D3.6 Enhanced toolbox of recommended data collection tools, monitoring methods and interventions including thresholds for the safety tolerance zone Safe tolerance zone calculation and interventions for driver-vehicle-environment interactions under challenging conditions

Abstract

The i-DREAMS project aims to establish a framework for the definition, development, testing and validation of a context-aware safety envelope for driving called the “Safety Tolerance Zone’ (STZ). Taking into account driver background factors and real-time risk indicators associated with the driving performance as well as the driver state and driving task complexity indicators, a continuous real-time assessment will be made to monitor and determine if a driver is within acceptable boundaries of safe operation. Moreover, safetyoriented interventions will be developed to inform or warn the driver in real-time as well as on an aggregated level after driving, through an app-and web-based gamification coaching platform (post-trip intervention). The conceptual framework of the i-DREAMS platform integrates aspects of monitoring (such as context, operator, vehicle, task complexity and coping capacity), to develop a Safety Tolerance Zone for driving. In-vehicle interventions and post-trip interventions will aim to keep the drivers within the Safety Tolerance Zone as well as provide feedback to the driver. This conceptual framework will be tested in simulator studies and three stages of field trials in Belgium, Greece, Germany, Portugal, and the United Kingdom with over 600 participants representing car, bus, truck, and rail drivers. This deliverable (D3.6) is an update of Deliverable 3.2, the second deliverable of the Operational Design Work Package of i-DREAMS. The original aim of D3.2 was to provide a more concrete description of the STZ by defining variables, values and thresholds associated with each phase of the STZ. D3.2 also made the initial steps in identifying mathematical models which had the potential to explain and analysis the resulting data – both from a realtime and post trip perspective. D3.2 was published early in the project (beginning of 2020) and since then there have been a number of further developments. These include, algorithm creation, technology updates and fitting test vehicles with the full technology to test performance of both the technology and STZ calculations. In addition the analysis work packages have commenced – WP6, analysis of risk factors and WP7, Evaluation of safety interventions. Together with the Mathematical Model Working Group (MMWG), a group formulated as part of WP3 to focus on evaluating and identifying the most appropriate analysis methodologies, WP6 and WP7 leaders have refined their plans based on the available data. This deliverable therefore constitutes an update to selected sections of D3.2. The original authors have worked with the WP3 partnership to update original text and added new text to this deliverable to reflect the above described developments in WP4, 6 and 7 as well as the MMWG work. The variables proposed in D3.2 have been confirmed with those available using the iDREAMS platform as developed by WP4. This has resulted in a list of variables that can be measured for which mode and that can be used to calculate STZ phases. The real-time warning strategies for the four performance objectives (Headway, Illegal overtaking, Speeding, Fatigue) that can be assigned variable thresholds are defined and threshold ranges are assigned to each STZ phase. For each of these four strategies additional variables can be used as indicators and/or modifiers and the types of real-time warnings are outlined. Driving style, in terms of ‘normal’ (STZ normal phase) and ‘abnormal’ (STZ danger and avoidable accident phase) is discussed and it was concluded that it is necessary to account for the possibility of the driver being in a ‘normal’ driving style for one performance indicator and an ‘abnormal’ driving style for another. A key aspect of defining the STZ, is measuring task complexity and (driver) coping capacity with safe driving defined as when these two dimensions are in balance. Therefore also which variables are associated with each of these are defined, alongside the method and frequency of recording, which mode are applicable and whether real-time or post-trip modelling methodologies are required for analysis. Alternative definitions of risk are discussed and described that relate to the STZ phases or the detection of an ‘event’ (discrete variables). In addition, ways in which the overall risk during a period of time are defined e.g. a composite STZ value or proportion of time spent in a STZ phase (continuous variables) Finally detailed descriptions of the relevant mathematical models (Dynamic Bayesian Network, Long Short-Term Memory, Discrete Choice Models, and Structural Equation Models) are provided with an explanation as to when they could be used for analysis. This depends on the variable type (discrete, continuous) and when the associated values are calculated (real-time or post-trip) were provided. For each model, the relevant independent variables or risk definitions that can feed into the model were defined and the relevant equations/functions were defined. Over the next six months, on road field trials will be conducted for the passenger car, bus and truck mode and simulator trials for the rail mode. Any learning from the simulator and field trials or changes to the platform that relate to this deliverable will be documented in the WP7 and WP6 deliverables that will be published at the end of the project.