Massachusetts Institute of Technology (MIT) researchers are developing an intelligent transportation system (ITS) algorithm that factors in models of human driving behavior to warn drivers of potential collisions and ultimately assumes control of the vehicle to prevent crashes.
MIT professor Domitilla Del Vecchio says that a common problem for ITS developers is designing a system that is safe without being too conservative and hypersensitive.
The researchers categorize driving actions into two primary modes--braking and accelerating, with a finite set of possible locations the vehicle could be in the future, depending on which mode a driver is in at a given moment. This set of potential positions, in conjunction with predictive human behavior models, were fed into the algorithm. The subsequent program can compute, for any two cars on the road approaching an intersection, a defined area in which the vehicles are in danger of colliding. The ITS-outfitted car then uses both internal and external sensor input to predict the other car's actions, and respond appropriately.
Tests with miniature vehicles revealed that failures to avoid collisions largely stemmed from communication delays, which can be resolved by reprogramming the algorithm.
From MIT News
View Full Article