An alternative approach to the estimation of the critical cornering sp
eed of large combination vehicles is presented. Existing techniques ar
e based on analytical models subject to a series of simplifying assump
tions, and do not explicitly consider grades and other three-dimension
al roadway features of the real problem. One result of this study indi
cates that these simplified models may overestimate critical cornering
speed by as much as 50%. The proposed technique combines detailed sim
ulation of the dynamics of the motion of large combination vehicles th
rough a three-dimensional roadway, with a Monte Carlo sampling of poss
ible vehicle configurations and component mechanical properties to pro
duce databases for statistical analysis. This approach is applied to t
he study of critical cornering speed at freeway-to-freeway connectors.
Databases for several samples of vehicles are generated, and a series
of regression models that predict critical cornering speed is created
. Results indicate that these regression models may be used to predict
critical cornering speed to a high degree of accuracy. Finally, criti
cal cornering speeds estimated using these models compare reasonably w
ith actual speed data collected at five freeway connectors.