Chinese
German
Spanish
Japanese
Korean
New capabilities, new skills: the story of how we made GNSS-denied localisation possible – interview with Paris Austin
How do next-generation inertial navigation solutions get built? If you’ve ever thought about building your own localisation product, or even just wondered how these things get made in the first place, this is the blog for you.
You may have seen that we’ve recently launched WayFinder, our new GNSS-denied localisation solution. It’s a complete solution that works out of the box, containing a GNSS/INS, LiDAR, and two cameras as well as on-board processing power that convert LiDAR data into localisation data that constrains position drift in challenging environments.
In our latest blog, we sat down with our Principle Product Owner, Paris Austin, to tell the story of how we brought GNSS-denied localisation and WayFinder to life.
OXTS: Let’s start at the beginning. Why did OXTS decide to create a GNSS-denied localisation solution?
Paris: There were two main driving factors that made us want to develop a GNSS-denied localisation solution. The first was that we could see that there was an unmet need in the spaces our customers operate in – in fact, it was a common need across a wide range of industries. Basically, everyone we spoke to who was using localisation data was grappling with the same two challenges: how do I get accurate localisation data without GNSS signal, and how do I transition between GNSS-based localisation and other methods?
It’s that second challenge that was the biggest one in many ways. Things like SLAM have existed for years and provide good indoor localisation data, but integrating that into a system that also functions outdoors hasn’t really been done in a commercially scalable way yet.
The second factor was a real desire to push the boundaries of what you can do with localisation technology. Our team has been at the forefront of inertial navigation technology for decades – long enough to recognise that the future in our space lies in multi-sensor fusion for GNSS-denied localisation. So we married that desire up with the challenge we were seeing in the marketplace, and that’s where everything started.
OXTS: What was the process like to create WayFinder?
Paris: The beginning of WayFinder was the development of LiDAR Inertial Odometry, a post-process method of using LiDAR data to create velocity data that could be used to improve position accuracy. We built that a few years ago now, and it delivers incredible results, both as a supplement to weaker GNSS signal and as a replacement for GNSS in environments like underground car parks.
We quickly realised that if we could make that real-time, then we’d have a powerful GNSS-denied localisation capability. So that was where we started: how do we create real-time LiDAR aiding and what hardware is needed to run that? That was the basis of LiDAR Boost, the technology that really makes WayFinder work.
Learn more: How LiDAR Boost supports GNSS-denied localisation
The goal was always to create a self-contained GNSS-denied localisation solution. Just like a GNSS/INS is a complete localisation solution, we wanted WayFinder to be completely self-sufficient too.
The very first thing we did, therefore, was to pick the hardware we needed. And that brings me to my biggest piece of advice, I think, for anyone in a similar endeavour – Get your hardware sorted first. There are so many variables with hardware that can affect your build that you really need to know which sensors, processors, and so on that you want to use. Even the variables between different units of the same LiDAR, for instance, can trip you up. By starting with the hardware, we avoided many more challenges down the road.
OXTS: How did the build go?
Paris: The build was, as these things are, a long grind, and I’m proud of how hard the team worked throughout it. And, as anyone who’s done this sort of thing before will know, different elements were moving at different paces. The real-time LiDAR algorithm, for instance, was ready long before the hardware was because we had already started building it beforehand. Other pieces of software were ready after the hardware was finished. It created lots of challenges in terms of coordinating different teams, hardware integration, and data integration, but the team handled it really well.
We also put a lot of work into our web interface, that lets users monitor the system and the data they’re getting. I’m sure other engineers will relate to the feeling of spending 80% of your effort getting the last 20% of the project – the user interface – done. It was a new area for our team of engineers, and I believe that the finished product is excellent – and it’s made our team even stronger.
OXTS: How has it made the team stronger?
Paris: building the web interface – and in fact multiple areas of the project – have pushed our engineers out of their comfort zones to explore new skills and ways of working. As I’ve already said, the web application was a new area for us, and our engineers have loved learning those new skills. The complexity of the build meant that our teams worked more closely than they have in the past – which has been great. Everyone is now so much more aware of what other teams bring to the table, and understands how their work fits into what everyone else is doing – it means we can work more efficiently and more effectively.
And at the end of the day, all that work and growth has also helped us deliver a new GNSS-denied localisation capability that’s genuinely exciting, both for us and for our customers.
OXTS: What excites you most about WayFinder?
Paris: Personally, it’s the level of integration that it achieves for our customers that I think is most exciting. It really does work out of the box. Anyone who has attempted sensor fusion to enable GNSS-denied localisation before will know how complex the work can be to get the data out of one sensor and into a format that’s usable by another sensor; the fact that WayFinder takes all that pain away from a project is something that I think many engineers will be very happy about!
I also love the localisation expertise we bring to the project. There are other people attempting real-time sensor fusion, but it’s an incredibly complex task – and one that OXTS has been tackling for twenty years. I think that unique perspective makes WayFinder one of the most well-rounded solutions out there, capable of working in – and across – any environment.
OXTS: What do you think the future holds for real-time sensor fusion?
Paris: WayFinder is proof that you can build a platform for real-time sensor fusion. The next step is to work out what we can build on that foundation.
LiDAR Boost is one of those things you can build on, but there are many other sensors out there that we could do similar things for. Camera-based odometry, or radar, for instance. I’d expect new algorithms to be written that enable users to plug in a wider variety of sensors, giving them flexibility in how they build their setup.
I think we’ll also see tighter integration with robotics workflows. A major use case for WayFinder is GNSS-denied localisation for autonomous navigation, which is generally part of a robotic control system. People are doing incredible things with NVIDIA Jetsons in robotic control; how can we intersect with that?
Looking above the technical level, I’m excited to discover what different GNSS-denied localisation challenges we can solve for customers, now that we can give them better localisation data in different environments. Could we use it for passive vehicle-to-vehicle detection in ADAS testing, for instance? In many ways, the sky’s the limit. And I know that OXTS will be looking to work with customers to work out where to take this technology next. We want to build the perfect products for people, so it makes sense to involve our customers in the process of evolving WayFinder.
Download the WayFinder Datasheet
WayFinder can help you to deliver your GNSS-denied localisation project accurately, reliably and on-budget – whatever environment you’re operating in. Find out the technical specifications and more key information here.
