Current map-matching algorithms for transport applications: State-of-the art and future research directions

TL;DR: The constraints and limitations of existing map matching algorithms are uncovered by an in-depth literature review and some ideas for monitoring the integrity of map-matching algorithms are presented.

Abstract: Map-matching algorithms integrate positioning data with spatial road network data (roadway centrelines) to identify the correct link on which a vehicle is travelling and to determine the location of a vehicle on a link. A map-matching algorithm could be used as a key component to improve the performance of systems that support the navigation function of intelligent transport systems (ITS). The required horizontal positioning accuracy of such ITS applications is in the range of 1 m to 40 m (95%) with relatively stringent requirements placed on integrity (quality), continuity and system availability. A number of map-matching algorithms have been developed by researchers around the world using different techniques such as topological analysis of spatial road network data, probabilistic theory, Kalman filter, fuzzy logic, and belief theory. The performances of these algorithms have improved over the years due to the application of advanced techniques in the map matching processes and improvements in the quality of both positioning and spatial road network data. However, these algorithms are not always capable of supporting ITS applications with high required navigation performance, especially in difficult and complex environments such as dense urban areas. This suggests that research should be directed at identifying any constraints and limitations of existing map matching algorithms as a prerequisite for the formulation of algorithm improvements. The objectives of this paper are thus to uncover the constraints and limitations by an in-depth literature review and to recommend ideas to address them. This paper also highlights the potential impacts of the forthcoming European Galileo system and the European Geostationary Overlay Service (EGNOS) on the performance of map matching algorithms. Although not addressed in detail, the paper also presents some ideas for monitoring the integrity of map-matching algorithms. The map-matching algorithms considered in this paper are generic and do not assume knowledge of ‘future’ information (i.e. based on either cost or time). Clearly, such data would result in relatively simple map-matching algorithms.

Summary

1 INTRODUCTION

• A range of intelligent transport system (ITS) applications and services such as route guidance, fleet management, road user charging, accident and emergency response, bus arrival information at bus stops, and location based services (LBS) require location information.
• Map matching not only enables the physical location of the vehicle to be identified but also improves the positioning accuracy if good spatial road network data are available (Ochieng et al., 2004).
• Different algorithms, however, have different strengths and weaknesses.
• First, an in-depth literature review of map matching algorithms is presented, followed by a presentation of the performance of some existing map matching algorithms.

2 LITERATURE REVIEW

• Most of the formulated algorithms utilise navigation data from GPS (or GPS integrated with deduced reckoning sensors) and digital spatial road network data.
• While this assumption is valid for most vehicles under most operating conditions, problems may be encountered for off-roadway situations such as car parks or on private land.
• Most of the studies also report that the digital spatial road network data used for map matching should be of a large scale in order to generate position outputs with fewer errors (e.g., Zhao, 1997, Quddus et al., 2006a).
• Procedures for map matching vary from those using simple search techniques (Kim et al., 1996), to those using more advanced techniques such as the use of an Extended Kalman Filter, fuzzy logic, and Belief Theory (Najjar and Bonnifait, 2005; Quddus et al., 2006b).
• The following sections briefly describe these algorithms.

2.1 Geometric Analysis

• A geometric map matching algorithm makes use of the geometric information of the spatial road network data by considering only the shape of the links (Greenfeld, 2002).
• It does not consider the way links are connected to each other.
• It is very sensitive to the way in which the spatial road network data was created and hence can have many problems in practice.
• Another geometric map matching approach is point-to-curve matching (Bernstein and Kornhauser, 1998, White et al., 2000).
• Secondly, it constructs piecewise linear curves using the vehicle’s trajectory, and determines the distance between this curve and the curve corresponding to the road network.

2.2 Topological Analysis

• In GIS, topology refers to the relationship between entities (points, lines, and polygons).
• Note however, that care must be taken in the use of vehicle heading calculated from the coordinate information of vehicle positions as GPS position fixes are less reliable when the speed is less than 3.0m/sec (Taylor et al., 2001 and Ochieng et al., 2004).
• Meng (2006) also uses a topological analysis of the road network to develop a simplified map matching algorithm.
• A number of conditional tests are applied to eliminate road segments that do not fulfil some pre-defined thresholds.
• The algorithm does not work well at junctions where the bearings of the connecting roads are not similar.

2.3 Probabilistic Map Matching Algorithms

• The probabilistic algorithm requires the definition of an elliptical or rectangular confidence region around a position fix obtained from a navigation sensor.
• The error region is then superimposed on the road network to identify a road segment on which the vehicle is travelling.
• While such criteria are conceptually beneficial, Zhao (1997) does not go into the details of their implementation.
• Ochieng et al. (2004) develop an enhanced probabilistic map matching algorithm.
• This method is more reliable as the construction of an error region in each epoch may lead to incorrect link identification if other links are close to the link on which the vehicle is on.

2.4 Advanced Map Matching Algorithms

• Some of these algorithms are briefly described below.
• As stated earlier, the point-to-curve method is not sufficient to select the correct link especially in dense urban road networks.
• The performance of this algorithm is evaluated against the performance of GPS.
• Syed and Cannon (2004) also describe a map matching algorithm based on a fuzzy logic model.
• In the first sub-algorithm, a fuzzy inference system (FIS3) is used to identify the correct link for the initial position fix.

2.5 Performance of existing map matching algorithms

• The formulations of most of the existing map matching algorithms are not accompanied by methodology for performance assessment.
• The few performances reported are captured in Table 1.
• The navigation sensors used in the algorithms, the test environments, the percentage of correct link identification, and the 2-D horizontal accuracy are shown in columns two, three, four, and five respectively.
• Very few studies state the scale of the digital spatial road network data used to estimate performance.
• As shown in Zhang et al. (2003) and Quddus et al. (2006a), the quality of spatial data can have a large impact on map matching performance.

3 CONSTRAINTS AND LIMITATIONS

• Some of the map matching algorithms discussed in the previous sections possess the capability to support the navigation module of many ITS applications and services.
• A positioning accuracy up to 5.5m (95%) is achievable within suburban areas using some algorithms.
• Among these algorithms, the fuzzy logic map matching algorithm provides the best performance both in urban and suburban areas.
• As reviewed in the previous sections, there are a number of issues that hinder the maximum exploitation of the current map matching algorithms.
• These are crystallised below, and ideas given to be explored to address them.

3.1 Problems with Initial Map Matching Processes

• Most of the map matching algorithms start with an initial matching process.
• The segments that are originated from (or are destined to) these nodes or shape points are considered as candidate segments.
• Moreover, the error circle is relatively easy to formulate compared to an error ellipse.
• This is a one-off offline process that could be performed for all road segments within the spatial digital map data.
• Therefore, the problem of identifying road segments within an error ellipse then transforms to the problem of identifying whether a line intersects a circle or not.

3.2 Problems with Threshold Values

• A variety of thresholds are used to make the correct decision during various decision-making processes within host map matching algorithms.
• The threshold for the minimum speed at which the heading of the vehicle from stand-alone GPS is incorrect is taken as 3 m/sec by Taylor et al. (2001) and Quddus et al. (2003).
• A more analytical approach may be needed to improve this value.
• Moreover, the values of various weighting parameters used in map matching algorithms can vary based on different operational environments.
• These values may be different when they are derived using data from another city and may also be dependent on the type of sensor equipment used.

3.3 Problems at Y-junction

• Given that a map matching algorithm identifies the correct link, AB, for the position fixes P1 and P2, the identification of the correct link for the fix P3 may be incorrect if the perpendicular distance from P3 to link BC and BD is almost equal, and the heading of the vehicle from the navigation sensor is 90 degrees.
• This type of hypothetical road network may be observed in motorway diverging scenarios.
• Further improvements of map matching algorithms should focus on this type of scenario.
• A positive heading change between fixes P3 and P2 implies that the vehicle is on link BD and a negative heading change implies that the vehicle is on link BC.

3.4 Consideration of Road Design Parameters in Map Matching

• Road design parameters such as turn restrictions at junctions, roadway classification (such as one-way, two-way), width of the carriageway, number of lanes, and overpass and underpass information are normally not included as inputs in the existing map matching algorithms as the data was not readily available.
• The availability of such attribute data could potentially improve the performance of map matching algorithms especially at junctions where the map matching sometimes give incorrect results.

3.5 Height Data from the Navigation Sensors

• Map matching algorithms normally do not make use of height data from a navigation sensor.
• This height data together with the data from a 3-D digital road network map can effectively identify the correct road segment at a section of roadway with fly-overs.
• This will largely depend on the accuracy of height data and the availability of a high-quality 3-D road map.

3.6 Spatial Road Network Data Quality

• The review of the literature suggests that spatial road network data have both geometric and topological errors (Noronha and Goodchild, 2000, Kim et al., 2000, Zhang et al., 2003).
• It is envisaged that the position fixes from a stand-alone GPS, specifically in an open-space environment, could be better than the map-matched positions if a poor quality map is employed in the map matching algorithm.
• Note that height data from GPS is not as accurate as horizontal positioning data.
• The accuracy (2D) of a spatial road network data (map) can be derived by a field experiments and can be incorporated in the map matching process.

3.7 Techniques used in the map matching processes

• The methods used in the map matching algorithms vary greatly from using simple search techniques to a highly mathematical approaches.
• The performance and speed of the algorithms in turn largely depend on the technique used in the algorithm.
• This may be desirable for many ITS applications.
• Therefore, the current methods introduce large errors in the location estimation, specifically the case of low resolution spatial road network data.
• A method can be developed so that the final positioning outputs from the map matching algorithm can optimally be determined anywhere within the edges of the carriageway.

3.8 Validation Issues

• Validation of a map matching algorithm is essential to derive statistics on its performance in terms of correct link identification and vehicle location determination.
• Very few existing map matching algorithms provide a meaningful validation technique.
• Kim et al. (2000) use code-based DGPS to obtain true vehicle positions.
• The performance of DGPS is strongly affected by signal multipath, among other factors, and varies according to the surrounding environments.
• The real-time availability of precise GPS satellite orbit and clock data has facilitated the development of a novel positioning technique known as Precise Point Positioning (PPP) (Heroux et al., 200; Shen and Gao, 2002).

3.9 Integrity (Level of Confidence)

• The integrity of a map matching algorithm directly reflects the level of confidence that can be placed in the map-matched position.
• The detection capability can be utilised to provide a timely warning to the driver that the position solution should not be used for navigation or positioning, and to aid the algorithms in recovery from the failure mode.
• Quddus (2006) describes a simple empirical method to derive the integrity of a map matching algorithm.
• It is essential to further investigate the performance of the integrity method with other test routes, especially in urban areas.

3.10 Implementation Issues

• It should be noted that a map matching algorithm could be implemented in two ways depending on the application: (1) an in-vehicle unit and (2) a dispatch terminal (i.e., a central server).
• In the case of implementing the algorithm in the in-vehicle unit (e.g., route guidance), each vehicle contains a map matching processor along with other navigation devices and spatial road network data.
• A communication with the central server is only required if other information (such as roadway traffic conditions) is essential for the application.
• The positioning data (easting, northing, speed, and heading) computed by these sensors are then sent to the dispatch terminal to obtain physical locations (i.e., addresses) and road related information using a suitable map matching algorithm.

4 IMPACTS OF GALILEO AND EGNOS

• The Galileo System will be an independent, global, European-controlled, satellite-based navigation system that will provide a number of services to users equipped with Galileocompatible receivers.
• Limitations on system performance and potential political considerations suggest that standalone GPS cannot always meet all the requirements of a range of ITS services and other safety-of-life (SOL) applications.
• The development of EGNOS began in the early 1990’s initiated by the Tripartite Group (ETG), comprising the European Space Agency (ESA), European Community (EC), and EUROCONTROL.
• The following sections examines the potential impacts of the forthcoming European Galileo and the EGNOS systems and how this may affect the performance of map matching algorithms and whether this will further improve the capability to support the navigation module of ITS services.

4.2 The Impact of EGNOS

• EGNOS was designed to provide potential users with the following services (Sauer, 2004).
• The following scenarios are considered to discuss the potential impact of EGNOS on the performance of map matching algorithms.
• SISNeT gives access to the corrections and the integrity information of EGNOS.
• A GPS receiver will be capable of receiving both differential correction data via SISNet and the additional data (GEO L1) directly from EGNOS.
• This will lead to a better positioning data at all times and a marginal improvement in geometry for some instances.

5 CONCLUSIONS

• The navigation function of an intelligent transport system could be supported by a map matching algorithm that integrates positioning data with spatial road network data.
• This paper has presented an in-depth literature review of map matching algorithms.
• These algorithms are not always capable of supporting the navigation module of some ITS applications such as bus priority at junctions, especially in dense urban areas.
• Therefore, to achieve the required navigation performance for some ITS services, further research and improvements to map matching algorithms are essential.
• These algorithm enhancements will be aided by the new systems EGNOS and Galileo to offer a significantly improved performance capable of supporting a wide variety of ITS services in different operational scenarios.

Figures

Figure 2: A hypothetical road network

Figure 1: No nodes within an error ellipse

Table 1: The performance of some existing map matching algorithms

Full text

This item was submitted to Loughborough’s Institutional Repository
(https://dspace.lboro.ac.uk/) by the author and is made available under the
following Creative Commons Licence conditions.
For the full text of this licence, please go to:
http://creativecommons.org/licenses/by-nc-nd/2.5/

1
Current Map Matching Algorithms for Transport Applications: State-of-
the art and Future Research Directions
Mohammed A Quddus
Transport Studies Group
Department of Civil and Building Engineering
Loughborough University
LEICESTERSHIRE LE11 3TU, United Kingdom
Email: m.a.quddus@lboro.ac.uk
; Phone: + 44 (0) 150 922 8545
Washington Y Ochieng and Robert B Noland
Centre for Transport Studies
Department of Civil and Environmental Engineering
Imperial College London
LONDON SW7 2AZ, United Kingdom
Email: w.ochieng@imperial.ac.uk
; Phone: + 44 (0) 207 594 6104

2
Current Map Matching Algorithms for Transport Applications: State-of-
the art and Future Research Directions
1 INTRODUCTION
A range of intelligent transport system (ITS) applications and services such as route guidance,
fleet management, road user charging, accident and emergency response, bus arrival
information at bus stops, and location based services (LBS) require location information. For
instance, buses equipped with a navigation system can determine their locations and send the
information back to a control centre enabling bus operators to predict the arrival of buses at
bus stops and hence improve the service level of public transport systems. The horizontal
positioning accuracy for such ITS applications is in the range of 1m to 40 m (2D positioning
accuracy at 95% of the time), with relatively high requirements on integrity, continuity and
system availability. Although most ATT services (navigation and road guidance, distance-
based road pricing etc.) requires a sample frequency of 1HZ, some ATT service (such as bus
arrival information at bus stops) only requires a sample frequency of 30 HZ or higher.
In the last few years, the Global Positioning System (GPS) has established itself as a major
positioning technology for providing locational data for ITS applications such as mobile
phones equipped with GPS for various applications. Zito et al. (2005) provide a good
overview of the use of GPS as a tool for intelligent vehicle-highway systems. Deduced
Reckoning sensors - commonly refereed to as dead-reckoning (DR) sensors - (consist of an
odometer and a gyroscope) are commonly used to bridge any gaps in GPS positioning
(Kubrak, et al., 2006). Spatial road network data are used to determine the spatial reference of
the vehicle location via a process known as map matching. For instance, the accuracy and
availability of positioning data using mobile phones will be greatly increased if the navigation
function of mobile phones is supported by GPS, DR, and spatial road network data integrated
by a map matching algorithm.
Map matching algorithms use inputs generated from positioning technologies (such as GPS or
GPS integrated with DR) and supplement this with data from a high resolution spatial road
network map to provide an enhanced positioning output. The general purpose of a map
matching algorithm is to identify the correct road segment on which the vehicle is travelling
and to determine the vehicle location on that segment (Greenfeld, 2002; Quddus et al, 2003).
Map matching not only enables the physical location of the vehicle to be identified but also

3
improves the positioning accuracy if good spatial road network data are available (Ochieng et
al., 2004). This means that the determination of a vehicle location on a particular road
identified by a map matching algorithm largely depends on the quality of spatial road map
used in the algorithm. A poor quality road map could lead to a large error in the map-matched
solutions. A map matching algorithm can be developed for all applications (i.e., generic) or it
can be developed for a specific application. For example, Taylor et al. (2006) develop a map
matching algorithm called Odometer Map Matched GPS (OMMGPS) applicable to a service
where the most likely path of a trip is known in advance. In this study, only generic map
matching algorithms will be considered. A map matching algorithm can be developed for a
real-time applications or a post-processing application. For instance, Marchal et al. (2005)
develop an efficient post-processing map matching method for large global positioning
systems data. In this study, a real-time map matching algorithms will be considered as most
ITS services require a map matching algorithm that can be implemented in real-time.
It is essential that the map matching algorithm used in any navigation module meet the
specified requirements set for that particular service. Although the performance of a map
matching algorithm depends on the characteristics of data inputs (Chen et al. 2005), the
technique used in the algorithm can enhance overall performance. For instance, the
performance of a map matching algorithm based on fuzzy logic theory may be better than that
of an algorithm based on the topological analysis of spatial road network data if all else are
equal. There are at least 35 map matching algorithms produced and published in the literature
during the period 1989-2006, most of which are recent reflecting the growth in the need for
ITS services. However, most of these algorithms were developed recently as the need for ITS
services has developed. The positioning accuracy and quality offered by these algorithms has
also improved over the years. This is mainly due to the use of advanced techniques in the
algorithms such as Kalman filtering, fuzzy logic, and belief theory, and the improvement in
the performance of the positioning sensors and the quality and quantity of the spatial road
network data.
Different algorithms, however, have different strengths and weaknesses. Some algorithms
may perform very well within suburban areas but may not be appropriate for urban areas and
vice versa. A review of literature suggests that existing map matching algorithms are not
capable of satisfying the requirements of all ITS applications and services. For instance, bus
priority at junctions requires a positioning accuracy of 5 m (95%) with integrity. None of the
existing algorithms can meet this positioning requirement, especially, within dense urban
areas. This implies that apart from other elements including input data sources, further
improvements to map matching algorithms are essential. To accomplish this, it is necessary to

4
identify the constraints and limitations of existing map matching algorithms for further
research. Therefore, the objectives of this paper are to perform an in-depth literature review of
existing map matching algorithms and then to uncover the constraints and limitations of these
algorithms. In addition to this, the paper also recommends ideas for future research to
overcome these limitations. The potential impacts of the European Geostationary Overlay
Service (EGNOS) and the forthcoming Galileo system on the performance of map matching
algorithms are highlighted also. It is important to emphasise that this paper is intended to
serve as a key reference for future research and development of map matching algorithms by
bringing together existing knowledge and defining future research directions.
The reminder of the paper is structured as follows. First, an in-depth literature review of map
matching algorithms is presented, followed by a presentation of the performance of some
existing map matching algorithms. The next section describes the constraints and limitations
of existing map matching algorithms. This is followed by the discussion on the potential
impacts of the Galileo and EGNOS on the performance of map matching algorithms.
Conclusions summarise the key constraints and limitations of existing algorithms and provide
some thoughts on future research directions.
2 LITERATURE REVIEW
As stated above, map matching algorithms are used to determine the location of a vehicle on a
road. Most of the formulated algorithms utilise navigation data from GPS (or GPS integrated
with deduced reckoning sensors) and digital spatial road network data. One of the common
assumptions in the literature on map matching is that the vehicle is essentially constrained to a
finite network of roads. While this assumption is valid for most vehicles under most operating
conditions, problems may be encountered for off-roadway situations such as car parks or on
private land. Most of the studies also report that the digital spatial road network data used for
map matching should be of a large scale in order to generate position outputs with fewer
errors (e.g., Zhao, 1997, Quddus et al., 2006a).
Procedures for map matching vary from those using simple search techniques (Kim et al.,
1996), to those using more advanced techniques such as the use of an Extended Kalman
Filter, fuzzy logic, and Belief Theory (Najjar and Bonnifait, 2005; Quddus et al., 2006b).
Approaches for map matching algorithms in the literature can be categorised into four groups:
geometric (Bernstein and Kornhauser, 1996), topological (White et al., 2000; Joshi, 2001;
Greenfeld et al., 2002; Chen at al., 2003; and Quddus et al., 2003), probabilistic (Zhao, 1997;
Ochieng et al., 2003), and other advanced techniques (Najjar and Bonnifait, 2005; Pyo et al.,