DETERMINING THE UNCERTAINTY OF A GPS-BASED COLLISION VEHICLE DETECTION SYSTEM

Open Access
Author:
Amin, Sanket R
Graduate Program:
Mechanical Engineering
Degree:
Master of Engineering
Document Type:
Master Thesis
Date of Defense:
July 15, 2011
Committee Members:
  • Sean N Brennan, Thesis Advisor
  • Sean Brennan, Thesis Advisor
Keywords:
  • Allan Variance
  • sensitivity analysis
  • LIDAR
  • collision warning system
  • GPS
  • friction coefficient estimation
Abstract:
Automotive manufacturers are researching forward collision warning systems (FCWS) to reduce the occurrence of rear-end collision accidents between vehicles. Traditionally these systems use forward scanning sensor technology such as RADAR or LIDAR to measure the distance between the equipped vehicle and other vehicles/obstacles ahead. The U.S. Army is using such technology on their ground vehicles but has noticed its performance is sometimes compromised due to environmental effects (caking of debris on sensors). This thesis presents the work of developing a FCWS that instead uses Global Position Satellite (GPS) technology and the available information associated as an alternative approach for collision avoidance in convoy situations. This approach however requires a vehicle to vehicle (V2V) network infrastructure to share local GPS data among vehicles. Sponsorship from U.S. Army and Penn State’s own Applied Research Laboratory (ARL) led to the fabrication of three low-cost, embeddable prototype units that were fielded on three Army Heavy Expanded Mobility Tactical Trucks (HEMTTs) vehicles navigating through desert test courses in convoy formation. These experiments proved the feasibility of such an alternative collision detection system. The primary goal of this thesis is to evaluate how measurement errors/uncertainty affects performance of a GPS-based convoy collision avoidance system. A simple analytical framework is presented for merging system sensitivity analysis and measurement input error characterization results to determine the uncertainty in the output. The resulting metric is a dimensionless parameter corresponding to a range in the probability of collision. To test this approach, field data were analyzed and applied within the proposed framework. A secondary focus of this thesis is to address a specific concern regarding the feasibility of GPS-based collision avoidance approach due to concerns about GPS accuracy. This thesis includes identification of dominant GPS stochastic error sources using Allan Variance analysis. The research experimentally compares inter-vehicle distance accuracy, which is a core measurement of the system, between the GPS proposed approach and the traditional LIDAR-based approach in an attempt to address accuracy concerns. As vehicular communication systems such as vehicle to vehicle (V2V) emerge in the near future, a GPS-based FCWS will naturally provide a lower-cost alternative, or even supplemental, solution to the scanning technologies currently implemented. This work thus offers an immediate and substantial opportunity to save lives. While the target application of the work discussed here was for rear-end collisions such as might be encountered in military convoy operations, the solution could be adopted for the civilian commercial sector via straightforward application of existing technology.