SENSPEED: SENSING G DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENV VIORNMENTS IEEE CONFERENCE PAPER REVIW CSC 8251 ZHIBO WANG
EXECUTIVE SUMMARY Brief Introduction of SenSpeed Basic Idea of Vehicle Speed Estimation Design of SenSpeed Results SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 2
Brief Introduction of SenSpeed Basic Idea of Vehicl e Speed Estimation Design of SenSpeed Results 3
WHAT IS SENSPEED Smartphone sensor designing Uses two kinds of sensors in smartphone accelerometers and gyroscopes. Natural driving conditions sensing Identifies unique reference points from natural driving conditions include making turns, stopping, and passing through uneven road surfaces. Vhil Vehicle speed estimating i Estimates instant vehicle speed utilizing reference points to qualify acceleration error and eliminate accumulative error. SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 4
WHY VEHICLE SPEED IS ESSENTIAL Analyzes increasing complex urban traffic flows Facilitates more intelligent driving experie ence including vehicle localization Enhances driving safety and driving behavior analysis Builds intelligent transportation system SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 5
Brief Introduction of SenSpeed Basic Idea of Vehicl e Speed Estimation Design of SenSpeed Results 6
BASIC IDEA FOR VEHIC LE SPEED ESTIMATE USING SMARTPHONE SENSORS Fig. 1. Illustration of vehicle coordination system and smartphone coordination system SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 7
BASIC IDEA FOR VEHIC LE SPEED ESTIMATE USING SMARTPHONE SENSORS (CONT D) SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 8
CHALLENGE OF THE IDEA Noise from sensor readings causes serious errors Engine vibrations Phone usage (driver or passenger picks up the phone and put it back) White noises Etc. Errors accumulated affects integral results over time See Fig. 2. example 9
CHALLENGE OF THE IDEA (CONT D) Fig. 2. The true speed, integral value of the accelerometer s reading and their difference in real driving environment SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 10
11
Brief Introductio n of SenSpeed Basic Idea of Vehicl e Speed Estimation Design of SenSpeed Results 12
OVERVIEW OF SENSPEED SYSTEM Identifies 3 types of reference points Sensors 3 kinds of natural driving conditions making turns, stopping, and passing through uneven road surface Updates speed estimation more frequently between two adjacent reference points Eliminates accumulative errors Proposes an online algorithm to achieve real-time speed estimation Uses exponential moving average giving more weight to latest data Considers other practical issues Reorients the coordinate systems Allows usage of phone Acquires the wheelbase information 13
OVERVIEW OF SENSPEED SYSTEM (CONT D) Fig. 3. SenSpeed System Architecture SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 14
SENSORING REFERENCE POINTS TURNS Fig. 4. Illustration of circular movement when a car makes a turn 15
SENSORING REFERENC CE POINTS TURNS (CONT D) Fig. 4. Illustration of circular movement when a car makes a turn 16
SENSORING REFERENCE POINTS STOPS Exact speed 0 can be obtained Stopping reference points determination Reading of accelerometer s z-axis is steady Standard deviation of acceleration on z-axis remains low Standard deviation of the acceleration on z-axis can be used to detect stopping reference points Fig. 5. Illustration of acceleration and standard deviation on z-axis when vehicle stops 17
SENSORING REFERENC CE POINTS UNEVEN ROAD SURFACES Fig. 6. Illustration of acceleration and auto- correlation results on z-axis when vehicle passes a speed bump. 18
INSTANT VEHICLE SPEED ESTIMATION SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 19
INSTANT VEHICLE SPEE ED ESTIMATION (CONT D) SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 20
INSTANT VEHICLE SPEE ED ESTIMATION (CONT D) SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 21
INSTANT VEHICLE SPEE ED ESTIMATION (CONT D) SENSPEED: SENSING DRIVING CONDITIONS TO ESTIMATE VEHICLE SPEED IN URBAN ENVIRONMENTS 9-Apr-17 22
Brief Introduction of SenSpeed Basic Idea of Vehicl e Speed Estimation Design of SenSpeed Results 23
OVERALL PERFORMANCE Offline speed estimation using all reference points has lowest error and better accuracy than GPS or use of one reference point under all types of roads and different period of day. Back road, Manhattan, Offline Highway, Manhattan, Offline Fig. 7. Average offline estimation error of vehicle speed in Manhattan 24
ACCURACY V.S. REFERENCE POINTS Average estimation error of using only one reference point on back road is lower than GPS. Average estimation error of using turns or stops on highway is higher than GPS. Average estimation error of using uneven n surface on all types of road is lower than GPS. 25
ACCURACY V.S. ROAD TYPE / DAY PERIOD Average estimation error on back road is lower than on highway. Average estimation error at peak time is slightly lower than at off-peak time. 26
REFERENCES H. Han J. Yu H. Zhu Y. Chen et al. SenSpeed: Sensing Driving Conditions to Estimate Vehicle Speed in Urban Environments in Proc. of INFORCOM 2014. Y. Wang J. Yang H. Liu Y. Chen et al. "Sensing vehicle dynamics for determining driver phone use " in Proc. of ACM MobiSys'13 2013. V. Cevher R. Chellappa and J. McClellan "Vehicle speed estimation using acoustic wave patterns " IEEE Trans. on Sonignal Processing vol. 57 no. 1 pp. 30-47 2009. D. Johnson and M. Trivedi "Driving style recognition using a smartphone as a sensor platform " in Proc. of IEEE Int. Conf. on Intelligent Transportation Systems 2011. 27
APPENDIX. ELIMINATING ACCUMULATIVE ERRORS 28
APPENDIX. AUTO-CORRELATION AND PEAK DETECTION 29
APPENDIX. AUTO-CORRELATION AND PEAK DETECTION 30