Alec Gorjestani

Evaluation of in-vehicle GPS-based lane position sensing for preventing road departure

We present a systematic method for quantifying the dynamic performance of differential GPS, and in particular the Novatel RT-20 DGPS unit, for determining a vehicles lateral position in a highway lane. Image processing is used. Novatels RT-20 Double Differencing Carrier Phase Measurement System is specified to achieve real-time positioning performance of better then 20 cm nominal accuracy. This paper documents the results from a series of dynamic tests carried out on the RT-20 to verify its actual accuracy while on a moving vehicle. The approach adopted here incorporates synchronized data acquisition using two separate computer systems, and experimental verification of the computational latency of the RT-20. The image processing scheme used for this analysis achieved high accuracy by taking advantage of subpixel resolution in the image processing algorithm. Our results indicate that the RT-20 system exhibited a mean error of 2.03 cm in the lateral direction, and 3.16 cm in the longitudinal direction (note that both lateral and longitudinal are with respect to the moving truck) while moving at speeds ranging from 15 mph through 40 mph. The corresponding error standard deviations were 1.98 cm and 34.87 cm respectively. Our main interest is in the lateral positioning performance of the RT-20, which turns out to be very good. Furthermore, we believe that the longitudinal error standard deviation exhibited by the RT-20 can be reduced further by using an algorithm that eliminates the outlier data points.

The Minnesota Mobile Intersection Surveillance System

Detailed crash analyses indicate that poor gap selection, rather than stop sign violation, is the primary causal factor in crashes at rural, unsignalized intersections. To determine under what conditions the gap selection process fails, a transportable rural intersection surveillance system has been designed and implemented. The system can be installed at any rural intersection, and can be used to collect data regarding the gap acceptance behavior of drivers at the rural intersection. Described herein is the design and performance of the transportable rural intersection surveillance system. This system will be deployed at eight rural intersections in eight US states, from April 2006 through December 2008. Data collected by the system will be used to determine whether regional differences in gap acceptance behavior exist. If differences exist, they will be quantified and used in the design of an intersection decision support system, a device under development designed to assist a driver at a rural intersection with the gap selection process

DGPS-based lane assist system for transit buses

Metro transit and the Minnesota DOT cooperatively operate a BRT-like system throughout the Twin Cities, Minnesota, metropolitan area. During peak congestion periods, buses operate on specially designated road shoulders (albeit at speeds significantly lower than limits posted for the adjacent highway). This allows buses to bypass congested roadways, enabling the bus to maintain its schedule regardless of traffic conditions. One of the problems faced by drivers using the shoulders is that the shoulders are typically no more than 3.1 m wide; a 12 m long transit bus measures 2.9 m across the rear view mirrors, and 2.6 m across the rear dual wheels. These narrow lanes require that a driver maintain a lateral error of less than 0.15 m to avoid collisions. This is a difficult task under the best conditions, and degrades to nearly impossible during conditions of bad weather, low visibility, high traffic congestion, etc. Metro transit drivers are not required to use the shoulders; shoulder use is left to their discretion. When poor conditions are encountered, many drivers choose not to use the shoulder. However, these poor conditions offer the greatest benefit of the bus-only shoulder use, creating an operational paradox. To minimize the effect of poor conditions on the use of bus-only shoulders, a lane assist system has been developed by the Intelligent Vehicles Lab at the University of Minnesota to help bus drivers under these difficult conditions. The system uses carrier phase, dual frequency differential GPS, a lane-level, high density, high accuracy geospatial database, and a lateral control algorithm for lateral assistance, radar for obstacle detection (critical in low visibility), and graphical, haptic, and tactile driver interfaces to provide guidance information to a driver. In addition to the system description, performance of the system on a operational bus-only shoulder is provided.

Rural Expressway Intersection Surveillance for Intersection Decision Support System

More than 30% of all vehicle crashes in the United States occur at intersections; these crashes result in nearly 9,000 annual fatalities, or approximately 25% of all traffic fatalities. Moreover, these crashes lead to approximately 1.5 million injuries per year, accounting for approximately 50% of all traffic injuries. In rural Minnesota, approximately one-third of all crashes occur at intersections. AASHTO recognized the significance of rural intersection crashes in its 1998 Strategic Highway Safety Plan and identified the development and use of new technologies as a key initiative to address the problem of intersection crashes. A study of 3,700 rural Minnesota intersections showed that right-angle crashes account for 36% of all rural intersection crashes. Approximately 50% of crashes at intersections that have higher than expected crash rates are right-angle crashes. Further investigation also found that poor gap selection is the predominant causal factor in these crashes. To address the problem of poor...

Range sensor evaluation for use in Cooperative Intersection Collision Avoidance Systems

The Intelligent Transportation Institute at the University of Minnesota is developing a Cooperative Intersection Collision Avoidance System - Stop Sign Assist (CICAS-SSA) for rural intersections as an alternative to signalized intersections. When deployed, the system will provide a driver stopped at a thru-stop intersection information about the available gaps in the mainline road traffic stream. The system uses surveillance sensors alongside the major road to determine the state1 of the intersection; this state information is used to determine whether the gaps that exist are unsafe, thereby triggering a warning to a driver not to initiate the desired maneuver. The system is capable of sending intersection state information to the vehicle (I2V) so that gap information can be displayed in the vehicle. Low cost automotive radar/laser sensors form the basis of the surveillance system. Described herein is a performance evaluation of a Delphi ESR radar sensor (ESR), an Ibeo Lux laser sensor (LUX), and a Smartmicro Umrr9 radar sensor (UMRR9). Each sensor was mounted adjacent to the shoulder on US 52 while a probe vehicle equipped with dual frequency, carrier phase DGPS was driven past. The accuracy of the position and speed measurements for each sensor were determined by comparison with the DGPS position and speed measured at the probe vehicle. An analysis was conducted to determine which sensor provided the best performance:cost ratio when used as a CICAS-SSA mainline sensor.

Advanced range sensor processing using DGPS and a geospatial database

Manufacturers of automotive radar typically use narrow beam angles to minimize the number of detected objects (traffic signs, guard rails, etc.) which ought not to pose a threat to the host vehicle. Although narrow beam angles are sufficient for some applications, namely automatic cruise control (ACC), wider fields of view are necessary for driver assistive systems. In order to make wide field of view range sensors perform well for driver assistive systems, a novel radar processing technique has been developed which integrates vehicle location provided by a high accuracy Differential Global Positioning System receiver and a highly detailed Geospatial Database map into the radar processing algorithm. Road objects such as road shoulders and road islands are used to delineate the driveable road surface. Objects detected by the range sensor which are located off of the driveable road surface are identified as such. Relevant vehicle systems (i.e., Heads Up Display or Collision Avoidance) can use this information to minimize false positives. This radar processor was implemented on an International snowplow and results from a series of experiments using this vehicle on Minnesota Trunk Highway 101 between Rogers and Elk River are presented. The system proved very effective at minimizing radar false positives.

A Demonstration of a Vision Enhancement System for State Patrol Vehicles

In conditions of poor weather and restricted visibility, it is necessary that emergency vehicles (e.g., police, ambulance, snowplows) are still able to operate and respond to critical events in the road environment. This report presents the result of a pilot study designed to evaluate a prototype Vision Enhancement System (VES) intended for use by state patrol vehicles. For this study, visibility was artificially restricted using smoked headlight covers during nighttime driving on a closed test track. The study examined the effect of preview distance and the presence of motion cues in a Head Up Display (HUD). Driving speed and self-reported data were the dependent measures related to safety and vehicle control. Despite the lack of a valid baseline, drivers reported that they felt their driving would be safer with the system than in the (imagined) case of unassisted driving in poor visibility. As a result of the apparent improvement in safety, all officers had a positive attitude toward the system.