Novel synthetic environment to design and validate future onboard interfaces for self-driving vehicles

This thesis presents a novel synthetic environment for supporting advanced explorations of user interfaces and interaction modalities for future transport systems. The main goal of the work is the definition of novel interfaces solutions designed for increasing trust in self-driving vehicles. The basic idea is to provide insights to the passengers concerning the information available to the Artificial Intelligence (AI) modules on-board of the car, including the driving behaviour of the vehicle and its decision making. Most of currently existing academic and industrial testbeds and vehicular simulators are designed to reproduce with high fidelity the ergonomic aspects associated with the driving experience. However, they have very low degrees of realism for what concerns the digital components of the various traffic scenarios. These includes the visuals of the driving simulator and the behaviours of both other vehicles on the road and pedestrians.  High visual testbed fidelity becomes an important pre-requisite for supporting the design and evaluation of future on-board interfaces. An innovative experimental testbed based on the hyper-realistic video game GTA V, has been developed to satisfy this need. To showcase its experimental flexibility, a set of selected user studies, presenting novel self-driving interfaces and associated user experience results, are described. These explore the capabilities of inducing trust in autonomous vehicles and explore Heads-Up Diplays (HUDs), Augmented Reality (ARs) and directional audio solutions. The work includes three core phases focusing on the development of software for the testbed, the definition of relevant interfaces and experiments and focused testing with panels comprising different user demographics. Specific investigations will focus on the design and exploration of a set of alternative visual feedback mechanisms (adopting AR visualizations) to gather information about the surrounding environment and AI decision making. The performances of these will be assessed with real users in respect of their capability to foster trust in the vehicle and on the level of understandability of the provided signals. Moreover, additional accessory studies will focus on the exploration of different designs for triggering driving handover, i.e. the transfer vehicle control from AI to human drivers, which is a central problem in current embodiments of self-driving vehicles.

Language

  • English

Media Info

  • Pagination: 54p
  • Serial:
    • TRITA-EECS-RP ;
    • Issue Number: 2018:3
    • Publisher: KTH Royal Institute of Technology, Sweden

Subject/Index Terms

Filing Info

  • Accession Number: 01739071
  • Record Type: Publication
  • Source Agency: Swedish National Road and Transport Research Institute (VTI)
  • ISBN: 9789177299912
  • Files: ITRD, VTI
  • Created Date: May 14 2020 9:41AM