Investigation of Using Statically Mounted Laser Scanners for Situational Awareness of Complex Transportation Environments, 10-R9767

Printer Friendly Version

Principal Investigators
Paul A. Avery
Joshua J. Curtis

Inclusive Dates:  11/13/07 – 03/13/08

Background - A typical deployment of a traffic management and incident detection system utilizes some combination of closed circuit television (CCTV) video cameras and inductive loops buried just under the street surface. These technologies can be used independently to identify a variety of traffic conditions or in tandem for increased reliability. SwRI targeted the use of this technology for integration into traffic surveillance and intersection signal control, such as adjustments to the signal phase and timing (SPAT) signal of an intersection. This project sought to identify the feasibility of using a laser incident detection and ranging (LIDAR) sensor to provide similar traffic management functionality as is provided today using video cameras and inductive loops in a complex, dynamic traffic environment.

Approach - Another internal research project, the Southwest Safe Transport Initiative (SSTI), allowed SwRI to acquire three LIDAR sensors from the sensor OEM. These sensors were chosen for the SSTI program for their post-processing ability. The rear sensor from the SSTI vehicle was removed and mounted in a specially-built stand and oriented toward the main four-way stop located on the SwRI grounds. Data was collected and algorithms were developed in C++ for the RTMaps rapid-prototyping software architecture to process the scan and object data returned from the sensor for analysis within a traffic management framework. The algorithms developed as part of this project use a reinforcing feedback dynamic to verify the presence of vehicles, vehicle lanes (queues), and the intersection itself. This enables the algorithm to learn the intersection configuration and dynamics without any prior information and then to begin observing the intersection from a traffic management perspective where vehicles in queues are counted and tracked for the duration of their time in the queue.

Accomplishments - The results of this work show that LIDAR sensor technology is a feasible alternative to video cameras and buried inductive loops for traffic analysis. The investigators also demonstrated the feasibility of using a LIDAR sensor to dynamically learn the configuration of a generic intersection based on the presence and behavior of vehicles; however, the algorithms were very sensitive to the orientation of the sensor. The algorithm begins by attempting to classify objects as vehicles. Once vehicles are identified, a vehicle queue is established, and each time a new vehicle is identified in the vicinity of an existing queue, its confidence level is increased. Depending on the rate at which vehicles are observed, the algorithm will begin to identify vehicle queues in approximately 30 to 60 seconds and will identify the location of the intersection within about two minutes. The results of this project were presented to technical and executive staff at the sensor OEM's headquarters in April 2008. The results were well-received, and tentative verbal agreements were made to work together with SwRI in the future to enhance the sensor's capabilities and market applicability.

2008 Program Home