AutoExp: A multidisciplinary, multi-sensor framework to evaluate the acceptability of self-driving cars

Carlos F Crispim-Junior (1) , Romain Guesdon (1) , Christophe Jallais (2) , Florent Laroche (3) , Stephanie Souche-Le Corvec (3) , Xuguang Wang (4) , Georges Beurier (4) , Laure Tougne Rodet (1)

1 imagine - Extraction de Caractéristiques et Identification
2 TS2-LESCOT - Laboratoire Ergonomie et Sciences Cognitives pour les Transports
3 LAET - Laboratoire Aménagement Économie Transports
4 LBMC UMR T9406 - Laboratoire de Biomécanique et Mécanique des Chocs

The arrival of self-driving cars (SDC) in our daily lives is imminent. However, research on the adoption of this technology is still an ongoing topic. Existing work focuses on situations where the driver may be asked to take back control of the vehicle. Therefore, we currently lack tools to analyze the behaviors of the occupants of SDC beyond driving-related activities, and studies in close to real-world situations. We propose a multidisciplinary, multi-sensor framework to evaluate the changes in the internal states and behaviors of SDC's occupants. To test the proposed framework, we carried out a four-day long experiment in July 2021 using a Renault Zoe car (electric supermini urban model). The experiment took place in the parking of the campus of Ecole Centrale de Nantes (ECN) in France, and the vehicle was robotized by the LS2N-ARMEN laboratory to behave as a SDC of level four (high driving automation). We acquired a multidisciplinary, multi-sensor dataset composed of recordings of 29 people (18 men/11 women) carrying out a daily domicile-work travel. Recruited participants presented a variety of body sizes, human traits, ages, and education levels. We hope that the proposed framework and the acquired dataset will encourage interdisciplinary research between AI field and Human and Social Sciences on the study of the acceptance of SDC, and of the transformations that our society is undergoing.

Mots clés: activity analysis; self-driving cars; computer vision; internal state assessment; value of time

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Carlos F Crispim-Junior, Romain Guesdon, Christophe Jallais, Florent Laroche, Stephanie Souche-Le Corvec, et al.. AutoExp: A multidisciplinary, multi-sensor framework to evaluate the acceptability of self-driving cars. [Research Report] LIRIS UMR 5205 CNRS/INSA de Lyon/Université Claude Bernard Lyon 1/Université Lumière Lyon 2/École Centrale de Lyon. 2022. ⟨hal-03825090⟩