From traffic conflict simulation to traffic crash simulation: introducing traffic safety indicators based on the explicit simulation of potential driver errors
URL: http://arxiv.org/abs/1808.01878v1
archive: archived pipeline: cataloged verified
Abstract
This paper introduces a general simulation framework that can allow the simulation of crashes and the evaluation of consequences on existing microsimulation packages. A specific family of simple and reproducible conflict indicators is proposed and applied to many case studies. In this approach driver failures are simulated by assuming that a driver stops reacting to an external stimulus and keeps driving at the current speed for a given time. The trajectory of the distracted driver vehicle is thus evaluated and projected, for the given time steps, for the established distraction time, over the actual trajectories of other vehicles. Every occurring crash is then evaluated in terms of energy involved in the crash, or with any other severity index (which can be easily calculated since the accident dynamics can be accurately simulated). The simulation of a driver error allows not only the typology of crashes to be included, normally accounted for with surrogate safety measures, but also many other type of typical crashes that it is impossible to simulate with microsimulation and traditional methodologies being caused by vehicles who are driving on non-conflicting trajectories such as drivers speeding at a red light, drivers taking the wrong lane or side of the street or just driving off the road in isolated accidents against external obstacles or traffic barriers. The total crash energy of all crashes is proposed as an indicator of risk and adopted in the case studies. Moreover, the concepts introduced in this paper allow scientists to define other relevant variables that can be used as surrogate safety indicators that consider driving errors. Preliminary results on different case studies have shown a great accordance of safety evaluations with statistical data and empirical expectations and also with other traditional safety indicators that are commonly used in microsimulation.
Summary
Methodological/conceptual paper (Astarita & Giofré, U. Calabria; preprint of 2019 Transportation Research Procedia article) introducing a new microsimulation framework for road safety evaluation that explicitly simulates driver errors (momentary attention/awareness lapses) and resulting crashes. The proposed framework projects a distracted vehicle's trajectory forward for a fixed time interval (ΔT) at constant speed and counts collisions with other vehicles, roadside obstacles, or barriers along that projection, scoring each by impact energy. Three case studies validate the approach against published crash data: (1) undivided sea-side road in Paola, Italy where SSAM with TTC=1.5s reports zero conflicts but the proposed indicator shows ~30% higher total collision energy than a straight road of equal length; (2) hybrid toll plazas (Florida) where ORT-on-mainline design D2 yields 9-20% higher total crash energy than D1, replicating Abuzwidah & Abdel-Aty (2018); (3) intersection layouts. The framework is integrated with VISSIM and TRITONE microsimulation packages and produces network 'danger maps' colored by crash energy density.
Key finding
Explicit simulation of driver errors (constant-speed projection during distraction interval ΔT) within microsimulation lets surrogate safety analysis capture crash types that conflict-only indicators (TTC, PET, SSAM) miss — including head-on crashes between non-intersecting trajectories on undivided roads and crashes against roadside barriers/obstacles. Results match observational crash statistics where conventional surrogate measures fail.
Methodology
Microsimulation-based methodology paper. Driver-error simulation framework integrated with VISSIM (case 1) and TRITONE (cases 2-3). For each case, multiple traffic scenarios (e.g., 200m undivided road, 500 veh/h/dir, 20 reps; toll plaza flows 600-2400 veh/h with 19% manual lanes; intersection 720+360 veh/h) were run; total collision energy was computed using a Z3-15-1/3 indicator. No human participants. Results compared against published observational crash data and against SSAM TTC<1.5s baseline.
Sample size: Simulation-based; no human participants. Multiple scenarios with 20 simulation repetitions per case study.
Quality score: 5 / 5