Intelligent Vehicles > Collision Avoidance > Lateral Avoidance

Printer-friendly version

• Introduction
• Technological Solutions
  • Principles of operation 
  • Sensing techniques  
• Systems in Development 
  • Alerting Systems
  • Correcting Systems 
• Project Evaluation
• References

INTRODUCTION

A major component of the National Highway Traffic Safety Administration's program for Collision Avoidance deals with lateral collisions. Two major types of lateral vehicle collisions, single vehicle roadway departure (SVRD) crashes and opposite direction (OD) crashes, account for approximately a quarter of the total number of collisions in the United States according to studies performed by the US Department of Transportation. The studies also concluded that OD and SVRD crashes are the most likely to be fatal, and that lane drifting due to driver inattention was the cause of 3.8% of the lane change/merger crashes, 6.8% of SVRD crashes and 17.8% of OD crashes. Numerous counts of these accidents occurred as a result of driver fatigue.

This paper presents current technological solutions to the Lateral Collision Avoidance problem, discussing systems in development and current techniques being used to improve vehicle performance and passenger safety. Project descriptions will include lateral sensing techniques, and human factors in Collision Avoidance Systems.

Figure 1: Lateral Collision Avoidance

TECHNOLOGICAL SOLUTIONS

Efforts are underway to develop technologies to support the driver's recognition and judgment during vehicle operation. Both public and private researchers have been developing technologies to increase the field of vision, and to enhance the basic maneuverability of vehicles to ensure safe driving. Lateral Collision Avoidance systems generally cover two areas of technology: systems that alert the driver to any foreseeable danger, and systems that assist and correct driver error. There are currently many projects being tested around the world to improve these specific areas of lateral collision avoidance.

Driver alerting systems prompt the driver with a warning when danger exists during a lane change or when the vehicle is having difficulties in lane-keeping. Allowing the driver to accurately recognize his environment constitutes the basis of preventative safety. Basic technologies to provide accurate recognition ensure an adequate field of vision in the vehicle. There are technologies currently in development that will appropriately locate roadway obstacles, ensure proper vehicle driving position, and optimize vehicle lighting.

Driver assisting systems are more complicated in nature, and involve a combination of roadway detection technology and in-vehicle mechanical systems to properly assess driver judgment and correct for driver error and inattention. Longitudinal Collision Avoidance technologies are further along in this area with the development of systems that assist vehicle operation such as Intelligent Cruise Control (ICC) and automatic braking systems. Driver assisting systems should noticeably reduce errors in recognition and judgment, without taking away the greater part of the driving responsibilities from the driver, as a purely automated system would.

Sensing Techniques

The main component of any Lateral Collision Avoidance System is the sensing mechanism used for monitoring lane position. Three basic sensor technologies have been investigated for the detection of lane boundaries or lane center: vision-based systems, laser scanning of lane markers, and coded magnetic arrays embedded in the roadway. The vehicle's lateral position relative to the road geometry is established by processing the raw sensor data.

Vision-Based Systems
Vision-based systems make use of complex algorithms to ensure proper lane-keeping of the vehicle. This type of Lateral CAS would use an on-board computer and employ binocular stereopsis to locate lane markers and obstacles that fall within the field of vision. The forward movement of the vehicle would continually update the information stored by the computer, which transforms obstacles and lane markers into geometric shapes that exist in the vehicle's field of vision.

Figure 2: Geometric transformation of obstacles using vision-based stereopsis

These vision-based systems are intended to work in conjunction with non-visual sensing techniques, such as magnetic sensors for lateral position measurement and active range sensors, for a more integrated approach to vehicle control. Those technologies are discussed next.

Magnetic Sensing
Another mechanism for lateral collision avoidance uses magnetic sensors to locate roadway markings, lane lines, intersection stop bars and shoulders. A lateral CAS requires a guidance system that determines the lateral position of the vehicle relative to these inputs. Magnetic sensing addresses this concern with minimal cost and exceptional accuracy, especially in adverse weather conditions where other systems tend to be less reliable.

A magnet-based lateral collision avoidance system uses a magnetometer to sense magnetic fields on the roadway left by magnetic markers that follow lane lines, shoulders, etc. The strength and direction of these fields is a function of the distance from the magnetic marker, and thus can be used to maintain a vehicle's course and alignment within a lane or along a designated right of way. Numerous methods have been tested using magnetic sensors. Magnetic markers have been placed in the center of lanes so that a magnetometer could position the vehicle precisely between lane lines. Other tests have magnetic sensors placed along lane lines so that a vehicle would stay a precise distance away from the magnetic field surrounding it.

The drawback for this lateral collision avoidance system is that it requires a roadway to be equipped with the magnetic markings before it can be used. Also, the magnetic sensing needs to be incorporated with a longitudinal collision avoidance system for optimal safety conditions.

Laser-Based Systems
Laser-based lateral collision avoidance systems have been developed to warn a driver when the vehicle begins to weave due to driver fatigue, drowsiness, or inability to properly control the vehicle for other reasons. These systems are based on laser technology that sends a laser beam to scan the roadway surface adjacent to the travelling vehicle. This takes advantage of the fact that highway lines and other markers are much more retro-reflective than the background surface, and that a very strong signal is obtained whenever the scanning beam crosses a roadway line.

Although much more costly than the other lateral collision avoidance systems previously mentioned, a laser-based system makes use of the existing roadway lane markers to detect lateral vehicle position. Specialized markers would not be a prerequisite for a roadway using this system. Also, a laser-based system is more compatible with vehicles already equipped with longitudinal collision avoidance systems employing laser-Doppler technology, such as Intelligent Cruise Control (ICC).

SYSTEMS IN DEVELOPMENT

There are currently many projects underway to develop and test various lateral collision avoidance systems. As was previously discussed, the technology can be divided into 2 major areas: systems that alert and systems that correct. The following are representative projects addressing these types of Lateral CAS.

California PATH and the University of California at Berkeley
The Machine Vision-Based Vehicle Control Project funded by California PATH, proposes a new approach for vision-based lateral control. Unlike past attempts incorporating vision-based systems that used single-camera ocular technology, this project uses binocular vision to locate lane markers and other obstacles. Knowing the position and alignment of the two cameras and their geometry to the road plane, the detection and measurement of the lane markers provides the positional parameters and the road curvature information that is needed for lateral vehicle control.

This system made use of a complicated lane marker detection algorithm to define a ground plane. From the image of the lane markers, geometric parameters in relation to the test vehicle were obtained. These parameters were continually updated through inputs from the cameras, and the lateral positioning of the vehicle was determined. This project was also successful in incorporating a longitudinal obstacle detection algorithm to define a 3-dimensional field and retrieve positional information of the test vehicle in relation to hazardous situations downstream.

3M and Honeywell
Researchers at Honeywell Technology Center in Minneapolis, MN along with 3M Corporation in St. Paul, have developed an all-weather system based on magnetic sensors. As discussed previously, magnetic lateral guidance works by providing a magnetic marker on or in the roadway surface to create a well-defined field pattern in a plane normal to the direction of travel in the traffic lane. 3M currently provides several different types of roadway marking products to the industry. These products, mostly 2 to 6 inch wide tape segments, provide optical lane delineation and are most useful at night when they provide excellent visibility due to the retroreflection of vehicle headlights. For this project, 3M has developed a tape which can be magnetized and surface applied, with the intention to produce a tape that is similar in application and optical performance to the existing tape, but with the additional magnetic feature.

One of the purposes of this project was to see how strong the magnetic field needed to be in order for the magnetometer to properly respond to lateral deviations. Trials were run using different magnetic field strengths and different magnetometer sensitivities. In the first set of experiments, Honeywell used a standard passenger car equipped with a magnetometer. More recently, experiments included real-world situations, where tests were run using a snowplow in adverse weather. A magnetometer was mounted on the curbside front bumper of a snowplow, and magnetic tape was applied as an edge marking along the sides of the roads. Results from both sets of experiments look very promising. In the future, this technology could be used as a means to reduce accidents on highways, with the tape serving as a high-tech rumble strip, with the vehicle providing an alert to the driver upon any lateral deviations from the lane.

As mentioned before, magnetic sensing costs less to implement than other forms of lateral vehicle control. Tape is gradually taking over from paint and other liquid systems for highway marking because of its durability. Though paint is initially cheaper, it needs replacement far more frequently. By applying magnetic sensing technology to standard highway markings, a significant advancement could be made in lateral collision avoidance with a minimal impact on cost.

Aerometrics
Aerometrics Incorporated, a firm from Sunnyvale, CA, is developing a laser diode-based device that provides a warning signal when a vehicle deviates from the center of its lane line. The device is based on a sensor that scans the roadway on either side of the vehicle and obtains the lateral position relative to the existing lines marking the lane. This beam also employs a multifaceted scanning mirror that receives the reflection of the laser beam back from the roadway surface and into a photodetector, which then processes the information to determine the lateral displacement of the vehicle.

Figure 3: Laser-diode scanning for roadway markers

The information obtained from the photodetector is transmitted to an on-board computer, which would be able to alert the driver with an audio or visual warning whenever the center of the lane is not maintained.

Although the laser-diode, on-board computer, and other testing equipment are quite expensive, no enhancements need to be made to existing roadways, so there would be no cost associated with system-wide infrastructure changes. But testing has shown some drawbacks, in particular, diode performance in adverse weather. When conditions are foggy or rainy, the photodetector has problems reading the positional data from the reflected light.

Correcting Systems

Cooperative Co-pilot with Active Steering Assistance
The concept of cooperation is to integrate both steering commands from the driver and the vehicle in order to keep a trajectory that is stable within the lane lines. Specifically, what is needed is a computerized assistant that can utilize the information of the road scene and driver inputs (steering torque, steer angle and speed) to exert a proper steer torque on the front wheels to correct for driver error.

Researchers in the Department of Power Mechanical Engineering at the National Tsing Hua University in Taiwan have been working on a new vehicular steering system named cooperative co-pilot. The concept is to develop a steering control system in which the vehicle's co-pilot can monitor the driver's actions and correct them as necessary. Using a single camera, the computerized co-pilot would analyze the roadway directly ahead of the test vehicle, and determine the bounds of feasible steering angles that the vehicle should make, and if need be, to correct for driver error in lane-keeping. The Command Integrator is the "brain" of the co-pilot, and uses a series of algorithms to determine the proper steering angle and speed. A sophisticated test vehicle was developed that was equipped with a steering actuator, which could apply torque to the front wheels in accordance with the calculations obtained from the Command Integrator.

Experiments proved quite successful, as a number of cases were tested (different road scenarios, different driver actions, speed, etc.). Results showed that the vehicle was able to keep the lane at low and high speeds, and that in normal situations, the vehicle was completely controlled by the driver without interference from the co-pilot. In dangerous situations caused by improper driver's actions or driver unawareness, the co-pilot was able to correct and maintain the vehicle steadily in the lane based on the strategy of the Command Integrator.

Oak Ridge National Laboratory and Scientific-Atlanta DASCAR Project
The Department of Energy's Oak Ridge National Laboratory (ORNL) and Scientific-Atlanta Inc. are working together on an advanced Data Acquisition System for Crash Avoidance Research (DASCAR). The project's goals are to reduce vehicle collisions by analyzing pre-crash situations and applying advanced technologies toward vehicular control. The lateral collision avoidance system used here is vision-based, but it is used in conjunction with a variety of technologies to maintain a satisfactory level of safety under many different situations.

In 1994, a prototype was made using a high-speed data processing unit, a four camera video imaging system, communications equipment, and vehicular controlling devices. Sensors strewn along the car's bottom and under the hood communicate with an on-board computerized system to record, process, and transmit data. DASCAR will gather near real-time information on the car's speed, pitch, lateral position within the lane and distance from cars ahead and behind. DASCAR also receives information from a control center to obtain characteristics of the roadway, changes in the weather, and traffic density. Using this abundance of information, the on-board computer runs multiple algorithms to ensure safe driving situations. The car can then be programmed to respond accordingly to any driver error. This technology is especially useful in lane deviations and car-following errors, where a majority of vehicular collisions occur.

PROJECT EVALUATION

The systems and technologies being applied in these projects show marked advances in Lateral Collision Avoidance. Each of the projects exhibited promising results, with the possibility of future expansion on a larger, more regional scale. Yet, it is hard to say what direction or technology Lateral CAS will follow (optical, magnetic, or laser). Whatever system is employed, that system needs to be reliable and compatible with systems in place, but also expandable and cooperative with up and coming IVHS technology.

REFERENCES

http://infosrv1.ctd.ornl.gov/Press_Releases/archive/crashes.html
"ORNL and Scientific-Atlanta Conduct Research on Avoiding Vehicle Crashes" Article by Carolyn Krause 1994

http://robotics.eecs.berkeley/edu/~jweber/MOU-131.html
"Machine Vision Based Vehicle Guidance: Lateral and Longitudinal Control"

Hsu, Jin-Chuan. "Cooperative Co-Pilot with Active Steering Assistance for Vehicle Lane Keeping" International Journal of Vehicle Design. 1998 Vol. 19 Number 1 Pages 78-107

Chachich, Alan C., Marten J. De Vries Transportation Sensors and Controls: Collision Avoidance, Traffic Management and ITS 1997

Samuel, Peter "Magnetic Strips Make Snow Ploughs Smarter" ITS International. Issue Number 8 January/February 1997 Pages 36-37

Bachalo, William D. "Advanced Laser-Based Tracking Device for Motor Vehicle Lane Position Monitoring and Steering Assistance" Collision Avoidance and Automated Traffic Management Sensors 1995 Pages 128-137

Stauffer, Dan, Mike Barrett, Nick Demma, Thomas Dahlin "Magnetic Lateral Guidance Sensors for Automated Highways" Collision Avoidance and Automated Traffic Management Sensors 1995 Pages 138-149

Author: Bryon Li. Date: 06/26/2000