D6.1 Analysis of task complexity factors

The main goal of the i-DREAMS project was to establish a framework that enables the definition, development, testing and validation of a context-aware safety envelope for driving called the ‘Safety Tolerance Zone’ (STZ). This could be accomplished through the implementation of a smart Driver, Vehicle & Environment Assessment and Monitoring System (i-DREAMS). With the i-DREAMS project, data was collected from car, truck and bus drivers during on-road trials conducted in Belgium, Germany, Greece, Portugal and the United Kingdom. 

The aim of this deliverable is to analyse the impact of task complexity on risk within the context of a four-phase on-road trial. The study consisted of four consecutive phases; Phase 1 involved observing driving behaviour without intervention following the installation of the i-DREAMS system. In Phase 2, in-vehicle real-time warnings were given using adaptive Advanced Driver Assistance Systems (ADAS) while monitoring continued. Phase 3 combined in-vehicle warnings with feedback via an app, and in Phase 4, gamification features were added to the app with the added support of a web dashboard. 

The aim of this report is to examine the impact of task complexity factors, such as road layout, traffic, time of day, weather, etc., on risk. The objectives are to determine which task complexity factors have the most significant impact on risk, create Structural Equation Models (SEM) to understand how task complexity affects the Safety Tolerance Zone (STZ) and compare the effects of task complexity on risk for different countries and transport modes during the four phases of the i-DREAMS road-trial. 

Task complexity relates to the current status of the real-world context in which a vehicle is being operated. Since this context is consistent of various individual elements which, together, determine the complexity of the task imposed on the vehicle operator, a multi-dimensional approach in further operationalizing this concept is adopted. In particular, task complexity context is monitored via registration of road layout (i.e., highway, rural, urban), time and location, traffic volumes (i.e., high, medium, low) and weather. 

In terms of the methodology, generalized linear and structural equation modelling techniques were utilized to investigate the factors that define task complexity and how it relates to risk. Both task complexity and risk were treated as latent variables, which are not directly observable. Despite a unified data collection design, technical issues such as sensor failures and driver availability arose during the data collection process in different countries. As a result, different datasets were obtained, and different variables were selected for the models to ensure their validity. 

The SEM analysis involved the development of four models per risk factor (e.g., speeding and headway), one for each phase, to identify any differences in the way task complexity impacts risk. However, due to the issues mentioned earlier, it was not possible to make a direct comparison between countries or transport modes. In some cases, not only the variables that represent task complexity vary, but also the variables that represent risk differ. Thus, the results could only be interpreted on a country and transport mode basis. It is noteworthy that age and gender were not significant factors in any of the models across different countries and transport modes. 

Measuring task complexity and relating this to risk was a challenging task as the number of variables that were collected and could be used was restricted and therefore, proxies were utilised. For instance, weather conditions were indicated by the use of the wipers and lighting conditions, or night-time driving was assessed by the use (or not) of the high beams. 

In general, the collection of the initially planned variables was proven to be trickier than anticipated. Future research should consider these challenges and attempt to incorporate information on factors like road configuration, traffic density, and other relevant metrics that would be very useful for establishing the complexity of the driving task and its association with risk.