■ TOP 　 ■ 概要 　 ■ 研究体制 　 ■ 研究成果 ■ アウトリーチ 　■ 解説講座 　 　■ リンク 　■ お問合せ

I. General study plan

1. Objectives of the study

a. Background and needs of the study

Although diverse safety measures have been taken over many years aiming to ensure the safety of vehicles, the accident ratio has scarcely decreased while the number of accidents has actually increased. It is estimated that 70 to 80% of accidents are attributable to driver errors, some of which are caused by complex, intertwined factors that are outside of the control of drivers. High recognition, judgment and handling skills are needed to control today’s fast-traveling, multi-function vehicles, and so aids appropriate to driving conditions and the driver’s state are indispensable to ensure safety.

b. Overview of the study throughout the entire study period

This study aims to develop a preventive safety system for detecting potential risks at an early stage and preventing them from affecting safety, as a novel means of reducing traffic accidents of commercial vehicles. The study involves the development of technologies for enhancing the abilities of drivers to recognize situations and for helping them to detect and avoid potential risks at an early stage, by collecting data on driving behaviors of trucks on roads, constructing a database, identifying potential risks in the environment, and analyzing the psychological states and intentions of drivers within a dynamic environment using methods for modeling behaviors, real-time sensing technologies, image understanding technologies and numerical methods.

c. General goals

The goals of the study are to develop various technologies needed to construct a preventive safety system that provides drivers with aids appropriate to each situation and the intentions of the drivers in order to greatly reduce traffic accidents. The technologies to be developed include: 1) real-time sensing systems for estimating the states of drivers and driving environment, 2) technologies for monitoring the driving behaviors, understanding road and traffic situations and driver’s intentions, and detecting increases in driving risk, 3) psychological aids, such as to reduce fatigue, 4) systems for giving advice appropriate to driving states and interfaces for helping drivers to recognize their states and ensure their safety, and 5) human-machine interactions and aids for the elderly, such as adaptive functional allocation technology for reducing the workload of drivers and ensuring safety in an emergency. Implementation of the system will be investigated by examining these functions under diverse situations which drivers may face.

 

2. Overview of the study

a. Studies on human-machine interaction

(1) Development of an adaptive functional allocation principle appropriate to situations and intentions

Lack of vigilance by drivers of fast, complex vehicles on the traffic environment and excessive trust in driving aids make it harder for them to recognize situations. To create a preventive safety system that provides aids to drivers appropriate to each traffic situation and driver’s intentions, adaptive functional allocation technologies will be developed for helping drivers to recognize situations and reduce their workload during normal driving, and to reduce the risks in an emergency. These technologies include those for 1) detecting high-risk psychological states in real time, 2) enhancing the ability of drivers to recognize situations, and 3) controlling vehicles safely in an emergency.

Excessive trust and lack of vigilance have been evaluated subjectively after accidents. In the “detection of high-risk psychological states in real time”, a technology will be developed to detect such psychological states in real time based on sensing data of driver’s posture, positions and movements of arms and legs, line of sight, etc. In the “enhancement of drivers’ ability to recognize situations”, information provision technologies and adaptive functional allocation technologies will be developed to improve recognition when the driver is suspected to be deviating from the normal state of alertness. In the “safety control in an emergency”, adaptive functional allocation technologies will be developed to minimize the driving risk and ensure safety when the vigilance level of the driver is insufficient and the vehicle is at risk of causing an accident.

 

b. Study on numerical information methods for understanding situations and intentions

(1) Development of probability statistical methods and video surveillance technology for understanding situations and intentions

(2) Development of technologies for enhancing vision using dynamic environment sensing

Many traffic accidents occur when drivers fail to correctly understand the surrounding situation and/or when communication between drivers and systems is insufficient. To prevent such accidents, aids will be created by developing element technologies, including methods for predicting essential conditions and situations in real time based on probability using imperfect data monitored by sensors and technology for estimating the states of drivers and the outside world from sensing information such as line of sight. A visual aid will also be developed for helping drivers to detect moving traffic bodies, dangerous movements of pedestrians and other states in a dynamic environment by detecting such movements that are difficult to see due to obstacles by installing road sensing devices, processing information received from the devices, and providing it to drivers.

Element technologies will also be developed for understanding situations and intentions, such as understanding and predicting situations from data of the driving behavior of the driver on an expressway and actual data such as video images of the driver and the road taken by on-board cameras. To develop a probability statistical method for understanding situations and intentions, methods will be investigated for extracting the dependency relationship between the causes and results of accidents from data on accidents as a probability cause-and-effect relationship and modeling the relationship using a probability network. The feasibility of the probability statistical method for analyzing driving behavior data will also be investigated. In the development of video surveillance technology, the existing system for detecting the front face will be expanded to include other face directions, and methods will be investigated for estimating the angle of the face from video images of drivers taken by on-board cameras.

To prevent vehicle-to-vehicle and vehicle-to-pedestrian collisions during right-turns, blind spots will be reduced by developing image processing technology for processing and providing images collected by roadside monitoring cameras to drivers in forms that are easy to see and technology for detecting pedestrians and moving bodies from videos; and a radio communication method between roadside monitoring infrastructure, such as cameras, and on-board displays will be investigated. Roadside infrastructure devices and on-board display units will also be developed, which will be tested at an intersection of university roads. Verification experiments will be conducted on eliminating blind spots for drivers making right turns and on detecting pedestrians to assess the effectiveness of the systems.

 

c. Study on modeling of driving behaviors

(1) Development of risk assessment technologies based on a database on long-distance driving behavior

(2) Development of technology for estimating driving state

During normal driving, drivers behave so as to not cause accidents, and the risk of accidents is lower than a certain level when the drivers drive appropriately. On the other hand, the risk rises when the drivers are distracted, lose concentration, etc. Therefore, a technology will be developed for assessing the risk of accident by measuring the behavior of long-distance driving and accumulating data in order to understand ordinary driving, and then detecting deviations from the normal states.

To achieve the goal, instruments will be developed for monitoring the driving behaviors of truck drivers during actual long-distance driving (driving, vehicle condition, distance from the car in front, position in the driving lane, etc.). The behavior of the drivers on expressways, etc. will be monitored and a database on driving behavior will be constructed. The data will then be used to construct models of normal driving behavior, based on which technology for detecting deviation from normal driving will be developed.

To contribute to safety education, safe driving behaviors demanded in terms of operation management will be identified, and a system will be developed for giving appropriate advice to operation managers and drivers to encourage safe driving in a traffic environment in which drivers see each other.

The study will involve surveys on actual states of operation management, compiling the requests of operation managers, and investigating and identifying desired safe driving behaviors. Simultaneously, driving behaviors will be monitored and analyzed including behaviors for checking the safety based on the constructed database on driving behavior and data of driving behaviors recorded by on-board units. The behaviors will be compared with the desired safe driving behavior, driving records and appropriate advice will be provided to the drivers and operation managers, and a support system for encouraging safe driving will be constructed.

 

d. Study on assessing the psycho-physiological states of drivers

(1) Development of chaotic algorithms for processing voice signals and a prototype on-board system

(2) Collecting evaluation voices and conducting evaluation experiments

(3) Physiological studies for verifying the reliability of the voice signal processing system

This study aims to establish a method for assessing the cerebral activity of truck drivers at work in real time, to identify changes in value of an evaluation index and the relationship between long-distance driving and fatigue, and to develop a preventive safety system that can be arbitrarily and complementarily combined with the results of other studies.

Indices to show the activity of the brain of a truck driver will be acquired by processing the speech of the driver using a chaotic method. In this study, methods will be established for collecting voices that can be processed from the driver’s seat environment, where the environmental noise levels are higher than those in offices, etc. Signal processing methods will also be established for calculating the index values from voices that may not always be of good quality or sufficient reliability.

The effectiveness of the indices in evaluating the fatigue of drivers will be verified by examining the correlation with indices used for conventional fatigue evaluation technologies, levels of stress-related hormones in the blood, and oxygen consumption by the brain, which can be measured by cerebral function diagnosis, in order to verify that the system to be created by the study will be sufficiently reliable.

 

e. Study on aids for the elderly

(1) Study on information provision methods appropriate for the elderly

(2) Study on learning aid mechanisms for improving driving behaviors

To reduce accidents among the growing population of elderly drivers, two studies will be conducted on the prevention of human errors: one on methods for presenting information to elderly drivers, and the other on learning for improving driving behaviors.

State-of-the-art driver’s aids are being installed on vehicles but may just confuse elderly drivers unless the systems match the physical and recognition characteristics of the drivers. Thus, in this study, optimum methods and timing of presenting information to elderly drivers will be identified by indoor and test course experiments. The effects of the information presentation on the driving behavior of the elderly drivers will also be investigated. The knowledge will then be used to identify the performance requirements of support systems that are easy to accept by elderly drivers.

Besides providing driving assistance and information, it is also effective to encourage elderly drivers to drive carefully by making them aware of their declining abilities. Methods will be investigated for providing information so that drivers themselves improve their driving behaviors. The provision of information appropriate to each driver will be investigated.

 

Mission statement (target results by system improvement)

1. Goals of the study

A number of analyses have been conducted on accidents involving moving vehicles, and 70 to 80% of the accidents have been found to be attributable to driver errors. Precise analyses of accidents have shown that a number of factors are involved in such errors in an interrelated manner, and many of them are outside the control of drivers. Furthermore, today’s fast, complex vehicles require drivers to have excellent driving skills and to detect risks and make decisions extremely quickly. Due to this requirement, no form of transportation, including vehicles, trains, ships, and airplanes, can reduce the ratio of accidents even though state-of-the-art technologies have been installed.

The difficulty of reducing the accident ratio can be resolved not by the conventional method of controlling direct causes of accidents but by developing and implementing a preventive safety system. This is indispensable for achieving a breakthrough toward the goal of reducing traffic accidents. Although this applies to all kinds of moving vehicles, this study focuses on developing systems for assisting drivers of commercial vehicles in manners appropriate to each situation and their intentions by identifying potential risks, conditions of the drivers, and their intentions using numerical and behavior models and considering the situations and psychological states of the drivers in order to reduce accidents.

Accidents involving commercial vehicles may cause many casualties and have increased by 50% in the last 10 years. This threatens the national goal of halving the number of traffic accident casualties, and so traffic safety measures for commercial vehicles need to be improved urgently.

For example, in fiscal 2001, there were 3,337 serious accidents caused by commercial vehicles. Of these, 55% could have been prevented if the drivers had correctly understood the driving conditions and their own states and had driven appropriately. Of accidents caused by limousine taxis and taxis, 65% were attributable to insufficient checking of safety, insufficient checking of movement, looking aside, erratic driving, failure to understand the situation, inappropriate intention, and fatigue. Drivers drive by detecting and predicting situations and formulating intentions appropriate to the situations. Inconsistencies among the situation, intention and actions increase the potential risk and may lead to accidents. Insufficient checking of safety, insufficient checking of movement, looking aside, erratic driving and fatigue are examples of such inconsistencies. Systems for preventing drivers from being exposed to such risk or detecting the risk at an early stage could prevent at least one half of all today’s accidents.

This study aims to greatly reduce traffic accidents by preventing drivers from being exposed to potential risk (such as insufficient checking of safety, insufficient checking of movement, looking aside, erratic driving and fatigue) and helping them to return to the normal state (safe driving) by sensing the traffic situation and the intentions and actions of drivers, and by detecting inconsistencies among situation, intention and actions in real time based on situation-intention correlation models developed using actually monitored data. As shown by the aforementioned statistics, this study will make a major contribution to reducing traffic accidents by half. The number of elderly drivers is increasing along with the aging of the Japanese population, and the number of fatal accidents involving elderly drivers is also increasing. Possible causes of accidents include declining physical and psychological capacities due to aging. This study also aims to develop information provision and safety technologies appropriate to the recognition, judging and operational capacities of each driver, which should help reduce accidents involving increasingly sophisticated commercial vehicles.

This study focuses on commercial vehicles since accidents caused by such vehicles have large social impacts (large number of casualties per accident and economic losses) and there are fewer restrictions on installing equipment on board (easier government control of the operators and sufficient space for installing the equipment). However, the technologies to be developed in this study will be deployable to ordinary vehicles in general, and so the study will help reduce accidents by all types of vehicles.

 

2. Concrete goals upon completion of the study

The five clusters of the preventive safety technologies to be developed in this study are: 1) real-time sensing systems, 2) technologies for monitoring the driving behaviors, and understanding road and traffic situations and driver’s intentions, 3) systems for evaluating the psychological states of drivers, 4) systems for giving advice appropriate to driving states and interfaces, and 5) human-machine interactions and aids for the elderly. Specific goals of each cluster are:

1) Real-time sensing

To present potential risks in each situation, technologies for processing images from environment monitoring cameras, roadside and on-board devices for monitoring the environment, radio communication methods between monitoring cameras and on-board units, and methods for identifying and tracking the direction of the driver’s face and the position of pedestrians using video images from on-board and roadside cameras will be developed. Verification experiments will be conducted using actual vehicles to test their effects on eliminating blind spots and to establish visual aids for drivers in dynamic environments.

2) Monitoring driving behaviors and understanding road and traffic situations and driver’s intentions

A device to be mounted on trucks will be developed that can monitor the driving behavior of long-distance driving for 10 hours. Data on actual driving will be accumulated to develop technologies for assessing the risks of driving following other vehicles based on a following driving model and the risks of negotiating curves based on handling and pedaling models at curves, both in three stages. A probability network will be constructed for forecasting accidents and estimating situations by evaluating the probability dependency relationship between the causes and results of accidents from data on driving behaviors and accidents in the past. A device will be developed based on presenting the estimated driving risks to drivers.

3) Evaluating the psychological states of drivers

A subsystem for collecting voices from the driver’s seat will be developed, and voice data will be accumulated to identify indices for evaluating the sensitivity of a chaotic voice signal processing system for judging the fatigue of drivers. Signal processing parameters to actualize sufficient sensitivity in a driving environment will be identified, and a system will be developed for detecting and warning the fatigue state of the drivers.

4) Giving advice appropriate to driving states and interfaces

Based on long-distance driving behaviors of drivers, high-risk driving behaviors will be identified, and a system for detecting deviation from ordinary driving from changes in face direction, etc. will be developed. Indices for evaluating safety driving will be established, and methods for giving appropriate advice to drivers and managers, when driving is not appropriate to the situation, will be developed based on the data of driving control monitoring with the objective of reducing the risk of accidents.

5) Human-machine interactions and aids for the elderly

A system will be developed for detecting a driver’s excessive dependency on driving aids in real time. An adaptive function allocation technology will be developed that dynamically changes the functional allocation (automation level) between driver and driving aids depending on the potential risk of dependency on and awareness of driving aids, driving loads and traffic situations and the available time for avoiding the risk.

Factors that determine the timing at which elderly drivers make judgments and their recognition characteristics, such as effective field of vision for detecting risks, will be identified to develop information provision systems for helping the elderly to recognize situations. Data will be accumulated on the reactions of the elderly to warming systems to identify appropriate combinations of information modes, timing of information provision, position of display, and the performances required of information systems.

 

3. Contributions to policies and impacts

Possible contributions and impacts of this study on achieving the policy goal of “Constructing a safe and comfortable society” include:

1) The preventive safety technologies for preventing potential risks will greatly reduce traffic accidents, the number of which has leveled off, and will improve the safety of people.

2) Accidents of commercial vehicles that involve many casualties can cause serious psychological impacts. A sharp reduction of accidents will make people feel safer.

3) A reduction of accidents will result in direct benefits of reducing economic losses caused by accidents, including 1.3 trillion yen for loss of human life and 1.6 billion yen for material loss.

4) The number of elderly drivers is steadily increasing and so is the number of fatal accidents caused by elderly drivers. Driving aids appropriate to the recognition, judging and operational capacities of elderly drivers will effectively reduce accidents among elderly drivers.

5) Technologies for estimating the psychophysical states and intentions of a person in real time will be applicable to diverse fields as well as for preventing and reducing traffic accidents.

6) Advanced information processing systems that are capable of recognizing their own states and making judgments will solve the problems and limitations (lack of ability to cope with unanticipated events and requests by people for flexibility) of today’s information processing systems, which are indispensable infrastructure.

7) Technologies to be developed in this study can be widely deployed to commercial vehicles and long-distance trucks, which are highly public in nature.

8) Technologies to be developed in this study will also be effective to prevent and reduce accidents of ordinary vehicles which are driven by amateur drivers who may not have had thorough training and who may be less skilled at recognizing situations, making judgments and controlling vehicles in an emergency.