Radar Localisation FAQs

Frequently Asked Questions

An answer to the most common questions on Radar Localisation.

Q. What Does The Radar Localisation Product Do?

A. For an autonomous vehicle, Radar Localisation aims to answer key question in autonomy “Where Am I?” using only a Radar in all weather and environmental conditions. Radar Localisation will tell the user where their radar is within a Radar map to centimetre level accuracy.

Q. How accurate is the system?

A. The solution will deliver pose estimates with an accuracy of better than 8cm in a huge variety of operating domains and conditions. It has been tested in feature-sparse tunnels, deep underground, in cities, on highways, even on high speed trains and has proven this level of performance in all these domains.

Q. How Does The Radar Localisation Product Work?

A. There are two key stages: Mapping and Localisation. First, a map of the environment is generated by driving a radar through the intended operating domain collecting data. Only one mapping run is required to build a map. The data is then processed and a radar map is built. At run-time the radar uses its current scans, along with the map as a reference, to provide the user with an estimate of the radar’s position.

Q. Does the solution rely on external HD Maps?

A. No. Map sizes are less than 20MB per kilometre and only use data from the radar.

Q. What Infrastructure is required to localise? How complex is the system?

A. The solution works completely infrastructure free, requiring only a single sensor per vehicle.

Q. Does it require maps to be created at different times of day/light conditions?

A. No. The solution uses only radar to create maps and localise. Radar maps are incredibly robust to changes of environmental conditions and totally unaffected by lighting conditions.

Q. I can’t mount the sensor with a 360 degree field of view. Does this matter?

A. No. The solution will work even when the field of view from the radar is less than 180 degrees. 360 degrees is ideal, but not necessary.

Q. How does the radar initialise its position? Is it a manual process?

A. The process of initialisation is fully automated. The radar will initialise its position automatically within the map that has been created.

Q. Great, I get a position estimate! But what co-ordinate frame is it in?

A. Short answer, any. By default the position estimates will be in the radar’s own co-ordinate frame, however can be translated to any reference co-ordinate system.

Q. So my product needs position estimates in UTM co-ordinate frame, how could a radar know where it is in the world?

A. Radar Maps are topological, able to store semantic information to nodes throughout the map. Therefore, if a user requires position estimates to be in a global frame of reference, then the radar map can associate UTM pose estimates to nodes during the mapping process. At runtime, localisation occurs relative to nodes within the map and in this way the system can output a localised position estimate relative not only to UTM pose, but any co-ordinate system the customer is using.

A good example of a ‘custom’ co-ordinate frame is the London Underground train network. The map is topological, where the public only care about the relationship between nodes in the map (and how far away the next train is). Our mapping style means that it would be well suited to output pose information in such a custom co-ordinate frame.