An Experimental Evaluation of a Model-Free Predictor Framework in Teleoperated Vehicles*

A teleoperated vehicle is a vehicle operated by a human from a distance by means of a communication network. One important challenge with teleoperated vehicles is that communication delays in the network can negatively affect the mobility performance of the vehicle. This paper adopts and further develops a model-free predictor framework to compensate for communication delays and improve vehicle mobility where the term “model-free” indicates that the predictor does not need to know the dynamic equations governing the system. This framework has previously been conceived and applied to the teleoperated vehicle domain; however, prior evaluations have been conducted with simulated drivers and for only the speed control of the vehicle. The contribution of this paper is two-fold. First, the framework is further developed to improve the transient response of the predictors by including a saturation and resetting scheme. Second, to evaluate the effectiveness of the predictor framework with human drivers and combined speed and steering control, a human-in-the-loop simulation platform is developed to emulate a driving task in a virtual environment. Using this platform, human-in-the-loop experiments are performed, where humans are tasked with driving a typical military truck as fast as possible while keeping it as close as possible to the center of the track. Three types of experiments are conducted: (1) without communication delays as a benchmark; (2) with communication delays, but without the predictor framework to quantify the mobility performance degradation due to delays; and (3) with communication delays and the predictor framework to evaluate the change in mobility performance due to the predictor framework. Three metrics are used to quantify performance; namely, track completion time and track keeping error are used to quantify the speed and lateral control performance, respectively, and the steering control effort is monitored to assess drivability. Five drivers repeated each type of experiment seven times, and Analysis of Variance (ANOVA) is used to statistically analyze the results. The conclusion is that the predictor framework improves the mobility performance of the vehicle and increases drivability significantly.