ANALYZING THE EFFECTS OF GEOMETRIC LANE CONSTRAINTS ON RADAR-BASED SENSING OF AVAILABLE VEHICLE HEADWAY USING MAPPED LANE GEOMETRY AND CAMERA REGISTRATION OF LANE POSITION
Open Access
- Author:
- Varadarajan Srinivasan, Krishna Prasad
- Graduate Program:
- Mechanical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 07, 2019
- Committee Members:
- Sean N Brennan, Thesis Advisor/Co-Advisor
- Keywords:
- RADAR-based sensing
Connected and Autonomous Vehicles
Headway estimation
Available free space in ego-lane
free space estimation for fuel economy
sensor fusion - Abstract:
- Commercial trucks are currently equipped with a single front-facing RADAR mounted on the front bumper, as this is a sensor useful for Cooperative Adaptive Cruise Control, Emergency Braking Systems, and many similar Connected and Automated vehicle functions that require longitudinal vehicle control. This paper investigates the use of a bumper-mounted RADAR to perform traffic characterization around the ego-vehicle, with a particular goal to obtain an an estimate of the furthest headway that can be considered as a reliable estimate of open maneuvering space, such that there are no vehicles within the same lane as the ego-vehicle, i.e. the ego-lane. This available headway in front of a vehicle is an important parameter in an ongoing study whose goal is to obtain improvements in fuel economy for highway driving of a tractor-trailer. But head- way availability depends on the correct attribution of RADAR-measured vehicles to be either within the ego-lane, or outside the lane. The attribution of lane designations to specific RADAR targets depend strongly on lane geometry and the ability to align RADAR measurements to the ego-lane. This work investigates how knowledge of lane geometries, as well as sensor performance characteristics, may improve the trust in a RADAR measurement of open headway distance in front of a vehicle. Specifically, several strategies for associating local traffic either within or excluded from the ego lane are considered, and the possible sources of error in headway calculations are investigated for each strategy. The strategies differ by the availability of lane geometry information and RADAR target association to lane constraints, which in turn is assumed to be supplied by the use of map data and secondary camera measurements. It is observed that combining RADAR with lane-geometry maps help detect headway up to and beyond RADAR range limit with exceptions tied to a particular road. These errors can further be minimized by using camera measurements subject to accuracies in registering lane markers. The results of analyzing highway routes reveals that: RADAR data without lane geometry strongly limits the trusted headway measurement from this sensor; that lane geometry maps can greatly increase the trusted range, usually beyond the RADAR’s range limits; and that image-based registration of lane markers to lane geometry should have accuracies allowing full-range trust in the RADAR target measurements of headway.