The Impact of Different Road Characteristics on Lane Keep Assist Performance

Author(s): Hildebrand, Stewart
Student Paper Competition: 1st Place

Slidedeck Presentation:




In order to fully exploit the safety promise of Lane Keep Assist technology that has emerged in most late model passenger vehicles, it is necessary for road authorities to understand what level of pavement marking performance is necessary for the systems to effectively delineate travel lanes.


The study evaluated different pavement marking schemes and road shoulder types under various light and weather conditions to identify those combinations where Lane Keep Assist (LKA) technologies were no longer able to discern the edges of travel lanes in order to warn distracted or inattentive drivers.


A total of 110 km of sampled sections of different functional classes of highways were driven using vehicles equipped with a Mobileye machine vision system (MVS) under various scenarios of light, weather, and pavement marking conditions. The LKA engagement rates were recorded and contrasted against site performance indicators such as pavement marking retroreflectivity and luminance contrast ratio.


It is necessary that an MVS system be able to detect both edges of a travel lane for ‘full’ LKA to be active. Good condition of the pavement markings (both in terms of retroreflectivity and contrast ratio) is key to ensuring high LKA engagement rates.

LKA engagement on the centreline of the sampled sections of Freeways, Arterials, Collectors and Locals was 100%, 98.8%, 98.5% and 89.4% during daytime testing, and 100%, 99.1%, 98.6% and 89.0% during nighttime testing, respectively. Sections classified as local roads achieved lower rates of full LKA engagement primarily the result of lower maintenance levels for centreline markings.

LKA engagement of the travel lane’s right edge was comparable to that observed for the centreline except where there were no edge pavement markings. Without a painted edge line, the type of shoulder and its condition greatly impacted LKA performance. Generally, LKA engagement increased along with the contrast between the travel lane and shoulder material and the sharpness of the boundary. Crushed rock surfaced shoulders and grassed shoulders only provided daytime LKA engagements averaging 6-10%, gravel 41%, and concrete curbs 57%.

Interestingly, the presence of artificial illumination and operating speed had little impact on the performance of LKA engagement. The presence of shadows or glare had a negative affect only when edge pavement markings were not present. The presence of salt over the road surface during winter served to reduce both retroreflectivity and contrast ratios to the point where LKA engagement rates were significantly reduced.


While LKA can, under the right contrast conditions, identify the right edge of a travel lane adjacent to a shoulder without the presence of an edge pavement marking, the percent of engagement is typically low thereby emphasizing the need for the provision of edge lines. LKA appears to work equally well with dashed or solid lines.


If we are to realize the full safety benefits of LKA technologies, it is incumbent on road authorities to ensure that travel lanes are delineated with well-maintained pavement markings, including edge lines adjacent to shoulders. Current suggested thresholds to enable LKA call for minimum retroreflectivity values of 34 mcd/lx/m2 and luminance contrast ratios that exceed 1.6. This research suggests that to achieve high percentages of LKA engagement under real-world conditions, these performance thresholds should be higher.