Activity and mood-based routing for autonomous vehicles
AActivity and mood-based routing forautonomous vehicles
Ankit Kariryaa
University of BremenBremen, [email protected]
Tony Veale
University College DublinDublin, [email protected]
Johannes Schöning
University of BremenBremen, [email protected]
Permission to make digital or hard copies of part or all of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full citationon the first page. Copyrights for third-party components of this work must be honored.For all other uses, contact the Owner/Author.Copyright is held by the owner/author(s).
Abstract
A significant amount of our daily lives is dedicated to driv-ing, leading to an unavoidable exposure to driving-relatedstress. The rise of autonomous vehicles will likely lessenthe extent of this stress and enhance the routine travel-ing experience. Yet, no matter how diverse they may be,current routing criteria are limited to considering only thepassive preferences of a vehicle’s users. Thus, to enhancethe overall driving experience in autonomous vehicles, weadvocate here for the diversification of routing criteria, byadditionally emphasizing activity- and mood-based require-ments.
Author Keywords
Routing preference; Autonomous vehicles;
ACM Classification Keywords
H.5.m [Information interfaces and presentation (e.g., HCI)]:Miscellaneous
Introduction & Motivation
A significant amount of our time is typically spent driving.For example, in 2014, 83.4% of total inland travel in theEU-28 was carried out in passenger cars [4]. The nega-tive effects of driving-related stress are well known [7], andwhile autonomous vehicles are expected to remove some ofthese stressors, such as e.g., the behavior of other drivers, a r X i v : . [ c s . H C ] A ug he cognitive (over)load of driving and other ergonomic fac-tors, other external stressors, such as e.g. environmentalconditions, traffic congestion, and peripheral noise, are in-dependent of the vehicles themselves. To create a pleasantdriving experience and to improve a user’s interactions withan autonomous vehicle, we believe users should be giventhe possibility to limit the influence of external stressors.The related literature provides a large body of work on cri-teria that optimize route-finding by distance, time, safety,simplicity, efficiency and even scenery and other aestheticpreferences. We propose here the enhancement of route-finding by activity and mood to improve the user’s overallexperience (and enjoyment) of travel in an autonomous ve-hicle. Related work
A plethora of routing techniques supplement the most fre-quently used fastest and shortest route criteria. Based onthe seminal work of Golledge in 1995 [6], Johnson et al.[8] categorizes these alternative routing techniques intothe positive , negative , topological and personalized . The positive category encompasses routes that are the mostappealing or attractive. One such technique is described byRunge et al. [14]. It is capable of generating scenic routesvia a classification of Google Street View images. Similarly,Ali et al. [3] present an exploration-based route plannerthat learns from the routes commonly taken by photogra-phers. As a result, it generates aesthetically-pleasing routesfor city exploration and/or photography sessions. Routesthat include the least number of unfavorable or adverseconditions to a particular user are classified as negative routes, not because the routes themselves are negative butbecause routing decisions are made on the basis of dis-cernible negatives. Shah et al. [15] demonstrated an algo-rithm for generating safe routes, deduced from the reportedlevels of crime in any given locality. Li et al. [11] created a routing service to be used in cases of natural and man-made disasters. It first optimizes on measures of surviv-ability, and only then on travel time. Topological routes givehigher preference to factors such as simplicity or efficiency.Duckham and Kulik [2] proposed a routing algorithm forsimple routes that are easy to describe and execute, whileGanti et al. [5] presented a fuel-efficient routing algorithm.Among all approaches to routing, only personalized routestake into account the preferences of individual users. Letch-ner et al. [10] introduced the concept of an individual inef-ficiency ratio (r) – which is the degree to which a particularuser deviates from the fastest route – to fulfil personal pref-erences. It is based upon a large dataset of personal GPStraces. This technique suggests routes by optimizing for thepreferred ratio “r” of the user. Delling et al. [1] proposed arouting algorithm that creates routes based upon a user’spreferences for speed level, type of road, and number andtype of turns. While there has been extensive research intosimple routing strategies, as well as recent research intopersonalized routing, we argue that the diverse needs ofusers cannot be satisfied with routing techniques that ex-ploit only the passive preferences of the user. Users of non-autonomous vehicles can always optimize the suggestedroute based upon their current needs, but such interven-tions become problematic in the context of autonomousvehicles where active control is passed from the user to thevehicle itself. Hence, we propose a routing approach thatconsiders these active elements and improves upon them. Activity and mood-based routing
Research in route planning with individualized preferencesshows that users prefer different routes in different condi-tions [1, 10]. However, little has been done to integrate themood and activity of the user in the driving preferences, thatcan play a crucial role in the overall driving experience. In-deed, there is an even greater potential to incorporate theood and activity of a non-driving user of an autonomousvehicle into routing preferences. While some of these pref-erences may be static – such as avoiding areas of extremeweather or natural/social hazard – we argue that manyother preferences are dynamic, and in many cases, userscan articulate their requirements explicitly (e.g., the goalof reaching a destination in the shortest time). However,in some cases a requirement cannot be mapped onto amathematical function that can then form the basis of opti-mization. The circumplex model of affect [13] maps humanemotions onto a valence and activation scale (see Fig. 1).Since different emotions are best suited to certain activities(as was demonstrated by [9]), our goal for an empathic au-tonomous vehicle is to alter the user’s mood to match thedesired activity. Here, we discuss some scenarios whereactivity- and mood-based routing will be required.Uplifting routes: Picard’s experiments (see [12]) have demon-strated the crucial role of affect in human perception andcognition, as well as its influence on such tasks as learningand decision-making. If a user experiences an unpleasantemotion while driving, one which the user would like to al-ter, the route itself can foster this change (see Fig. 1). Forinstance, the route may change to visit places with asso-ciated personal value, such as the user’s favorite restau-rant, lively centers of public performance, or places of nat-ural beauty, in order to nurture more pleasant emotions. Ascenic route can be chosen so as to maximize the effect ofsurroundings and to highlight ambient environmental condi-tions, such as e.g. a sunset.Sleep-aiding routes: If the user of an autonomous vehiclewishes to sleep during a journey, then the route preferencecan be optimized for the lowest level of noise and the lowestamount of other road-related disturbances, e.g. a route withthe fewest turns, ensuring a smooth and continuous drive. It may also factor in additional factors such as the sleeppattern and the required sleep time of the user to promote asatisfying sleep.Movie routes: If the users are watching a movie, then theroute and speed of an autonomous vehicle can be variedto match the duration of the movie. It can also be chosen tosuit its theme and to minimize environmental disturbances,e.g. at night a country road can be given a higher prefer-ence in order to minimize outer light disturbances. In cer-tain cases, it may also be possible to route through loca-tions relevant to (or analogous to) scenes in the movie, attimes that match the timing of scenes in the movie, to fur-ther enhancing the experience.Educational routes: The autonomous vehicle can alsopropose routes based upon the educational needs of theusers. For example, a route can pass through important ar-eas and landmarks to familiarize users with the history ofan area. A route can also follow historical events, such asthe course taken by an important procession or march, togive a deep insight into the locality. The route can also bechosen based upon the needs of a specific group, e.g. itcan include farms, parks, museums or galleries.
Discussion and future works
To enhance the user experience of autonomous vehicles,we have proposed new routing criteria, including both thecurrent activity/goal and mood of the user, that should betaken into account when selecting the “best” route to auser’s destination. These dynamic factors will diversify theroute-finding process and reduce the influence of hiddenexternalities.Subsequently, an archive of chosen routes for a particu-lar user can be used as a basis for engaging user interac-tion, by including ideas from both the user and the system. igure 1:
The route preference can be adapted to drive theemotions according to the requirement. Based upon the graphicalcircumplex model of affect in [13].
A creative route-finding system might come up with cre-ative new types of route, such as routes whose road-namesspell the name of a known person, or routes comprisingthe oldest or most historical roads, or routes whose road-names provide the answers to riddles and quiz questions.Likewise, a system may creatively choose a route basedupon the plot of novel or a TV-show/movie (e.g.
Game ofThrones ) that was filmed in or, set in, the locality, and en-tertain the user with insights about both the plot and thesurrounding area. A creative system may also engage withthe user in novel ways, for example, by generating a routethat passes through the largest number of a certain typeof building or establishment (e.g. bookshops, bars, restau-rants) and play a game with users by asking them to keeptrack of the such establishments. In fact, a creative route-finding system should serve many of the same functionsas a creative travel-companion, one that possesses muchmore than a map and an eye for traffic.
Acknowledgements
We would like to thank Nicolas Autzen and Tetiana Gren forvaluable discussion on the topic. This work was supportedby the Volkswagen Foundation through a Lichtenberg pro-fessorship.
References [1] Daniel Delling, Andrew V. Goldberg, Moises Gold-szmidt, John Krumm, Kunal Talwar, and Renato F.Werneck. 2015. Navigation Made Personal: InferringDriving Preferences from GPS Traces. In
Proceed-ings of the 23rd SIGSPATIAL International Confer-ence on Advances in Geographic Information Systems(SIGSPATIAL ’15) . ACM, New York, NY, USA, Arti-cle 31, 9 pages.
DOI: http://dx.doi.org/10.1145/2820783.2820808 [2] Matt Duckham and Lars Kulik. 2003. Simplest paths:utomated route selection for navigation. In
Inter-national Conference on Spatial Information Theory .Springer, 169–185.[3] Abdallah El Ali, Sicco NA Van Sas, and Frank Nack.2013. Photographer paths: sequence alignment ofgeotagged photos for exploration-based route plan-ning. In
Proceedings of the 2013 conference on Com-puter supported cooperative work . ACM, 985–994.[4] Eutostat. 2014. Passenger transport statistics. (2014). http://ec.europa.eu/eurostat/statistics-explained/index.php/Passenger_transport_statistics [5] Raghu K. Ganti, Nam Pham, Hossein Ahmadi,Saurabh Nangia, and Tarek F. Abdelzaher. 2010.GreenGPS: A Participatory Sensing Fuel-efficientMaps Application. In
Proceedings of the 8th Interna-tional Conference on Mobile Systems, Applications,and Services (MobiSys ’10) . ACM, New York, NY,USA, 151–164.
DOI: http://dx.doi.org/10.1145/1814433.1814450 [6] Reginald G Golledge. 1995. Defining the criteria usedin path selection.
University of California Transporta-tion Center (1995).[7] E Gulian, AI Glendon, G Matthews, DR Davies, andLM Debney. 1990. The stress of driving: A diary study.
Work & Stress
4, 1 (1990), 7–16.[8] Isaac Johnson, Jessica Henderson, Caitlin Perry, Jo-hannes Schöning, and Brent Hecht. Beautiful...butat What Cost? An Examination of Externalities in Ge-ographic Vehicle Routing. In
Submitted to UbiComp’17: Proceedings of the 2017 ACM International JointConference on Pervasive and Ubiquitous Computing .[9] Ashish Kapoor, Selene Mota, Rosalind W Picard, andothers. 2001. Towards a learning companion that rec-ognizes affect. In
AAAI Fall symposium . 2–4. [10] Julia Letchner, John Krumm, and Eric Horvitz. 2006.Trip router with individualized preferences (trip): In-corporating personalization into route planning. In
Proceedings of the National Conference on ArtificialIntelligence , Vol. 21. Menlo Park, CA; Cambridge, MA;London; AAAI Press; MIT Press; 1999, 1795.[11] YiRu Li, Sarah George, Craig Apfelbeck, Ab-deltawab M. Hendawi, David Hazel, Ankur Teredesai,and Mohamed Ali. 2014. Routing Service with RealWorld Severe Weather. In
Proceedings of the 22NdACM SIGSPATIAL International Conference on Ad-vances in Geographic Information Systems (SIGSPA-TIAL ’14) . ACM, New York, NY, USA, 585–588.
DOI: http://dx.doi.org/10.1145/2666310.2666375 [12] Rosalind W Picard and Roalind Picard. 1997.
Affectivecomputing . Vol. 252. MIT press Cambridge.[13] J. Posner, J. A. Russell, and B. S. Peterson. 2005. Thecircumplex model of affect: an integrative approachto affective neuroscience, cognitive development, andpsychopathology.
Dev. Psychopathol.
17, 3 (2005),715–734.[14] Nina Runge, Pavel Samsonov, Donald Degraen, andJohannes Schöning. 2016. No more autobahn!:Scenic route generation using googles street view.In
Proceedings of the 21st International Conference onIntelligent User Interfaces . ACM, 147–151.[15] Sumit Shah, Fenye Bao, Chang-Tien Lu, and Ing-RayChen. 2011. CROWDSAFE: Crowd Sourcing of CrimeIncidents and Safe Routing on Mobile Devices. In
Pro-ceedings of the 19th ACM SIGSPATIAL InternationalConference on Advances in Geographic InformationSystems (GIS ’11) . ACM, New York, NY, USA, 521–524.