Layered Semi-Markov road generation for driving simulations

Abstract

Driving simulators offer several advantages in driver behavioral research, including low cost, repeatable experiments, and the ability to simulate scenarios that would be too dangerous for real cars. An effective simulation must elicit the same responses from drivers as a real car. Several factors contribute to this realism including a good physics engine, a realistic environment, and expected behavior from simulated traffic and drivers. In this paper, we present a novel method of generating a wide variety of realistic roads called Layered Semi-Markov Model. These roads are tested on five drivers with a driver distraction detection experiment. The results of this experiment matched the expected behavior and subjective feedback showed a favorable response.