Precision Roadway Feature Mapping Jay A. Farrell, University of California-Riverside James A. Arnold, Department of Transportation

Transcription

1 Precision Roadway Feature Mapping Jay A. Farrell, University of California-Riverside James A. Arnold, Department of Transportation February 26, 2013 ESRA Fed. GIS Outline: Big picture: Positioning and Mapping Precision Mapping Real-time Vehicle Position Estimation Project: Results and Applications Interesting Directions GPS/IMU LIDAR panoramic camera

2 Precision Automated Mapping & Positioning Automated Highway Inventory mapping and characterization enables: More efficient management of highway assets Improved availability of lane-level positioning accuracy Positioning with high degrees of: Accuracy Less than one meter Continuity High sample rate Availability Works in diverse environments Accomplished at Low Cost Enables ITS applications with transformative improvements for safety, mobility, and environmental performance Automated, sensor-based mapping for accurate and low-cost roadway map production for where-in-lane accuracy ITS Safety & Mobility Applications: Lane-departure warning, Freeway merge assistance, Curve over-speed warning, Adaptive cruise control, Congestion warning systems, Signal Phase and Timing, Lanelevel Intersection management and collision avoidance, Headlight steering into curves

3 Precision Mapping & Real-time Positioning Mapping Data Accumulation Mapping DataBase Development Application Development Vision GPS/INS Lidar Vision Sensor??? GPS/INS Raw Data Smoothing/ Feature Extraction Offline Processing Database Management Tool Application Software: Navigation + Feature Detection Feature Data Database Database Export Database

4 Precision Roadway Map: Data Acquisition sensor platform Vision Mapping Data Accumulation GPS/INS Raw Data Lidar Offline Processing GPS IMU LIDAR Smoothing/ Feature Extraction Feature Data Database Panoramic Camera

5 Mapping Process Equipment: LIDAR Camera set IMU GNSS Receiver High capacity HD Roof Platform Power supply Matlab DGPS station Data Communication GIS High Performance CPU 5

6 Driving and data collection

7 IMU and GPS data are smoothed providing a continuous sensor platform trajectory 7

8 Trajectory Estimation Solve for Integer Ambiguity Smooth trajectory using inertial measurements, pseudo-range, and Doppler measurements Use smooth trajectory to find integer intervals Extend integers from known intervals to adjacent intervals Final Trajectory Smoothing Use integers when available, otherwise, use pseudo-range and Doppler Can be Utilized Extensively in Mapping See: A. Vu et al., (2012) Improved Vehicle Trajectory and State Estimation using Raw GPS and IMU Data Smoothing, IEEE Intelligent Vehicle Symposium, Madrid Spain, June 2012.

9 Trajectory Estimation Cont.

10 Features (e.g., road signs, lane markings, road curves, etc.) are extracted and identified from the raw data 10

11 Traffic Sign Extraction

12 Plane Feature Estimation see corresponding sign numbers and normals

13 Estimated Plane Position vs. Surveyed Plane Position Sign Post # Survey North (m) Survey East (m) Max Survey ECEF Std (m) Computed North (m) Computed East (m) Abs. Horiz. Error (m)

14 Sign Identification Process

15 Feature Extraction and Identification (2) Curve Extraction Primary sensor: LIDAR 64-laser, 20 Hz laser range finder Processing step 1: Ground plane image projection from measurements Output: intensity as a function of position Processing step 2: Curve detection using projected vehicle trajectory on ground plane images Output: Curves composed of a set of sequential 3D positions in world frame Curve Parameterization Primary input: curves from previous step Processing step: Curve Fitting Output: roadway curve format suitable for GIS

16 LIDAR Intensity Point Cloud

17 Map Representation in ArcGIS Bing Map with TFHRC aerial image overlay

18 Georectified LIDAR-based intensity image

19 Bing Map, TFHRC Aerial Image, LIDAR-based intensity image overlay

20 Finding Lanes and Curves

21 Curve Fitting

22 Map Database Mapping Data Accumulation Mapping DataBase Development Application Development Vision GPS/INS Lidar Vision GPS/INS Raw Data Smoothing/ Feature Extraction Offline Processing Database Management Tool Application Software: Navigation + Feature Detection Feature Data Database Database Export Database

23 Mapping Data Accumulation Applications Mapping DataBase Development Application Development Vision GPS/INS Lidar Vision GPS/INS Raw Data Smoothing/ Feature Extraction Offline Processing Database Management Tool Application Software: Navigation + Feature Detection Feature Data Database Database Export Database

24 inexpensive GPS/IMU Sensor platform for positioning No LIDAR sensor platform inexpensive rectilinear camera No panoramic camera

25 Application Data Flow

26 Positioning: Aided Navigation High rate sensors: provide high bandwidth, continuously available vehicle state Candidate sensors: Encoder or IMU Aiding measurements: maintain vehicle state accuracy Candidate sensors: GPS, Camera, Lidar, Radar, DSRC, Signals-of-Opportunity See J.A. Farrell, Aided Navigation, McGraw-Hill, 2008.

27 Map Representation in Positioning Graphical User Interface (GUI)

28 Application Graphical User Interface

29 Lane Departure Warning

30 Curve Overspeed Warning

31 Aiding Sensor Categories GNSS: Proven with open skies. Challenging in urban environments. TRN: Shows great promise as physical layer timing advances FB: Assessed as viable if precision roadway feature maps are available. Research leading to reliable and accurate FB positioning demonstrations was a project focus.

32 Sign Aiding

33 Project Summary Automated sensor-based mapping is necessary for nationwide lane-level map production This project developed software and demonstrated automate sensor-based mapping is feasible with centimeter-level accuracy Three lane-level applications built on the foundation of lane-level maps were demonstrated using decimeter-level positioning techniques Lane departure warning Curve overspeed warning Signal Phase and Timing, at lane-level

34 Some Ideas for Future Work Positioning DSRC modem based time-of-flight positioning Direct contacts with manufacturers are in place Terrestrial RF Signals of Opportunity: DTV, Digital radio and cellular Enhance performance and robustness of feature based methods, especially RADAR and LIDAR Investigate safety & integrity monitoring for data fused vehicle applications Mapping Improved and more fully automated mapping processes Mapping additional feature types Thorough evaluation in less-structured, more-dynamic environments Maintenance of the precision map Crowd sourcing Targeted updates Large scale computer or cloud implementation to map larger environments Positioning and Mapping Thorough evaluation in less-structured, more-dynamic environments