Probabilistic Approach to Consistency in Geometric Design

Abstract

The practice of evaluating consistency in geometric design is addressed. The current practice is based on comparing the speed of a certain percentile value at the points of interest. The underlying reasoning is to control the amount of speed change incurred by the drivers. This paper shows that comparing the two speed distributions does not give the full picture of all the speed changes that the drivers incur. In the extreme case, it is argued that, theoretically, even when the two distributions are identical, it is still possible that each driver will experience a speed change. Hence, it is proposed to obtain and analyze the distribution of the speed differences. Using an empirical example, it is shown that the current approach is likely to underestimate the amount of the speed changes, and hence, the current approach is prone to accept inconsistent designs as consistent.