Using LiDAR Data to Create Intelligent Freeways
The Dallas-Fort Worth area in Texas has been one of the fastest growing areas in the United States. In the decade from 1990 to 2000, the population of this area grew by 29.9 percent, bringing it to 6.3 million people, the fourth most populous region in the country. With such rapid growth comes long, slow commutes in heavily congested traffic and the bane of most boomtowns: gridlock.
In an effort to ease gridlock and reduce traffic congestion, high occupancy vehicle (HOV) lanes have been introduced on some of the major highways of the Dallas-Fort Worth area. The Texas Department of Transportation (TxDOT) reports that 314 HOV locations existed throughout the U.S. as of 2007. But not all HOV lanes are created equal; designs can vary from highway to highway.
In some areas, HOV lanes are demarcated by white stripes painted on the road surface. Others use a curb that separates the HOV lane from the regular lanes, pylons placed in regular intervals between the lanes, or concrete barriers that restrict access to the HOV lane. The HOV lane design depends on the type of highway, the volume and mix of traffic (car-to-truck ratio) and the particulars of the local geography.
All these variables mean that there is no “one-size-fits-all” HOV lane design. They also make HOV lane design, construction and maintenance an ideal place to apply LiDAR data.
One of the primary concerns with HOV lanes is commuter safety. Studies examining accident rates before and after the construction of HOV lanes in an effort to quantify the impact on traffic flow and safety have provided conflicting results. For example, after HOV lanes were introduced to sections of I-35E and I-635 in north Texas, a safety review carried out by the Texas Transportation Institute under the auspices of TxDOT demonstrated that the vehicle crash rate increased by 56 percent on I-35E and by 41 percent on I-635 in the first year of operation. A regional review of the HOV lanes stated that the increased rates were attributed to “speed differential and subsequent access, either exiting or entering, between HOV lanes and general purpose lanes.”
A comparison of crash rates on I-35E and I-635 with those on route U.S. 75 indicated that 27 percent fewer crashes occur on U.S. 75. The crash rate reduction is attributed to differences in the HOV design and prevailing road conditions. HOV access points on the section of U.S. 75 were located in areas where the speed differential is not significant. Also, the U.S. 75 HOV lanes use a buffer zone combination of curb and pylons that “provides a forgiving traffic control device” that discourages drivers from “arbitrarily crossing the double white lines,” as noted in the review.
Along with safety, the other key concern with HOV lanes is the question of enforcement. How can transportation authorities ensure that HOV lanes are not misused by vehicles carrying less than the required minimum number of occupants? The Dallas-Fort Worth regional review of HOV lanes noted, “A safe and effective HOV system requires regular enforcement of HOV lane rules. Enforcement is critical to discouraging illegal crossing of the pylon barrier and/or white line buffers.”
In the Dallas-Fort Worth region, the Dallas Area Rapid Transit (DART) was appointed to operate and maintain the HOV lanes, with responsibilities that include enforcing the minimal-occupant-per-vehicle requirement. One HOV lane that opened in 2009 on I-30 in Dallas started as westbound-only, but plans called for its conversion into a reversible managed-lane system later. In this plan, single-occupant vehicles will be allowed access to the HOV but will be required to pay a higher toll than vehicles carrying more occupants. This HOV was built with separation barriers and relies on local law enforcement agencies to police the minimum occupancy requirement.
As the designs and uses of HOV lanes become more complex, the challenges in operating, maintaining and enforcing proper use increase. Clearly, the increase in vehicle crash rates on I-35E and I-635 was an unintended consequence of the HOV lane designs. Concerns have also been raised about relying on local police agencies to enforce HOV regulations. Issues over manpower availability, safety, labor costs and the priority of assigning personnel to HOV enforcement instead of more serious crimes all cloud the picture. Furthermore, there is a question of whether police presence deters HOV lane misuse only as long as it is visible.
Recognizing these uncertainties, both DART and TxDOT continue to investigate automated enforcement solutions such as mounted cameras and other sensors to monitor, detect, record and report HOV violations. Known as intelligent transportation systems, these solutions promise to increase both enforcement and safety. However, the placement of the cameras is important to maximize the efficiency and capabilities of the system.
It is theoretically possible to use highway design drawings and plans when planning installation sites for camera monitors, but 2D files never perfectly represent as-built infrastructure and can’t display line-of-sight information. Accurately georeferenced 3D point cloud data captured with LiDAR technology reveals infinitely more about the actual site. Having such information allows designers to optimize HOV lane design and determine the best location for traffic cameras to obtain the optimal field-of-view.
A case in point can be found in a recent project in Arlington, Texas, near Dallas-Fort Worth. In April 2011, Woolpert Inc. conducted a survey on part of IH-30. The Texas Department of Transportation (TxDOT) was planning to install intelligent high occupancy vehicle (IHOV) lanes on IH-30 from the Trinity River West to N. Davis Drive in Arlington, and wanted a new, updated surface. Woolpert deployed Optech’s Lynx V200 Mobile Mapper to acquire and process approximately 140 lane miles of data for TxDOT’s Dallas-Fort Worth District. The LiDAR data was used to produce and provide a highly accurate surface along the roadway from the edge of pavement (EP) to EP.
To facilitate complete right of way (ROW) to ROW coverage, the mobile mapping system (MMS) dataset was merged with existing autocorrelated surface (ACS) data collected previously by the civil engineering firm Bohannan Huston based on aerial photography obtained in 2007. Autocorrelated surfaces are created from similar points between multiple image pairs in a photogrammetric process. These points then become XYZ points, typically based on a user-defined grid spacing. An ACS can be thought of as a very low point density surface similar to what might be derived from aerial LiDAR data, except that it does not provide a true ground surface where there is vegetation. To tighten up the accuracy of an ACS, a technician can manually add breakline data.
The ACS had previously been determined as suitable to produce 2-foot contours. Merging surfaces is not difficult and can be done in a commercial mapping/CAD workflow. Bentley MicroStation software is typically used when working on TxDOT projects and was used by Woolpert for this project, as well. The most current version of MicroStation effectively handles the large LiDAR datasets.
Merging the current mobile LiDAR data with the previous ACS data allowed TxDOT to maximize the return on its 2007 investment in acquiring the ACS while allowing the team to gather new, extremely high-accuracy data (+/- .06 foot at 1 sigma) where it most applied to the expected IHOV design work. In addition, new aerial photography was captured and orthorectified to the merged surface to completely cover the project from ROW to ROW.
On Woolpert’s IH-30 survey, the first step was to set up the drive path based on the width of the data collection area. Once this path was established, planners could determine where reflective reference targets should be placed. Before the survey crew went into the field to set up the physical targets, a preliminary drive-through of the route was done to note all factors that could affect the safety and efficiency of the survey crew and to identify obstacles that could interfere with GPS signal reception during data acquisition.
After this “dry run,” the path plan was set up in KML, the file format used by Google to display geographic information. Once the path plan was determined, the surveyors knew exactly where they needed to place the reflective targets. Approximately 500 reflective targets were set up at the edge-of-pavement in each direction of the traffic flow as well as in the center of some HOV lanes. The targets were spaced approximately 600 to 1,200 feet apart, depending on potential GPS interference. It took three two-person crews around four days to set the targets.
Once the targets were set up, the survey crew ran levels through the targets to establish the project’s control. They carried out the mobile survey both in daylight and at night. Because the LiDAR sensor is active (i.e., the laser provides the light), the point cloud data was easily captured at night. The photographic imagery, however, had to be acquired during daylight hours.
The survey vehicle drove the path plan at normal highway speeds and spent about five four-hour days to cover approximately 140 lane miles of highway. Delays were intermittently caused by heavy rains, but when the weather cleared, the survey proceeded on schedule.
Approximately one month was spent processing the Lynx data, which included geo-referenced XYZ point cloud and intensity data. The team established the smoothed best estimate trajectory (SBET) through differential GPS processing, in which GPS data from ground control stations is co-processed with the position info from the Lynx’s onboard GPS receiver. Position information was further refined by processing the Lynx vehicle’s dynamic movements (heading, roll and pitch) during the survey. When processed with ancillary data from the Lynx’s Distance Measurement Instrument (DMI), the position accuracy was refined to the 0.1-foot level achieved in the final results. Extracting features occurred as processed LiDAR data was delivered to the extraction team. Orthorectification of aerial photogrammetry took about a week and was completed concurrently.
To confirm the accuracy of Woolpert’s mobile LiDAR survey, TxDOT used a truck with two GPS units mounted on the cab. This quality control vehicle drove three passes at highway speeds along the 18-mile project area: eastbound, westbound and the HOV lane. A driver and a technician collected data from two Trimble R8 GNSS receivers and two Trimble TSC3 data collectors, gathering data simultaneously at 500-foot intervals. Using two GPS units at the same time provides a check of one receiver against the other. This process produces higher accuracy for checking the surface.
Next, TxDOT imported the GPS points into CAICE software to model the surface from the vehicle to check the accuracy of the mobile mapper. The DTM created from the mobile mapper was overlaid on a DTM created from the vehicle. Then, the void area from the two DTM surfaces was extracted. TxDOT found a root mean square difference of no greater than .08 foot.
TxDOT then created 1-foot contour intervals from the vehicle GPS and overlaid these on the mobile contours to illustrate the accuracy of the data. GPS shots were also captured to check targets that Woolpert set. TxDOT tied these directly to the VRS system and found a maximum accuracy error of .03 foot.
Deliverables included LAS files, TINs and KML files. As part of Woolpert’s innovative QA/QC procedure, photographic imagery was captured during the mobile LiDAR data acquisition using Lynx’s two 5-megapixel cameras, which, in Woolpert’s customized configuration, were aimed in both the forward and rearward directions. These photographic images are a deliverable as well as an aid in Woolpert’s extraction phase process, which incorporates TopoDOT and other software tools in the workflow. Instead of simply providing these images as an indexed file of jpegs, Woolpert adds value by providing the images as a “Google Street View” style dataset accessed through a KML viewer in Google Earth. Clients appreciate seeing their project imagery vividly displayed in Google Earth in a “Street View” stream. The most exciting part is seeing all of the information gained from one collection effort: imagery, orthos and, of course, point cloud data. Serving this information will be the next big leap in the industry.
The project was completed on schedule, and checks against the Lynx Mobile Mapper data averages yielded an RMSE .01 foot. What’s more, the data clearly shows everything in the surrounding area of interest, including guard rails, pavement markings, signage, bollards, pylons, lamp standards (on which camera monitors are often installed), overhead power lines, over- and underpasses, bridges, berms, retaining walls--all in a permanently searchable, editable database, available for manipulation, analysis and as a baseline for project monitoring via change detection comparisons. It’s an indispensable tool for quickly and safely providing the kind of data essential to forward-thinking highway engineering projects--particularly intelligent high occupancy vehicle lane design and construction.