system includes an acquirer that repeatedly acquires an observed value indicative of a driving state of the vehicle and / or a driver driving operation of the vehicle. The system includes a mode probability calculator that, each time an observed value is acquired at a given acquisition timing as a target acquired value, calculates a mode probability for each of a plurality of driving modes as a function of one or more previous observed values acquired prior to the target acquired value. Each of the plurality of driving modes is defined by modeling a group of normal driving behaviors.The mode probability for each of the plurality of driving modes represents a probability that a target driving mode corresponds to a corresponding one of the plurality of driving modes at the given acquisition timing. The system includes a deviation calculator that acquires, for comparison with the target acquired value, a predicted observed value for each of the plurality of driving modes using a model for normal driver behavior defined for a corresponding one of the plurality of driving modes, and calculates a deviation of the target observed value from the predicted observed value for each of the plurality of driving modes. The system includes an abnormality determination calculator that determines whether there is at least one abnormal driver behavior based on the mode probability for each of the plurality of driving modes and the deviation calculated for each of the plurality of driving modes.
[0013] According to a second exemplary aspect of the present disclosure, a program product usable for a system for detecting abnormal driving behavior for a vehicle is provided. The program product includes a non-transitory computer-readable medium; and a set of computer program instructions embedded in the computer-readable medium. The instructions cause a computer of a security system to: to repeatedly obtain an observed value indicative of a driving state of the vehicle and / or a driver driving operation of the vehicle; each time an observed value is acquired at a given acquisition timing as a target acquired value, calculate a mode probability for each of a plurality of driving modes as a function of one or more previous observed values acquired prior to the target acquired value, each of the plurality of driving modes being defined by modeling a set of normal driving behaviors, the mode probability for each of the plurality of driving modes representing a probability that a target driving mode at the given acquisition timing corresponds to a corresponding one of the plurality of driving modes; obtain a predicted observed value for each of the plurality of driving modes using a model for normal driver behavior defined for a corresponding one of the plurality of driving modes for comparison with the target obtained value; calculate a deviation of the target observed value from the predicted observed value for each of the plurality of driving modes; and to determine whether there is at least one abnormal driver behavior based on the mode probability for each of the plurality of driving modes and the deviation calculated for each of the plurality of driving modes.
[0014] The configuration of both the first and second exemplary aspects of the present disclosure determines whether there is at least one abnormal driver behavior based on the mode probability for each of the plurality of driving modes and the deviation calculated for each of the plurality of driving modes. Each of the plurality of driving modes is defined by modeling a group of normal driving behaviors. The mode probability for each of the plurality of driving modes represents a probability that a target driving mode corresponds to a corresponding one of the plurality of driving modes at the given acquisition timing. Thus, it is possible to determine whether there is at least one abnormal driver behavior without using abnormal behavior models, each of which is acquired by modeling driver driving behaviors when they are abnormal.Thus, the determination that there is at least one abnormal driver behavior can be carried out with higher accuracy and a simpler procedure.
[0015] The foregoing and / or other features and / or advantages of various aspects of the present disclosure will be further appreciated in light of the following description taken in conjunction with the drawings. Different aspects of the present disclosure may include and / or exclude different features and / or advantages where appropriate. Furthermore, different aspects of the present disclosure may combine one or more features of other embodiments where appropriate. The descriptions of features and / or advantages of specific embodiments are not intended to be considered as limiting other embodiments or the claims. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Further aspects of the present disclosure will become apparent from the following description of embodiments with reference to the drawings. They show: Fig. 1 is a block diagram schematically illustrating an example of the overall configuration of a driving support system SS according to an embodiment of the present disclosure; Fig. 2A is a schematic view of an AR-HMM (Autoregressive Hidden Markov Model) expressing a model of driver driving behavior according to the embodiment; Fig. 2B is a graph schematically illustrating parameters of a Gaussian distribution according to the embodiment; Fig. 3 is a flowchart schematically illustrating an example of an abnormal behavior detection task performed by a detection device included in Fig. 1 is illustrated, according to the embodiment; Fig. 4 is a flowchart schematically illustrating an example of a task for detecting a bad driving operation performed by the detecting device shown in Fig. 1 is illustrated, according to the embodiment; Fig. 5A is a view schematically illustrating a relationship between a first observed value and each of the first and second driving modes according to the embodiment; Fig. 5B is a view schematically illustrating a relationship between each of the first observed value and a second observed value and the first and second driving modes according to the embodiment; Fig. 6A shows graphs each schematically illustrating a relationship between a sequence of an observed value and driving modes selected therefor, according to the embodiment; Fig. 6B is a view schematically illustrating estimated driving modes while a motor vehicle is traveling on a circuit track such that each of the estimated driving modes correlates with a corresponding position of the circuit track; and Fig. 7 is a graph schematically illustrating a relationship between two sequences of observed values when a first driving mode is changed to a second driving mode; two sequences of predicted values corresponding to the observed values when the first driving mode is changed to the second driving mode; and a normal range for the first driving mode. DETAILED DESCRIPTION OF THE EMBODIMENT
[0017] An embodiment of the present disclosure will be explained below with reference to the drawings.
[0018] According to Fig. 1 illustrates a driving assistance system SS to which the embodiment of the present disclosure is applied. The driving assistance system SS is installed in a motor vehicle, referred to simply as a vehicle V. The driving assistance system SS includes sensors 2 installed in the vehicle V, an abnormal driving behavior detection system 1, and an information provider 3.
[0019] Some of the sensors 2, which serve as, for example, an acquisition device, are operative to detect the driving conditions of the vehicle V, and some of the sensors 2, which serve as, for example, an acquisition device, are operative to detect the driver operation conditions of the vehicle V. The driving conditions of the vehicle V detectable by some of the sensors 2 include the speed of the vehicle V, the longitudinal and horizontal accelerations of the vehicle V, the relative speed between the vehicle V and a preceding vehicle or a forward vehicle, etc. The driver operation conditions of the vehicle V include the rate of change of the acceleration operation component such as the accelerator pedal of the vehicle V, the pressure of the brake master cylinder of the vehicle V, the steering angle of the vehicle V, etc.Since the sensors 2 for detecting the above-mentioned driving conditions and the driver operation conditions of the vehicle V are known, additional descriptions thereof are omitted.
[0020] The abnormal driving behavior detection system 1 is communicatively connected to the sensors 2 and operative to determine whether there is at least one abnormal driver behavior as a function of the measured results of the sensors 2.
[0021] The information providing device 3 is equipped, for example, with a visible information output device and an audible information output device. Specifically, the information providing device 3 is operable to receive the determined results of the abnormal driving behavior detection system 1 and convert the determined results into visible information such as text information, geometric information, light information, or the like, and / or audible information such as sound information, alarm information, or the like. Then, the information providing device 3 is operable to provide the visible information and / or the audible information to a vehicle occupant such as the driver by means of a corresponding one of the visible information output device and / or the audible information output direction.
[0022] The abnormal driving behavior detection system 1 is configured to perform different operations based on an autoregressive hidden Markov model (AR-HMM) as one of models for observed sequence data.
[0023] First, parameters used by an AR-HMM are explained.
[0024] Fig. Figure 2A is a schematic view of an AR-HMM expressing a model of driver behavior. In Fig. 2A, the reference symbol t represents a present time, the reference symbol x t represents an observed value at the current time t, ie a measured value of a corresponding sensor 2 at time t, and the reference symbol z t represents a state variable that indicates a corresponding one of modes, ie, driving modes, that indicates a driver driving behavior at the current time t. A state variable z tis in a “hidden” state that is not directly observed. For this reason, a sequence Z t hidden state variables z1, z2, z t , which are represented by Z t = {z1, z2, z t ) is given based on several sequences Xt of observed values x1, x2,···,x t predicted, each of which is determined by X t ={x1, x2, x t} is given, measured by the corresponding sensors 2.
[0025] In the following, each of the individual state variables z1, z2,···, z t categorized into a given number of driving modes M1 to Mm (m is an integer not less than 2). That is, a given state variable z in the sequence Z t the state variables z1, z2, ···, z tbelongs to one of the driving modes M1 to Mm. Since these driving modes M1 to Mm are not directly observed, they are manipulated as hidden state variables. Specifically, normal driving behaviors and normal driving operations observed in different situations are grouped to be modeled as the driving modes M1 to Mm. In each of the driving modes M1 to Mm, the observed normal driving behaviors and observed normal driving operations included therein show a similar driving behavior or driving operation tendency. In other words, time-series behaviors of each piece of data observed by the sensors 2 are categorized into multiple groups, and each of the driving modes M1 to Mm shows an index of a corresponding one of the groups.For example, turning right or turning left as a normal driving operation can be divided into several driving steps, such as depressing the brake pedal and lightly turning the steering wheel, and the driving modes M1 to Mm respectively conceptually show the indexes of the divided driving steps.
[0026] Furthermore, in this embodiment, driving operations, for example, show actual operations of operating devices of the vehicle V, such as the accelerator pedal, the brake pedal, and the steering wheel. Driving behaviors, for example, show, in addition to observed driving behaviors, observed values of the operating states of the vehicle V, such as the vehicle speed and the acceleration of the vehicle V.
[0027] That is, the sequence of state variables z1, z2, ···, z t, each of which corresponds to one of the driving modes M1 to Mm, exhibits a corresponding driving behavior and / or driving operation. Thus, the state variables z1, z2, ···, z t , each of which corresponds to one of the driving modes M1 to Mm, primitive driver driving factors of a corresponding driving behavior and / or driving operation.
[0028] As in Fig. As illustrated in Figure 2A, autoregressive models can be constructed for each of the driving mode groups M1 to Mm and a sequence X t observed values can be expressed by the following equations (1) to (3). xt+1=Azxt+ε ε∼N(ε|μz,∑z) zt+1∼πz where: A zrepresents a driving behavior model, ie a model of normal driving behavior, which is a model in which average driving behaviors during normal driving, ie without any abnormal driving, in a corresponding one of the driving modes M1 to Mm; ε represents Gaussian-distributed noise, ie a normal distribution of a corresponding one of the driving mode groups M1 to Mm; µ z represents an average of a Gaussian distribution of a corresponding one of the driving modes M1 to Mm; Σ z represents a variance of a Gaussian distribution of a corresponding one of the driving modes M1 to Mm, where the average µ z and the variance Σ z are called mode distribution parameters that define a Gaussian distribution; N(ε|µ z ,Σ z ) represents a Gaussian distribution of noise ε, which is based on the mode distribution parameters µ z and Σ z is defined; and π zrepresents a mode transition probability, which is a transition probability between adjacent driving mode groups.
[0029] Note that the left side of each of equations (2) and (3) represents, with respect to the symbols ~, sampled values from a distribution defined by the right side of a corresponding one of equations (2) and (3).
[0030] These equations (1) to (3) can be set up under the following conditions: the distribution of observed values in each of the driving modes M1 to Mm follows a Gaussian distribution; and the mean and the variance, which are parameters describing a probability distribution, ie a Gaussian distribution, of observed values observed in a driving mode, are each given by µ z and Σ z (cf. Fig. 2B).
[0031] The abnormal driving behavior detection system 1 is configured to use as learned data values of these parameters A z , π z , µ z and Σ z to learn, ie train, and determine using observed values measured by the sensors 2 during normal driving, ie without abnormal driving, and a known learning algorithm such as a forward-backward algorithm.
[0032] Schematically, the abnormal driving behavior detection system 1 assigns one of the driving modes to each piece of learned data and calculates the mode distribution parameters µ based on the pieces of learned data to which the same driving modes are assigned. z and Σ zthe distribution of each of the driving modes. This allows the noise ε to be obtained for each of the driving modes according to equation (2) and hence the parameter, ie the driving behavior model parameter A z , can be obtained for each of the driving modes according to equation (1).
[0033] Furthermore, the abnormal driving behavior detection system 1 predicts the sequence of each driving mode and counts the number of transitions between each driving mode based on the predicted sequence of each driving mode. Then, the abnormal driving behavior detection system 1 calculates the mode transition probability π based on the results of the counting operation. z .
[0034] In particular, the system 1 learns and determines values of these parameters A to detect abnormal driving behavior z , π z , µ z and Σ zusing a well-known learning algorithm based on a Beta Process Autoregressive Hidden Markov Model (BP-AR-HMM). This allows these parameters A z , π z , µ z and Σ z and to automatically determine the number of driving modes in addition to this.
[0035] How these parameters A z , π z , µ z and Σ z A detailed description of how the number of driving modes and the number of driving modes can be determined using BP-AR-HMM is described, for example, in EB Fox, EB Sudderth, MI Jordan, and AS Willsky, "Sharing features among dynamical systems with beta processes," Advances in Neutral Information Processing Systems, Vol. 22, pp. 549–557 (2009). Additional descriptions in this regard are omitted.
[0036] Next, an example of the structure of the detection system 1 will be described with reference to Fig. 1 explained.
[0037] According to Fig. 1, the acquisition system 1 includes an observed value acquisition unit referred to as an acquisition unit 30, a storage unit 10, and a processing unit 20 communicatively coupled to the acquisition unit 30 and the storage unit 10.
[0038] The acquisition unit 30 is configured to cyclically acquire a measured value from each of the sensors 2 as an observed value x t and cyclically obtain an observed value x t from each of the sensors 2 to the processing unit 20.
[0039] The storage unit 10 stores these parameters, which define the driving modes M1 to Mm.
[0040] The processing unit 20 is configured to perform different operations including an operation that is performed as a function of the parameters stored in the storage unit 10 and the plurality of sequences Xt of observed values x1, x2, ···, xt obtained from the sensors 2 determines whether there are abnormal driver driving behaviors and / or poor driver driving operations.
[0041] For example, a bad driving operation is an operation of vehicle V that is poorly performed by a driver because the driver lacks the necessary skills for that operation, such as parking in a very confined space. An abnormal driving operation, for example, is an operation that no driver would normally attempt or perform.
[0042] In particular, the storage unit 10 consists of a mode transition probability memory 11, a mode distribution memory 12 and a behavior model memory 13.
[0043] The mode transition probability memory 11 is operative to store the mode transition probability π z to save.
[0044] The mode distribution memory 12 is operative to store the mode distribution parameters µz and Σ z for each of the driving modes M1 to Mm.
[0045] The behavior model memory 13 is operative to store the behavior model A z for each of the driving modes M1 to Mm.
[0046] The processing unit 20 consists of a mode probability calculation device 21, a deviation calculation device 22 and a detection device 23 which are operatively connected to the mode probability calculation device 21 and the deviation calculation device 22.
[0047] The mode probability calculation means 21 is operative to calculate a mode probability p(z t |X t ) for each of the driving modes M1 to Mm based on a target sequence Xt of observed values x1, x2, ···, x t , which have been obtained up to the present time t, the mode transition probability π z, which is stored in the probability memory 11, and the mode distribution parameters µ z and Σ z for a corresponding one of the driving modes M1 to Mm stored in the mode distribution memory 12.
[0048] The deviation calculation device 22 is operative to provide a normalized deviation d z,t for each of the driving modes M1 to Mm as a function of the behavior model A z for a corresponding one of the driving mode groups M1 to Mm stored in the behavior model memory 13, the mode distribution parameter µ z and Σ z for a corresponding one of the driving modes M1 to Mm and the target sequence Xt of observed values x1, x2, ···, x t to calculate.
[0049] The normalized deviation d z,t for each of the driving modes M1 to Mm shows a deviation of the target sequence X t observed values x1, x2, ··· , x tfrom one of the driving modes M1 to Mm.
[0050] The detection means 23, which serves, for example, as an abnormality determination means, is operative to determine whether there is at least one abnormal driver behavior as a function of the mode probability p(z t |X t ) for each of the driving modes M1 to Mm and the normalized deviation d z,t for a corresponding one of the driving modes M1 to Mm.
[0051] The mode probability calculation means 21, the deviation calculation means 22 and the detection means 23 are configured to perform these operations for each of the plurality of sequences Xt of observed values x1, x2, ···, x t to execute.
[0052] The processing unit 20 is designed, for example, as a microcomputer unit (programmed logic unit) consisting of a CPU 20a and / or a memory 20b (which is, for example, a non-volatile computer-readable storage medium) including a ROM and / or a RAM. The functional blocks included in Fig. 1 can be implemented by executing at least one program P stored in the memory 20b by the CPU 20a. As another example, the processing unit 20 can be designed as a hardware circuit consisting of hardware units each corresponding to the functional blocks shown in Fig. 1, or as a hardware / software hybrid circuit, wherein some of these functional blocks are implemented by some hardware units and the remaining functional blocks are implemented by software to be executed by the CPU 20a.
[0053] Next, further explanations will be given of the mode probability calculation means 21, the deviation calculation means 22, and the detection means 23.
[0054] Upon receiving a first observed value x1 of a target sequence X t the mode probability calculation means 21 calculates based on the mode probability parameters µ z and Σ z for each of the driving modes M1 to Mm, a probability p(x1|z) of the first observed value x1 is generated in each of the driving modes M1 to Mm. Then, the mode probability calculation device 21 sets the probability p(x1|z) for each of the driving modes M1 to Mm as an initial value p(z1|X1) of the mode probability p(z t |X t ) for a corresponding one of the driving modes M1 to Mm.
[0055] If the mode probability calculator receives a next, i.e. a second observed value x2 of the target sequence X t , the mode probability calculation means 21 calculates based on the mode distribution parameters µ z and Σ z for each of the driving modes M1 to Mm a probability p(x2|z) of the second observed value x2 generated in a corresponding one of the driving modes M1 to Mm.
[0056] Then, the mode probability calculation means 21 estimates, for each of the driving modes M1 to Mm, a probability P(z) of a corresponding one of the driving modes M1 to Mm at the acquisition timing of the second observed value x2 based on the mode transition probability π z and the initial value p(z1|X1) of the mode probability p(z t |X t) for a corresponding one of the driving modes M1 to Mm. Thereafter, the mode probability calculator 21 obtains, for each of the driving modes M1 to Mm, a mode probability p(z2|X2) for a corresponding one of the driving modes M1 to Mm using Bayesian estimation, the probability p(x2|z) as a likelihood, and the probability P(z) as a priori probability.
[0057] In particular, each time the mode probability calculation means 21 detects an observed value x t a target sequence X t at a current sampling cycle t, configured to: based on the mode distribution parameters µ z and Σ z for each of the driving modes M1 to Mm a probability p(x t |z) of the second observed value x t to calculate, is generated in a corresponding one of the driving modes M1 to Mm; for each of the driving mode groups M1 to Mm, a probability P(z) that a corresponding one of the driving modes M1 to Mm at the current sampling cycle t, ie at the acquisition timing of the observed value x t , based on the mode transition probability π z and the previous mode probability p(z t-j |X t-1 ) in the previous sampling cycle (t-1) for a corresponding one of the driving modes M1 to Mm; and for each of the driving modes M1 to Mm a mode probability p(z t |X t ) for a corresponding one of the driving modes M1 to Mm by using: Bayesian estimation; the probability p(x t |z) as a likelihood; and the probability P(z) as an a priori probability.
[0058] The mode probability p(z t |X t ) for each of the driving modes M1 to Mm: a first characteristic that when the driver driving operations are currently performed in one of the driving modes M1 to Mm, the mode probability p(z t |X t ) for one of the driving modes M1 to Mm, ie a current driving mode, a value becomes significantly higher than a value of the mode probability p(z t |X t ) for each of the remaining driving modes; and a second characteristic that, if the driver driving operations are not carried out in one of the driving modes M1 to Mm, no significantly high values of the corresponding mode probabilities p(z t |X t ) for driving modes M1 to Mm.
[0059] In particular, when driver driving operations are not performed in any of the driving modes M1 to Mm, the mode probabilities p(z t |X t) for all driving modes M1 to Mm intermediate values between the significantly high value of the mode probability p(z t |X t ) for the current driving mode and one of the values of the mode probabilities p(z t |X t ) for the remaining driving modes.
[0060] The deviation calculation device 22 consists, for example, of a storage module 221 for observed values, a deviation calculation module 222 and a normalization module 223. The storage module 221 for observed values is in Fig. 1 simply illustrated as MEMORY.
[0061] The observed value storage module 221 is operative to store an observed value x t-1 a target sequence X t-1 at a previous sampling cycle t-1 for each of the driving modes M1 to Mm.
[0062] The deviation calculation module 222 is operative to: an observed value xt' for each of the driving modes M1 to Mm based on the observed value x t-1 for a corresponding one of the driving modes M1 to Mm and the behavior model A z for a corresponding one of the driving modes M1 to Mm; and a deviation ε z,t of the predicted observed value xt' for each of the driving modes M1 to Mm from an observed value x t measured at the current sampling cycle t for a corresponding one of the driving modes M1 to Mm.
[0063] The normalization module 223 is operative to calculate the deviation ε z,t for each of the driving modes M1 to Mm using a probability of an observed value x t , which is the deviation ε z,t which is generated, thereby producing a normalized deviation d z,t for each of the driving modes M1 to Mm.
[0064] In particular, the deviation calculation module 222 calculates the deviation ε z,t according to the following equation (4), and the normalization module 223 calculates the normalized deviation d z,t according to the following equation (5): εz,t=xt'−Azxt−1 dz.t=1N(εz,t|μz,∑z) where N(ε z,t |µ z ,Σ z ) a probability of the observed value x t with the deviation ε z,t , generated in each of the driving modes M1 to Mm.
[0065] That is, the more the deviation ε z,t for each of the driving modes M1 to Mm from the average µ z of a corresponding driving mode M1 to Mm, the more the probability N(ε z,t |µ z ,Σ z ) for a corresponding one of the driving modes M1 to Mm.
[0066] Thus, using equation (5), the normalization module 223 calculates the normalized deviation d z,t for each of the driving modes M1 to Mm as the inverse of the probability N(ε z,t |µ z ,Σ z ) for a corresponding one of the driving modes M1 to Mm. This is because the more the deviation ε z,t for each of the driving modes M1 to Mm from the average µ z of a corresponding one of the driving modes M1 to Mm, the normalized deviation d z,t for a corresponding one of the driving modes M1 to Mm increases even more.
[0067] The normalized deviation d z,t simply as a deviation d z,t designated.
[0068] It should be noted that the deviation d z,t for each of the driving modes M1 to Mm by the deviation calculation means 222 for each sampling cycle and the obtained deviations d z,tfor the corresponding sampling cycles, for example, in the memory 20b.
[0069] The detection device 23 is configured to perform a task 231 for detecting abnormal behavior and a task 232 for detecting poor driving operation.
[0070] First, task 231 for detecting abnormal behavior is explained.
[0071] The detection device 23 performs the abnormal behavior detection task 231 when the mode probability p(z t |X t ) and the deviation d z,t for each of the driving modes M1 to Mm based on the observed value x t a target sequence X t , which is measured at a current sampling cycle t.
[0072] The abnormal behavior detection task 231, which is carried out by the detection device 23, performs a weighted addition of the deviation d z,t for each of the driving modes M1 to Mm using the mode probability p(z t |X t ) as a weighting coefficient for a corresponding one of the driving modes M1 to Mm according to the following equation (6), whereby an expected value E t the deviation d z,t for each of the driving modes M1 to Mm in step S110 of Fig. 3 is calculated: Et=∑zdz,tp(zt|Xt)
[0073] If the driver driving operations follow one of the driving modes M1 to Mm as a current driving mode, the mode probability p(z t |X t ) for the current driving mode, a value that is significantly higher than a value of the mode probability p(z t |X t) for each of the remaining driving modes. In contrast, the deviation d z,t for the current driving mode a lower value and the deviation d z,t for each of the remaining driving modes is a higher value. Thus, the expected value E t the deviation d z,t for the current driving mode, which is based on the multiplication of the corresponding mode probability p(z t |X t ) and deviation d z,t obtained is kept at a lower value.
[0074] On the other hand, if the driver driving operations do not follow any of the driving modes M1 to Mm, there are no driving modes M1 to Mm whose mode probabilities p(z t |X t ) have a significantly high value. This means that the mode probabilities p(z t |X t ) for all driving modes M1 to Mm, intermediate values between the significantly high value of the mode probability p(zt |X t ) for the current driving mode and one of the values of the mode probabilities p(z t |X t ) for the remaining driving modes. This results in the expected value E t the deviation d z,t for each of the driving modes M1 to Mm, whichever is a higher value.
[0075] After the operation at step S110, the abnormal behavior detection task 231 determines at step S120 whether the expected value E t the deviation d z,t for each of the driving modes M1 to Mm, which is calculated in step S110, is equal to or greater than a first threshold value. The first threshold value is previously set in step S120 to be sufficiently lower than the higher value of the expected value E t the deviation d z,t for each of the driving modes M1 to Mm.
[0076] Starting from the determination that the expected value E t the deviation d z,tfor each of the driving modes M1 to Mm calculated at step S110 is lower than the first threshold value (NO at step S120), the abnormal behavior detection task 231 determines that there are no abnormal driver behaviors, and thus the abnormal behavior detection task 231 is completed.
[0077] On the other hand, starting from the determination that the expected value E t the deviation d z,tfor each of the driving modes M1 to Mm calculated at step S110 is equal to or greater than the first threshold value (YES at step S120), the abnormal behavior detection task 231 determines that there is at least one abnormal driver behavior. Then, at step S130, the abnormal behavior detection task 231 sends the determined result indicating the detection of at least one abnormal driver behavior to the information provider 3, thereby causing the information provider 3 to provide visible and / or audible alarm information to a vehicle occupant such as the driver. After the operation at step S130, the abnormal behavior detection task 231 is completed.
[0078] Next, task 232 for detecting a bad driving operation is explained.
[0079] The detection device 23 performs the task 232 for detecting a bad driving operation every time a predetermined number of deviations d z,t , which corresponds to the same number of sampling cycles, has been stored for each of the driving modes M1 to Mm in the memory 20b (see YES at step S210). In other words, the detection device 23 does not execute the poor driving operation detection task 232 when the predetermined number of deviations d z,t for each of the driving modes M1 to Mm has not been stored in the memory 20b (see NO at step S210). For example, once the predetermined number of deviations d z,t for each of the driving modes M1 to Mm stored in the memory 20b used for the task of detecting a bad driving operation, they are deleted from the memory 20b or retained therein.
[0080] The predetermined number of deviations d z,tfor each of the driving modes M1 to Mm is determined such that an average value of the predetermined number of deviations d z,t for each of the driving modes M1 to Mm, in task 232 for detecting poor driving operation, a statistically reliable and sufficient value is obtained.
[0081] After positive determination at step S210, the poor driving operation detection task 232 calculates an average value of the predetermined number of deviations d for each of the driving modes M1 to Mm at step S220. z,t for a corresponding one of the driving modes M1 to Mm, stored in the memory 20b. The average value is referred to as a mode-to-mode averaged distribution.
[0082] After the operation at step S220, the poor driving operation detection task 232 determines at step S230 whether there is at least one driving mode whose mode-to-mode averaged distribution calculated at step S220 is equal to or greater than a second threshold value.
[0083] Based on the determination that there is no driving mode whose mode-to-mode averaged distributions calculated at step S220 are lower than the second threshold value (NO at step S230), the poor driving operation detection task 232 is completed.
[0084] On the other hand, based on the determination that there is at least one driving mode whose mode-by-mode average distribution calculated at step S220 is equal to or greater than the second threshold value (YES at step S230), the bad driving operation detection task 232 extracts the driver driving operations included in the at least one driving mode as bad driver driving operations at step S240. Then, the bad driving operation detection task 232 sends the determined result indicating the extracted bad driving operations to the information provider 3 at step S240, thereby causing the information provider 3 to provide visual and / or audible information about the extracted bad driving operations to a vehicle occupant such as the driver. After the operation at step S130, the bad driving operation detection task 232 is completed.
[0085] Next, overall operations of the abnormal driving behavior detection system 1 will be explained assuming that, for better understanding, the driving modes M1 to Mm are a first driving mode M1 and a second driving mode M2, ie, m = 2. This results in the distribution of each of the first driving mode M1 and the second driving mode M2 being based on the mode distribution parameters µ z and Σ z for a corresponding one of the first driving mode M1 and the second driving mode M1 M2.
[0086] According to Fig. 5A, a first observed value x1 of a target sequence X t by the processing unit 20 at a current time t = 1. In Fig. Figure 5A shows an ellipse E(M1) representing a distribution of observed values obtained in the first driving mode M1. The distribution is based on the corresponding mode distribution parameters µ z (M1) and Σ z(M1). In addition, an ellipse E(M2) shows a distribution of observed values obtained in the second driving mode M2. The distribution is based on the corresponding mode distribution parameters µ z (M2) and Σ z (M2) defined.
[0087] Under this assumption, we consider whether the first observed value x1 is generated in the distribution of the first driving mode M1 or in that of the second driving mode M2.
[0088] As explained above, the probability p(x1|M1) shows a probability that the first observed value x t is generated in the first driving mode M1, and the probability p(x1|M2) shows a probability that the first observed value x t in the second driving mode M2. In a Fig. In the case illustrated in Figure 5A, the probability p(x1|M1) is higher than the probability p(x1|M2), so that the hidden state, ie the driving mode, corresponding to the first observed value x t probably the first driving mode is M1.
[0089] Next, according to Fig. 5B a second observed value x2 of a target sequence X t by the processing unit 20 at a current time t = 2. As in the case where t = 1, we consider whether the second observed value x2 is generated in the distribution of the first driving mode M1 or in that of the second driving mode M2. In a case that Fig. As illustrated in Figure 5B, the probability p(x2|M1) is lower than the probability p(x2|M2), so there is a high probability that the hidden state, ie, the driving mode, corresponding to the second observed value x2 is the second driving mode M2.
[0090] At this time, the system 1 is configured to detect abnormal driving behavior, the mode probability p(z t |X t ) for each of the first driving mode M1 and the second driving mode M2 based on the mode transition probability π z which is the transition probability between the first driving mode M1 and the second driving mode M2. In particular, in a case where the mode transition probability π z has a lower value, which shows that the second observed value x2 is likely to be kept within the first driving mode M1, even if the probability p(x2|M2) is a higher value, the mode probability p(z t |X t ) for the second driving mode M2 is kept low. Thus, if the series of observed values x1 and x2, which are Fig. 5B, a high probability that the abnormal driving behavior detection system 1 determines that a series of hidden states estimated by the series of observed values x1 and x2 do not correspond to the first driving mode M1 and the second driving mode M2, but each correspond to the first driving mode M1.
[0091] Fig. Figure 6A illustrates schematically: a first sequence X t (1) observed values of observed values x t according to the rate of change of the acceleration operation component illustrated as “ACCELERATING PEDAL”; a second sequence X t (2) observed values of observed values x t according to the pressure of the brake master cylinder, which is illustrated as “BRAKE”; and a third sequence X t (3) observed values of observed values x t according to the steering angle.
[0092] It should be noted that for each of the observed values x t one of the driving modes M1 to Mm that has the highest mode probability p(z t |X t ) is selected.
[0093] Fig. 6A demonstrates that observed values each corresponding to the same driving mode exhibit a similar tendency, e.g., variation.
[0094] Fig. Figure 6B schematically illustrates the estimated driving modes based on observed values x t a sequence X t observed values measured while the motor vehicle V is traveling on a circuit track, so that each of the estimated driving modes correlates with a corresponding position on the circuit track at which the motor vehicle V is traveling. As in Fig. 6B, how the driving modes vary is based on the shape of the circuit, and how the driver driving operations vary is also based on the shape of the circuit.
[0095] Fig. Figure 7 schematically illustrates a graph showing: a first relationship between the second sequence X t (2) observed values of observed values x t according to the pressure of the brake master cylinder and the third sequence X t (3) observed values of observed values x t according to the steering angle when the first driving mode M1 is changed to the second driving mode M2; a second relationship between a sequence of predicted observed values xt' equal to the values A z x t-1 , which comes from the behavior model A z calculated that of the second sequence X t (2) observed values, and a sequence of predicted observed values xt' equal to the values A z x t-1 , which comes from the behavior model A z calculated that of the third sequence X t (3) observed values when the first driving mode M1 is changed to the second driving mode M2; and a normal range, ie an acceptable range, for the first driving mode M1, within which the deviation d z,t of each of the observed values x t the second sequence X t (2) observed values from a corresponding one of the predicted observed values xt'(Azxt−1) and the deviation d z,t of each of the observed values x t the third sequence X t (3) observed values from a corresponding one of the predicted observed values xt'(Azxt−1) are lower than the first limit value when the first driving mode M1 is changed to the second driving mode M2.
[0096] In particular, as in Fig. 7 is illustrated, if the deviation d z,t of each of the observed values x t the second sequence X t (2) observed values from a corresponding one of the predicted observed values xt'(Azxt−1) and the deviation d z,t of each of the observed values x t the third sequence X t (3) observed values from a corresponding one of the predicted observed values xt'(Azxt−1) lower than the first limit value to be within the normal range for the first driving mode M1, it is determined that there are no abnormal driving behaviors of the driver in the first driving mode M1.
[0097] In contrast, if either the deviation d z,t of at least one of the observed values x t the second sequence X t(2) observed values from a corresponding at least one of the predicted observed values xt'(Azxt−1) (A z x t-1 ) or the deviation d z,t of at least one of the observed values x t the third sequence X t (3) observed values from a corresponding at least one of the predicted observed values xt'(Azxt−1) is equal to or greater than the first threshold value to be outside the normal range for the first driving mode M1, determines that there is at least one abnormal driving behavior of the driver in the first driving mode M1.
[0098] As described above, the abnormal driving behavior detection system 1 is configured to: the mode probability p(z t |X t ) for each of the driving modes M1 to Mm as a function of a target sequence X t observed values x t, the mode transition probability π z and the mode distribution parameters µ z and Σ z for a corresponding one of the driving modes M1 to Mm; the deviation d z,t for each of the driving modes M1 to Mm and the expected value E t the deviation d z,t for a corresponding one of the driving modes M1 to Mm as a function of the model A z of normal driving behavior in a corresponding one of the driving modes M1 to Mm and the mode probability p(z t |X t ) for a corresponding one of the driving modes M1 to Mm; and to determine whether there is abnormal driving behavior of the driver as a function of the expected value E t the deviation d z,t for each of the driving modes M1 to Mm.
[0099] Thus, this configuration can detect abnormal driver behaviors without using abnormal behavior models, each of which is obtained by modeling driver behaviors when they are abnormal. Thus, it is possible to determine whether there is abnormal driver behavior with greater accuracy and a simpler procedure.
[0100] In particular, the abnormal driving behavior detection system 1 according to this embodiment is configured to calculate the mode probability p(z t |X t ) for each of the driving modes M1 to Mm with respect to the mode transition probability π z This configuration allows for determining with greater accuracy whether there are any abnormal driver behaviors involving abnormal mode transitions, while ensuring the robustness of System 1.
[0101] Furthermore, the abnormal driving behavior detection system 1 is configured such that the driving modes M1 to Mm and the model A z of normal driving behavior in a corresponding one of the driving modes M1 to Mm based on learning using a Beta Process Autoregressive Hidden Markov Model (BP-AR-HMM). This results in an automatic determination of the number of driving modes during learning using the BP-AR-HMM. The determined number of driving modes can be easily processed by computer. Thus, it is possible to further improve the distinguishability of the driving modes compared to a case where the number of driving modes is artificially determined.
[0102] The abnormal driving behavior detection system 1 is further configured to: an average value of a sufficient number of deviations d z,t to be calculated for each of the driving modes M1 to Mm; to determine whether there is at least one driving mode whose average value is equal to or greater than the second limit value; and based on the determination that there is at least one driving mode whose average value is equal to or greater than the second threshold value, extracting the driver driving operations included in the at least one driving mode as bad driving operations.
[0103] In particular, the state variables z1, z2, ···, z t A sequence, each of which corresponds to one of the driving modes M1 to Mm, constitutes primitive driver driving factors of a corresponding driving behavior and / or driving operation. Therefore, it is necessary to determine that there is at least one driving mode whose average value corresponds to a sufficient number of deviations d z,tfor the at least one driving mode is equal to or greater than the second threshold, it is possible to determine that the driver driving operations included in the at least one driving mode are bad driving operations.
[0104] The embodiment of the present disclosure has been described, but the present disclosure is not limited thereto. Specifically, the embodiment can be freely changed or modified within the scope of the present disclosure. For example, one or more functions included in a block of the abnormal driving behavior detection system 1 included in Fig. 1, to other blocks of the system 1 for detecting abnormal driving behavior, which Fig. 1. Furthermore, functions implemented in corresponding blocks of the system 1 for detecting abnormal driving behavior that is Fig.1 can be integrated into one block thereof. A part of the abnormal driving behavior detection system 1 according to the present embodiment can be replaced by a known structure having the same functions as the part of the abnormal driving behavior detection system 1.
[0105] For example, in the embodiment, the information providing device 3 provides the determined results of the abnormal driving behavior detection system 1 as visual information and / or audible information to a vehicle occupant such as the driver, but the present disclosure is not limited thereto. In particular, the abnormal driving behavior detection system 1 or another device may be configured to control one or more actuators, such as a brake actuator and / or a steering motor, to assist the driver's driving behaviors and / or operations according to the determined results of the abnormal driving behavior detection system 1.
[0106] While an illustrative embodiment of the present disclosure has been described, the present disclosure is not limited to the described embodiment, but includes any embodiments with modifications, omissions, combinations (e.g., aspects of the various embodiments), adaptations, and / or alternations that would be apparent to those skilled in the art based on the present disclosure. The limitations in the claims are to be interpreted broadly based on the language used in the claims and are not limited to examples discussed in the present specification or during the practice of the application, which are to be considered non-exclusive.
Claims
[1] System (1) for detecting abnormal driving behavior for a vehicle (V), the system (1) comprising: an obtaining device (2) that repeatedly obtains an observed value indicating a driving state of the vehicle (V) and / or a driver driving operation of the vehicle (V); a mode probability calculation device (21) which, each time an observed value is acquired at a given acquisition timing as a target acquired value, calculates a mode probability for each of a plurality of driving modes (M1 to Mm) as a function of one or more previous observed values acquired prior to the target acquired value, each of the plurality of driving modes (M1 to Mm) being defined by modeling a group of normal driving behaviors, the mode probability for each of the plurality of driving modes (M1 to Mm) representing a probability that a target driving mode at the given acquisition timing corresponds to a corresponding one of the plurality of driving modes (M1 to Mm); a deviation calculation device (22) which, for comparison with the target acquired value, obtains a predicted observed value for each of the plurality of driving modes (M1 to Mm) using a model for normal driver behavior defined for a corresponding one of the plurality of driving modes (M1 to Mm), and calculates a deviation of the target observed value from the predicted observed value for each of the plurality of driving modes (M1 to Mm); and an abnormality determination device (23) that determines whether there is at least one abnormal driver behavior based on the mode probability for each of the plurality of driving modes (M1 to Mm) and the deviation calculated for each of the plurality of driving modes (M1 to Mm). [2] System (1) for detecting abnormal driving behavior according to claim 1, characterized by that the abnormality determination device (23) based on the mode probability for each of the plurality of driving modes (M1 to Mm) and the deviation calculated for each of the plurality of driving modes (M1 to Mm), calculates an evaluation value of the deviation calculated for each of the plurality of driving modes (M1 to Mm); and Based on the evaluation value of the deviation calculated for each of the plurality of driving modes (M1 to Mm), determines whether there is at least one abnormal driver behavior. [3] System (1) for detecting abnormal driving behavior according to claim 2, characterized by that the abnormality determination device (23) determines whether the evaluation value of the deviation calculated for each of the plurality of driving modes (M1 to Mm) is equal to or greater than a first threshold value; and determines that there is at least one abnormal driver behavior based on determining that the evaluation value of the deviation calculated for at least one of the plurality of driving modes (M1 to Mm) is equal to or greater than the first threshold value. [4] System (1) for detecting abnormal driving behavior according to claim 2, characterized by that each of the multiple driving modes (M1 to Mm) is defined by modeling the set of normal driving behaviors using a Beta Process Autoregressive Hidden Markov Model. [5] System (1) for detecting abnormal driving behavior according to claim 2, characterized bythat the abnormality determination means (23) performs a weighted addition of the deviation for each of the plurality of driving modes (M1 to Mm) using the mode probability for a corresponding one of the plurality of driving modes (M1 to Mm) as a weighting coefficient, thereby calculating the evaluation value of the deviation calculated for each of the plurality of driving modes (M1 to Mm). [6] System (1) for detecting abnormal driving behavior according to claim 1, characterized by , that the deviation calculation means (22) calculates a predetermined number of deviations of a corresponding number of the target observed values from a corresponding number of the predicted observed values for each of the plurality of driving modes (M1 to Mm), the system (1) for detecting abnormal driving behavior further comprises: an average value calculating means that calculates an average value of the predetermined number of deviations for each of the plurality of driving modes (M1 to Mm); and a device for determining a bad operation, the determines whether the average value for each of the plurality of driving modes (M1 to Mm) is equal to or greater than a second threshold; and based on the determination that the average value for at least one of the plurality of driving modes (M1 to Mm) is equal to or greater than the second threshold, determines that there is at least one poor driving operation of the driver in the at least one of the plurality of driving modes (M1 to Mm). [7] A program product usable for a system (1) for detecting abnormal driving behavior for a vehicle (V), the program product comprising: a non-transitory computer-readable medium; and a set of computer program instructions embedded in the computer-readable medium, the instructions causing a computer of a security system to: repeatedly obtaining an observed value indicative of a driving state of the vehicle (V) and / or a driver driving operation of the vehicle (V); each time an observed value is acquired at a given acquisition timing as a target acquired value, calculating a mode probability for each of a plurality of driving modes (M1 to Mm) as a function of one or more previous observed values acquired prior to the target acquired value, wherein each of the plurality of driving modes (M1 to Mm) is defined by modeling a set of normal driving behaviors, the mode probability for each of the plurality of driving modes (M1 to Mm) representing a probability that a target driving mode at the given acquisition timing corresponds to a corresponding one of the plurality of driving modes (M1 to Mm); obtaining a predicted observed value for each of the plurality of driving modes (M1 to Mm) using a model for normal driver behavior defined for a corresponding one of the plurality of driving modes (M1 to Mm) for comparison with the target obtained value; calculate a deviation of the target observed value from the predicted observed value for each of the plurality of driving modes (M1 to Mm); and to determine whether there is at least one abnormal driver behavior based on the mode probability for each of the plurality of driving modes (M1 to Mm) and the deviation calculated for each of the plurality of driving modes (M1 to Mm).
Citation Information
Patent Citations
Photoacoustic image generating apparatus and method
JP2013052115A
Apparatus for estimating driver's wakefulness and method for estimating wakefulness
DE602004010968T2
Driver state estimation device and program
JP2009154675A
System and method for detecting driver drowsiness including detection of steering rotation and reversal
US4604611A
JP002009154675A
