Experimental Slip-based Road Condition Estimation

Abstract: Heavy traffic loads on the California highways have given birth to the development of automated highways. With vehicles traveling without human interaction, tighter spacing between cars canbeachieved without jeopardizng safety, leading to improved highway throughput. Since no human driver is present to make judgements about velocity and spacing, knowing the road condition is important in order to maintain safety. This project aims to, based on experimental measurements, give information about the road condition, and in this thesis a slip-based method is used. Slip is defined as the relative difference in velocity between the wheels and the vehicle. The data acquired from a Lincoln Towncar introduced di°culties due to very noisy measurements. A number of different approaches of extracting road surface information from the noisy slip data was examined and an observer was developed that signifcantly reduced unwanted effects caused by tire elasticity. The resulting road classifier could distinguish between dry and wet asphalt roads with 16% error probability. The classifier did only work for newly wet roads, most likely since roads are known to be the most slippery right after it has started to rain.