Dr. Lyudmil Vladimirov
Chief Technology Officer

Joined Voyant: June 2025

Expertise: Advanced tracking, sensor fusion, Bayesian estimation and resilient navigation

Can you tell us a little about your background and the path that led you to Voyant?

I completed an MEng in Computer Science and Electronics at the University of Liverpool, before staying on to do a PhD and then move into a postdoctoral role focused on signal processing, tracking and Bayesian estimation. A significant part of my research has been through applied defence research contracts, working on real-world tracking, sensor fusion and state estimation problems. I have also been closely involved with Stone Soup, the open-source tracking framework, which has given me the opportunity to combine research, software engineering and collaboration with a much wider community.

Voyant felt like a natural next step. It offered a great opportunity to transition from academia into an industrial setting while continuing to work on technically challenging problems, with a much stronger focus on turning research into deployable capability and making a tangible difference in the real world.

How has your research helped shape Voyant’s technology and the problems it is designed to solve?

My PhD focused on using particle filters to deliver operational advantage in the commercial maritime domain. I was fortunate that the PhD was funded by an industrial partner, which meant I was working at the boundary between academia and industry from the outset and had the opportunity to apply my research to real data, real systems and practical operational problems. That experience shaped how I approached research during both my PhD and postdoctoral work: not just asking whether an algorithm was theoretically interesting, but whether it could solve a meaningful real-world problem. Much of the research that I, and others in the group, worked on during that period, particularly around particle filtering, now forms the foundations of the technology that Voyant is looking to commercialise across tracking, sensor fusion and alternative navigation.

Particle filters are central to Voyant’s approach. How would you explain, in non-technical terms, what they can do that conventional approaches such as Kalman filters cannot?

A useful way of thinking about it is that Kalman Filters generally represent belief as a single estimate surrounded by a cloud of uncertainty, effectively describing where something is most likely to be, give or take. This works extremely well for many problems, but can lose information when the real world is highly nonlinear or when several explanations are genuinely plausible.

Particle filters instead use many individual hypotheses, or “particles”, to represent the probability distribution. You can think of them as a crowd of tiny detectives, each backing a slightly different theory about what is going on. As new evidence arrives, the stronger theories gain support and the weaker ones gradually drop away. This allows them to maintain several competing possibilities at once and let later evidence determine which is most likely.

A good example is terrain-based navigation. Imagine trying to work out where you are from the shape of the landscape around you, but several places on the map look very similar. A Kalman Filter will tend to represent one best guess, while a particle filter can scatter thousands of tiny detectives across the terrain. If a new measurement suggests you are near a river, detectives nowhere near one start to disappear, while those in plausible locations multiply. As more evidence arrives, the crowd gradually converges on the most likely position, even if you started with very little idea where you were.

For anyone interested in going a little deeper, there’s a great introduction in Prof. Simon Maskell’s (our CSO) paper, A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking

What excites you most about taking advanced algorithms out of the research environment and applying them to real-world defence challenges?

What excites me most is seeing ideas move beyond papers, simulations and controlled datasets into environments where they have to deal with the messiness of the real world. That is where the interesting problems often appear. Imperfect sensors, unexpected conditions, integration constraints and all the things you cannot fully reproduce in the lab. There is something very rewarding about seeing a solution you have worked on become part of a real system and contribute to solving an operational problem. In defence especially, the potential impact is tangible, and that makes the challenge of turning research into reliable, deployable capability particularly exciting.

How do you see Voyant’s technology evolving – what are the next major challenges you want it to solve?

I see the next stage for Voyant as both expanding where our technology can be applied and making the existing capability more resilient, mature and proven. That means broadening our work across different domains and platforms,  while continuing to develop and expand our capabilities. At the same time, we need to demonstrate that these technologies can perform consistently in real operational environments and integrate effectively with different sensors and systems. The challenge is to take what has already shown strong results in research and trials, extend it to a wider range of use cases, and turn it into dependable capability that can be deployed with confidence.

Away from research and technology, what do you enjoy doing?

There is probably a fair bit of overlap between my work and my hobbies. I enjoy tinkering with IoT and home automation, usually building and coding little gadgets to make them work exactly the way I want them to. Recently, my fiancée gave me the honorary title of “fixer of everything”, which probably says quite a lot. I’ve also, somewhat unexpectedly, developed an interest in lawn care. Between coding, fixing things and pretending to be a lawn care specialist, I enjoy swimming and squeezing in the odd gym session.

Current focus

Development of resilient airborne alternative (terrain-referenced) navigation technologies that help aircraft operate with confidence in GNSS-denied environments, where satellite navigation signals may be degraded, disrupted or unavailable