Assessing expected accuracy of probe vehicle travel time reports

The use of probe vehicles to provide estimates of link. travel times has be en suggested as a means of obtaining travel times within signalized network s for use in advanced travel information systems. Past research in the lite rature has provided contradictory conclusions regarding the expected accura cy of these probe-based estimates, and consequently has estimated different levels of market penetration of probe vehicles required to sustain accurat e data within an advanced traveler information system. This paper examines the effect of sampling bias on the accuracy of the: probe estimates. Pin an alytical expression is derived on the basis of queuing theory to prove that bias in arrival time distributions and/or in the proportion of probes asso ciated with each link departure turning movement will lead to a systematic bias in the sample estimate of the mean delay. Subsequently, the potential for and impact of sampling bias on a signalized link is examined by simulat ing an arterial corridor. The analytical derivation and the simulation anal ysis show hat the reliability of probe-based average link travel times is h ighly affected by sampling bias. Furthermore, this analysis shows that the contradictory conclusions of previous research are directly related to the presence or absence of sample bias.