Abstract
The development of autonomous driving cars is a complex activity, which poses challenges about ethics, safety, cybersecurity, and social acceptance. The latter, in particular, poses new problems since passengers are used to manually driven vehicles; hence, they need to move their trust from a person to a computer. To smooth the transition towards autonomous vehicles, a delicate calibration of the driving functions should be performed, making the automation decision closest to the passengers’ expectations. The complexity of this calibration lies in the presence of a person in the loop: different settings of a given algorithm should be evaluated by assessing the human reaction to the vehicle decisions. With this work, we for an objective method to classify the people’s reaction to vehicle decisions. By adopting machine learning techniques, it is possible to analyze the passengers’ emotions while driving with alternative vehicle calibrations. Through the analysis of these emotions, it is possible to obtain an objective metric about the comfort feeling of the passengers. As a result, we developed a proof-of-concept implementation of a simple, yet effective, emotions recognition system. It can be deployed either into real vehicles or simulators, during the driving functions calibration.
Highlights
The development of Autonomous Vehicles (AVs) poses novel problems regarding ethics, safety, cybersecurity, and social acceptance
We developed a tool, called Facial Expressions Databases Classifier (FEDC) [20], able to perform different operations on the selected databases images in order to prepare them for the training of the neural networks
For the network in [22], we did not encounter any problems, while, for the network in [21], we faced an ambiguity in the “e-block” layer because, in the paper, it is not clearly described how to implement the relative “crop and resize” operation
Summary
The development of Autonomous Vehicles (AVs) poses novel problems regarding ethics, safety, cybersecurity, and social acceptance. It is possible to inject the fault we want to assess, engaging its failure effect mitigation algorithm At this point, thanks to the proposed emotions recognition system, it is possible to evaluate the effect of the algorithm on the passengers’ feelings, obtaining an objective metric on how the performance degradation can be acceptable by the users’ perspective. The idea is to provide objective classifications of the passengers’ reactions to autonomous vehicle decisions, helping the driving functions calibration on the basis of an analysis that is less sensitive to the subjectivity and variability of post-drive responses to questionnaires To achieve this result, we developed a system to recognize the emotions that the passengers are feeling with different calibrations in an objective and automatic manner.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
