Abstract

This paper proposes VehILux, a new vehicular mobility model based on real traffic counting data. It relies on two freely available sources of real information for the country of Luxembourg. The first source is traffic data collected by counting devices located on the Luxembourgian road network, while the second is geographical information about different types of areas: residential, industrial, commercial and other services. VehILux models vehicles commuting around the city of Luxembourg by considering two types of traffic, outer traffic with vehicles entering in the defined geographical area and inner traffic starting from residential zones located inside the geographical area. One part of the collected traffic data is used as input traffic, while another part is used to control the produced traffic and to fine-tune the model. VehILux is coupled with the microscopic road traffic simulator SUMO to produce realistic vehicular traces.

Abstract

International audience; The measurement of the operators' workload is an important aspect of usage-oriented design of professional systems. In domains such as avionics, air traffic management or mission systems, being able to quantify the operators' workload under stress, and in potentially demanding physical and mental conditions, is mandatory to anticipate overload and prevent human errors. Current approaches to workload estimation rely mainly on experimentation in simulation as an approach that has proven its efficiency for the identification of bad system and/or user interface design. Even if one cannot expect to totally avoid experimenting, given the complexity of the issue of workload computation, a priori estimation of workload might be an interesting tool to pre-validate a design in order to save some time in the experimentation phase and facilitate the analysis of overload situations that appear during experimentation. Various approaches to the a priori measurement of workload have been proposed: performance-based, physiological and subjective measures. Although performance and physiological measures of workload may be more precise, subjective measures are more practical, easier and less costly to use. For these reasons, they have been applied to many complex domains. The experience, the skills and the level of training of the operator have been identified in the literature as being important human factors. Nevertheless, these parameters have not been deeply analyzed in the context of workload estimation. In this paper, we develop a predictive workload model based on the analysis of the tasks assigned to a human operator. We propose to use mental representations of tasks, human actors, human roles, knowledge and abilities. We then propose to estimate the operator's workload with reference to his experience and training, the load over time and the task complexity. Our approach is illustrated on an airborne maritime surveillance use-case, in the context of the French Medusa project.