On course for accuracy

A UK team plans to develop more accurate satellite location technology that could improve SatNav systems and help future road charging schemes operate efficiently.

Researchers at Loughborough University are investigating map-matching algorithms that integrate locational satellite data with the spatial road network data to identify the physical position of the vehicle on the road in a variety of environments. They will focus particularly on algorithms for dense urban areas.

'A navigation device like a TomTom normally receives latitude and longitude data from a satellite, which is then superimposed on to an electronic map data inside it,' said Dr Mohammed Quddus at Loughborough.

'If you superimpose electronic map data with satellite data, you will find there is an error or mismatch between them, so the data from the satellite does not fall exactly on the road network. In urban areas, when the satellite signals are blocked due to high buildings and trees, the error increases up to 100m.'

Map-matching algorithms reduce the error by reading the error sources in the satellite and electronic data, then matching the two datasets to estimate the most likely road on which a vehicle is travelling. The researchers hope to make a prototype device by 2011 that will contain the new algorithms and comprise off-the-shelf products, including a satellite data system such as GPS and a gyroscope to give local position data.

Quddus said most map-matching algorithms have been developed for rural areas, where open spaces mean that satellite signals are less likely to be blocked. While some have been designed for urban areas, no-one has addressed the algorithms necessary for built-up areas because of problems such as tall buildings and complex road layouts, including four-lane junctions.

However, the researchers aim to integrate a number of algorithms from which a suitable one can be selected depending on the environment (gauged from information such as building height) rather than come up with one algorithm for all situations.

'This project will identify a set of representative map-matching algorithms. It would not be wise to use a single algorithm. There is no point in using complicated mathematical algorithms that work in urban areas in a rural area where a simple map-matching algorithm would work fine,' said Quddus.

The best existing map-matching algorithm has an error margin of 10m. The Loughborough researchers want to reduce this even further to provide lane-level accuracy — that is an accuracy of two to three metres. This would be generally useful, but especially useful for what the researchers refer to as 'liability applications'.

'Liability applications means things like road user charging,' said Quddus.

'The government is considering replacing road tax and petrol duty with distance-based road user charging. If you are driving on a road which is parallel to the motorway, and the motorway is charging £1.50 a mile and the parallel road maybe 50p per mile, if the algorithm wrongly identifies a particular link, the bill will be incorrect to the users.'

In addition to improving accuracy, the researchers will address the reliability of information provided by the algorithm to the driver.

'We are developing a technique that can also give you a reliability indicator on a zero to 100 scale. That means if the scale shows 100, the positioning information is the most reliable information. If it is below, let's say, 20, the user should not believe this information as there may be something wrong with the system, the satellite or the electronic road network data,' said Quddus.

As a future extension of the project, the researchers would also consider increasing the accuracy to less than one metre. This would be useful in the development of autonomous vehicles, as well as improve the capabilities of collision-avoidance systems and adaptive cruise control in modern cars.

Meanwhile, sensors in cars with collision- avoidance systems, such as gyroscopes and wheel sensors, will help the researchers in their project. They are looking to calibrate vehicle-based sensors with the satellite sensor to help its prototype navigation device maintain continuous positioning information, for example, if satellite data is lost because the vehicle is travelling in a tunnel.

The project is being supported by technical and business consultant Helios, which will provide expertise on how to implement the eventual product.

Anh Nguyen