Generalized LRS Approach
for Interoperability of As-Is Legacy Data
Dr. Paul Scarponcini
Bentley Transportation
66 Willowleaf Dr.
Littleton, CO 80127
paul.scarponcini@bentley.com
Abstract
It is time for a fresh approach for how we look at Linear Referencing. On the one hand, we have software products which are too simplistic, supporting only a single LRM which itself has proven to be too unstable and too limiting as a foundation for attribute data. On the other hand, we have theoretical approaches which are too complex for comprehension, let alone implementation. Though technically sound, they do not address the practical considerations of software, process and data migration costs. Most DOTs have vast amounts of data tied to multiple LRMs with associated business processes for utilizing this data. Even if they could be convinced that there is a consummate model, they would be reluctant to convert all of their legacy data and processes. Instead, what we need is a generalized approach which is both simple enough to understand and robust enough to support LRM interoperability using legacy data in its as-is condition.
The Generalized Approach is a way of looking at seemingly disparate LRMs from a common view. What is common is that every LRM has as its basis a linear element against which multiple attribute values may be referenced. What is different is the linear element itself. In a Route/Reference LRM, it is a route; in Link/Node it is a link. By mapping linear elements from different LRMs to each other or to a user preferred set, referenced attributes can be correlated. Carrying this broad generalization a step further, attributes could be stored directly against anchor sections and topological links, simplifying the NCHRP model. Engineering design data referenced to survey alignments could also be included. The Generalized Approach enables interoperability across existing and emerging LRMs, is compatible with an open, accessible data base foundation approach, and supports legacy data as it currently exists, minimizing data and process conversion.
Introduction
When viewed from a life cycle perspective, the management of transportation facilities can be divided into the six phases of planning, design, construction, operation, maintenance, and retirement. Political and technology barriers exist between these phases (see Figure 1), restricting the flow of information between them.
The Transportation Research Board (TRB) believes that "data integration across application areas is an urgent, long-standing need of DOTs."[NCHRP 1993]. To fully support an integrated facility life cycle approach, the flow of information should be in four directions (see Figure 2):
The TRB goes on to state that "the concept of location … can serve as an integrative concept across a wide variety of data, both geographic and of other kinds." This would be trivial except for the fact that, in most DOTs, there is no single, consistent way of describing locations.
The National Cooperative Highway Research Program (NCHRP) has defined a Linear Referencing Method (LRM) as a "way to identify a specific location with respect to a known point." [NCHRP 1974] Numerous LRMs exist, even within a single DOT. They further define a Linear Referencing System (LRS) as "a set of office and field [policies, records, and] procedures that includes … highway reference method[s]" as well as the transformations between these methods.
NCHRP 20-27(2) Conceptual Model
NCHRP Project 20-27(2) provided a five-level LRS conceptual model which handles multiple LRMs (see Figure 3) [NCHRP 1997]. In this model, business data (e.g., an accident, a pavement type) is characterized as events occurring on a roadway. Events are located against the LRM which is mapped onto a topological network which is itself mapped onto a linear datum. Various cartographic representations can also be mapped onto the linear datum.
The linear datum is comprised of anchor points and anchor sections. Anchor points represent well known, unambiguous point locations in the real world. Anchor sections connect pairs of anchor points and are representative of the roadway segment which exists between the two real world locations. The only attribute associated with an anchor section is its length. The linear datum provides a (possibly) incomplete link/node network, as anchor sections are permitted to cross without an intervening anchor point (see Figure 4).
Topological networks are comprised of nodes and the links which connect them. Nodes need not occur at anchor point locations. In order to associate a network with the linear datum, nodes are given an offset distance along an anchor section. In the case where crossing anchor sections do not have an explicit anchor point of intersection, the corresponding intersection node would be referenced against both anchor sections.
The next level in the NCHRP model is comprised of any number of LRMs. Because of different terminology associated with each, the term traversal was chosen to describe the LRM backbone. This is the object along which locations are measured, e.g., route. A traversal is mapped to one or more network links and is therefore only connected transitively to the linear datum.
Events can be pointil or linear; pointil events occur at a single point location (e.g., accident) whereas linear events occur across an interval bounded by two point locations on a traversal (e.g., asphalt pavement). When associated with an LRM, these point locations are expressed as traversal reference points on a traversal. The same event could have its same point location expressed as two different traversal reference points, each for a different LRM. Locating an event on the linear datum involves projecting the event location onto the LRM, then network link, and then anchor section.
A cartographic representation (e.g. GIS map) is comprised of lines. There are no constraints on where lines begin and end, allowing discontinuities (though careless overshoots and undershoots are discouraged). Break points do not have to occur at anchor point, network node, or traversal reference point locations. Instead, lines are located against offsets along anchor sections. The model thereby enables the transitive projection of events onto lines through the LRM, network, and datum levels.
NCHRP 20-27(2) provides a good theoretical model. It helps to standardize terminology which is always welcomed in a new technology area. It stabilizes the LRS backbone with the concept of a linear datum instead of the current practice of using a volatile route-reference system. The model accommodates multiple network and multiple graphic representations which is essential. However, it begs a simpler yet flexible implementation model.
Linear Reference Methods
Again, an LRM is "way to identify a specific location with respect to a known point." We interpret this to mean the method of measuring along some type of linear element. Unfortunately, the manner in which this offset distance is expressed varies with each LRM. In addition to the method of measurement, a Linear Reference Method also specifies the type of linear elements used. The actual linear element instances along which the measurement is taken is purposely de-coupled from the LRM.
Several LRMs use absolute distances to describe locations; the distance is measured from the beginning of the linear element used by the LRM. Figure 5 shows how an event occurring at the midpoint of a 5 mile long linear element would be located by various absolute distance schemes. A milepoint LRM would measure the event location as 2.5 miles from the start. The metric equivalent meterpoint distance would be 4 kilometers, again measured from the start of the linear element. A percentage LRM would measure this location as 50%, i.e., 50% of the way from the start to the end of the linear element.
Figure 6 provides three LRMs which utilize relative distances. The location of an event is described as an offset distance from a pre-defined point along the linear element; the reference point need not be at the beginning of the linear element. A milepost LRM measures distance from the closest preceding milepost. Mileposts are located at one mile increments along the linear element. The event specified above would be at 2 + .50 miles, i.e., .50 miles beyond milepost 2. Milepost 2 is exactly 2.0 miles from the start of the linear element.
Reconstruction can alter the alignment of a roadway, thereby impacting the length of the linear element used to represent the roadway. Instead of relocating all subsequent mileposts to maintain their one mile interval, the mileposts can be considered to be reference posts. Reference posts provide relative measurement like mileposts but without the one mile constraint on inter-post distances. If reference post 1 is 1.2 miles from the start of the linear element, and reference post 2 is an additional 0.9 miles, then the event location would be 0.4 miles from reference post 2. This would be expressed as 2 + .400.
A county boundary may act as a reference point for defining relative distances. If the county (C) boundary crosses the roadway at a point 0.7 miles from the start of the roadway, the County milepoint distance of the event above would be © + 1.8 miles. Of course County boundaries may change and there is the difficulty of pinpointing the location of a County boundary in the field.
Figure 7 shows three additional LRMs. Stationing is a method of measurement typically used by surveyors and engineers during roadway design and construction. Here distances are measured in feet from some start point. The distance is expressed as if it were offset from posts 100’ apart. If the measuring starts at 0+00 at the start of the linear element, the event occurs at 132+00, i.e., 13,200 feet (2.5 miles) along the linear element.
There may be discontinuities in the stationing. For example, stationing could go for the first mile from 0+00 at the start of the linear element to 52+80. Stationing is then resumed from this point, but beginning with 10+00. To get to the event, another 1.5 miles (7,920 feet) is required. The 7920 feet from 10+00 results in a station of ‘A’ 89+20. The ‘A’ signifies a station ahead of the station transition point. The equivalence of ‘B’ 52+80 and ‘A’ 10+00 is called a station equation.
In urban areas, street addresses can be used to approximate distances. This is based on a linear proration of address numbers between the first and last address number on a street block. If the linear element (street block) begins at address 20 and ends with address 90, then the event occurring midway along the block would be approximately in front of the building at address 55.
Other distance measurement schemes exist, representing additional LRMs. There may also be slight departures from the ones mentioned above.
Distance Expression
We shall call the way in which distances are expressed distance expressions (DX). In the nine LRM examples above, nine DX’s describe the same event location:
2.5 miles 2 + .50 miles 132 + 00
4 km 2 + .400 ‘A’ 89 + 20
50% © + 1.8 miles 55
The LRM provides a context for interpreting a DX.
Linear Elements
A linear element (LE) is any one-dimensional object along which locations can be specified. The distance denoted by a distance expression is measured along a linear element. In addition to the method of measurement, a Linear Reference Method specifies the type of linear elements used. Figure 8 demonstrates several common linear element types.
Routes are the most common type of linear element, despite the fact that they are so unsuitable as a foundation element; in addition to changes to the underlying roadway alignment, routes may be re-designated, moved to another roadway, or share roadways with other routes. Routes may have mileposts, reference posts, or county boundaries associated with them to enable relative distance LRMs.
A less precise linear element type is street. A street it comprised of one or more street blocks, demarcated by intersecting streets. Associated with each street block is the minimum and maximum street address for the block for each side of the block. The assumption is that there is uniform distribution of addresses along the street block.
The design/construction survey alignment is a less traditional linear element type, as most LRMs have arisen either from the planning, operation or maintenance phases of the transportation facility lifecycle. However, a considerable amount of design / construct information is stored against alignments which would be useful to combine with other LRM-based information during the other life cycle phases.
The NCHRP conceptual model disallows direct assignment of event locations to network links, linear datum anchor sections, and cartographic representation lines. However, many DOTs either already have or plan to implement LRSs which include links, anchor sections, and/or lines as linear element types. In addition to functioning as linear elements, links also have start and end nodes to support network topology ; anchor sections have start and end anchor points of known location in the real world, and lines have geometry and location attributes.
Figure 9 shows the most common usage of measurement methods for each of these linear element types. Milepoint, meterpoint, milepost, reference post, and County milepoint are common ways to measure along a route. Though percentage measurement is possible, it is not as prevalent as the others, probably because of the long length of some routes. Length also poses a problem for the other absolute distances as well as a milepost relative distance. Changes in route length arising from re-alignment near the start of a route must be propagated to all event locations for the entire route.
Links, lines, and anchor sections typically use percentage or mile/meterpoint measurement. For lines, mile/meterpoint measurement is scale dependent and graphic precision dependent. Generally, a true roadway length is stored with a line in addition to its graphic length to enable pro-ration of event locations. Alignments are typically stationed. Streets are typically addressed.
Location Expressions (LX)
Events can occur at a single location or throughout an interval specified by a start and end location. Because of the variety of distance expressions expounded above, a more robust location expression is required. This expression provides the context for DX by including the LRM used, as well as the linear element against which the DX is measured:
LX = ( LRM, LE, DX )
where LX is a linear expression comprised of the three-tuple of linear referencing method (LRM), linear element (LE), and distance expression (DX). A linear element instance includes requisite adjunct information, such as the location of reference posts along a route. Recognize that this a logical specification for LX; a physical implementation would not require storing the LRM with each individual event location.
Expressing the previous nine examples of distance expressions within the broader context of LRM and LE would result in the following nine location expressions:
These nine location expressions specify the same event location. Distance expressions are now disambiguated by the inclusion of the LRM. The actual linear element instance is also included.
Generalized Model
As a result of exploring similarities in structure of apparently disparate LRMS, we can now propose a simplification to the NCHRP 20-27(2) conceptual model. This logical model is referred to as the Generalized Model.
Recall the five-level NCHRP model shown in Figure 3. The lower four levels are all comprised of linear elements, based upon the definitions and examples supplied above (see Figure 10). LRMs use traversals, networks have links, the linear datum has anchor sections, and the cartographic representation has lines. If the constraint against locating events directly on a link, anchor section, or line is relaxed, the NCHRP model can be compressed into the two-level model shown in Figure 11.
The Generalized Model has the following characteristics:
There are several departures from the NCHRP model. The Generalized Model has been reduced to two levels by realizing similarities within and between each of the four lower levels of the NCHRP model. Networks are not required as an intermediary between event locations and the linear datum, as long as network tracing is not required or if the linear datum is complete with respect to connectivity. The Network, Linear Datum, and Cartographic Representation levels become LRMs, with linear elements which can directly support event locations. Though not mandatory, the linear datum LRM is still recommended in order to simplify the translation between multiple LRMs. The Generalized Model does not preclude a literal, five-level implementation of the NCHRP model.
The Generalized model has the following advantages:
Summary
A Generalized Method for the linear referencing of event locations is proposed in support of transportation facility life cycle integration. The simple, flexible model supports legacy data as-is, without mandating change. The solution is an NCHRP 20-27(2) compatible logical model.
References
NCHRP 1974: Highway Location Reference Methods, Synthesis of Highway Practice 21, National Academy of Sciences, Washington, D.C.
NCHRP 1993: Adaptation of Geographic Information Systems for Transportation, NCHRP Report 359, National Academy Press, Washington, D.C.
NCHRP 1997: A Generic Data Model for Linear Referencing Systems, NCHRP Research Results Digest #218, Transportation Research Board, Washington, D.C.
NCHRP 1999: Functional Requirements for a MultiModal MultiDimensional Location Referencing System Data Model (Work in Progress), Draft Synthesis Report, NCHRP 20-27(3), Transportation Research Board, Washington, D.C.
Figure 1. Existing Transportation Facility Life Cycle
Figure 2. Integrated Transportation Facility Life Cycle
© 1998 Bentley Systems, Inc.
Figure 3. NCHRP LRS Data Model Conceptual Overview
Figure 4. NCHRP Mappings
Figure 5. LRMs with Absolute Distance
Figure 6. LRMs with Relative Distance
Figure 7. LRMs with Other Distances
Figure 8. Linear Elements
Figure 9. LE – LRM Usage
Figure 10. NCHRP Model Generalization
Figure 11. Generalized Model