master thesis smartphone crash notification drivers ... · to make a commercially attractive...

2
Master’s thesis at Department of Electrical Engineering, spring 2018 Supervisors: Stefan Candefjord and Leif Sandsjö Examiner: Bengt Arne Sjöqvist 1(2) Development of a smartphone-based automatic crash notification and driving characteristics recognition system for motorcycle drivers Background The injury statistics for motorcycle drivers show that only in Sweden around 40 people are killed each year, and over 200 are severely injured, with statistics being much worse globally. Severe motorcycle crashes leading to death and incapacitating injuries do not decrease at the relatively fast pace observed for car crashes. Based on our line of research we stand before an opportunity to establish smartphones as a tool for improving the safety of motorcycle driving. One of our main goals is to evaluate the feasibility of using smartphones for implementing an emergency alarm, utilizing an algorithm that automatically can detect a crash or incident. Modern smartphones can measure acceleration (accelerometers), rotation (gyros), speed (GPS) and other parameters (e.g. magnetometers), which can be useful for assessing body movements in real-time and automatically distinguish normal driving from a crash (Figure 1). The concept is based on jalp! – an automatic emergency alarm for bicyclists (see http://jalp.se/en ). Furthermore, we are exploring the possibility to use advanced algorithms, including machine learning, to recognize relevant driving characteristics/parameters, such as driving style, road characteristics, curve handling, etc. Figure 1: Motorcycle crashes are common and often have severe consequences. A crash can be detected by using sophisticated algorithms to analyze body movements measured by the sensors in modern smartphones (picture used with permission, from https://www.flickr.com/photos/72334647@N03/11493300164). In collaboration with Chalmers School of Entrepreneurship we are now attempting to make a commercially attractive smartphone-based service for the motorcycle community, including crash detection and advanced statistics and real-time feedback of driving characteristics parameters that could be run in real-time on smartphones. The goal is to create a start-up company in the year of 2018.

Upload: others

Post on 14-Sep-2019

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Master thesis Smartphone crash notification drivers ... · to make a commercially attractive smartphone-based service for the motorcycle community, including crash detection and advanced

Master’sthesisatDepartmentofElectricalEngineering,spring2018Supervisors:StefanCandefjordandLeifSandsjöExaminer:BengtArneSjöqvist

1(2)

Development of a smartphone-based automatic crashnotificationanddrivingcharacteristicsrecognitionsystemformotorcycledrivers

BackgroundThe injury statistics for motorcycle drivers show that only in Sweden around 40peoplearekilledeachyear,andover200areseverelyinjured,withstatisticsbeingmuchworseglobally.Severemotorcyclecrashesleadingtodeathandincapacitatinginjuriesdonotdecreaseattherelativelyfastpaceobservedforcarcrashes.Basedonour lineof researchwestandbeforeanopportunity toestablish smartphonesasatool for improving the safety of motorcycle driving. One of our main goals is toevaluatethefeasibilityofusingsmartphonesforimplementinganemergencyalarm,utilizing an algorithm that automatically can detect a crash or incident. Modernsmartphones can measure acceleration (accelerometers), rotation (gyros), speed(GPS)andotherparameters(e.g.magnetometers),whichcanbeusefulforassessingbodymovements in real-time and automatically distinguishnormal driving fromacrash(Figure1).Theconceptisbasedonjalp!–anautomaticemergencyalarmforbicyclists (see http://jalp.se/en ). Furthermore,we are exploring the possibility touseadvancedalgorithms, includingmachine learning, torecognizerelevantdrivingcharacteristics/parameters, such as driving style, road characteristics, curvehandling,etc.

Figure1:Motorcyclecrashesarecommonandoftenhavesevereconsequences.Acrash

canbedetectedbyusingsophisticatedalgorithmstoanalyzebodymovementsmeasuredbythesensorsinmodernsmartphones(pictureusedwithpermission,from

https://www.flickr.com/photos/72334647@N03/11493300164).

IncollaborationwithChalmersSchoolofEntrepreneurshipwearenowattemptingtomakeacommerciallyattractivesmartphone-basedserviceforthemotorcyclecommunity,includingcrashdetectionandadvancedstatisticsandreal-timefeedbackofdrivingcharacteristicsparametersthatcouldberuninreal-timeonsmartphones.Thegoalistocreateastart-upcompanyintheyearof2018.

Page 2: Master thesis Smartphone crash notification drivers ... · to make a commercially attractive smartphone-based service for the motorcycle community, including crash detection and advanced

Master’sthesisatDepartmentofElectricalEngineering,spring2018Supervisors:StefanCandefjordandLeifSandsjöExaminer:BengtArneSjöqvist

2(2)

AimTherearethreeprimaryaimsofthisthesis:

1. Toevaluatenaturalisticdatacollectedfrommotorcycletestdrivers,anddatafrom real/simulated crashes, and assess the accuracy of existing crashdetection algorithms adapted to motorcycle, possibly compared to thestudents’owndevelopedalgorithms.

2. To develop algorithms for automatic recognition of drivingcharacteristics/parametersthatareindicatedtobeofcommercialinterestforthe motorcycle community, in collaboration with the team from ChalmersSchoolofEntrepreneurship.

3. To implementand test thealgorithmsdeveloped in the thesis (aim1and2above)inasmartphoneapp(iOSand/orAndroid), incollaborationwiththeteamfromChalmersSchoolofEntrepreneurship.

MaterialsandmethodsDatacollectionisongoingandwillcontinuethroughspringandsummerof2018.Thedataisstoredinacloud-basedserver,andnewdatarecordingsbecomeavailableassoonasfetchedfromthetestdriver´ssmartphones(expectedtobethesamedaythedrivingwasrecorded).Themain tool for analysiswill beMatlab and/or other statistical/computing software thestudent(s) favor.Fordesigning thecrashdetectionand/or thedrivingdetectionalgorithm,tools for pattern recognition such as support vectormachinesmay be used. The resultingalgorithmsshouldbeimplementedinasmartphoneappandcanbeevaluatedinfieldtestsincollaborationwiththeteamfromChalmersSchoolofEntrepreneurship.

InformationYou are welcome to do this work at MedTech West’s facilities at Sahlgrenska UniversityHospital (students should use their own computers if possible). Start date is as soon aspossible.MajorsubjectsEngineeringGroupsize1-2Prerequisites Students who have background and/or interest in traffic safety andmotorcycledriving,machinelearning/signalprocessing/statisticalanalysis,arewelcometoapply. Good programming skills (mainly using Matlab and/or other powerful computingsoftware,andsecondaryprogrammingofsmartphoneappsinAppleiOS/GoogleAndroid)isadvantageous.Supervisors

• StefanCandefjord,PhDBiomedicalEngineering,ChalmersUniversityofTechnology,073-3821537,[email protected]

• Leif Sandsjö, PhDHuman-centered technology, University of Borås, 073-4606633,[email protected]

Examiner• BengtArneSjöqvist,ProfessorofPractice,ChalmersUniversityofTechnology,

070-7877797,[email protected]