Giving you sensors that work, rather than sensors that fail.

Technology and Applications

Our models use data from multiple sensors, such as LiDAR, radar and cameras. Our algorithms compress the vast amounts of data generated from these sensors; using artificial intelligence (AI) they then feed information back to the sensors in real-time, so they can adapt and make intelligent decisions (Adaptive Loop Processing). This is the case even in poor visibility or GPS denied locations, such as deep underwater or underground, giving you sensors that work, rather than sensors that fail.

Sensors are the fundamental building block of the internet of things (IoT) network. Many of them will fail in challenging environments, and they are not optimised for environments they haven't come across before. In addition, each type of sensor faces its own set of challenges under different environmental conditions.

Our key areas of focus are:

GPS Denied Locations

One of the key problems in GPS denied locations is: how do you know where you are? An autonomous vehicle can usually rely on satellite signals to find its way around; in areas such as tunnels where satellite signals are blocked, sensors currently don’t offer much of a solution. Our AI module can process the data from the sensors in real-time, meaning a map of the terrain can be produced and fed back to the sensors, enabling the vehicle to move around effectively.

Thanks to simultaneous location and mapping (SLAM) and our adaptive sensor control technology you can see beyond the visual line of sight (BVLOS). This is useful for railways, drones, deep sea environments, confined spaces and locations in complete darkness.

Our software allows sensors to adapt in real-time to their environment,
enabling them to work in conditions where they would usually fail.

Structures and Materials

We have expertise across the full electromagnetic spectrum. Using different forms of light, in combination with information from other sensors, our AI module can identify different types of materials and surfaces. This is a very valuable feature in infrastructure monitoring, for example: to identify corrosion levels, determine whether a surface needs painting, or to identify different types of building materials: brick, stone, wood, plastic, or metal.

Material identification is also of use to manufacturing processes and in factories: when checking for quality control and contamination, or for health and safety: monitoring chemical spills and slip hazards. Material identification has been vital in the inspection and analysis of many types of products, predominantly using spectroscopy or other techniques using specialised equipment. We are bringing it to robotic sensing for enhanced situational analysis.

Optimised sensing powered by AI means that using our software sensors can work in poor conditions, enabling intelligent decisions to be made.

Sensing & Mapping

In ideal environments regular 3D sensing is not easily done. It is even harder to do accurately using low cost sensors. Our technology will enable even low cost sensors to adapt and take accurate measurements in real-time.

Each type of sensor faces its own challenges under different environmental conditions. Using our software we can combine the inputs from different sensors and relay it to individual sensors, in real-time, so that they can work well even in extreme conditions.

Conditions where light-based sensors usually fail.

AI Safety

We take AI Safety seriously. It is difficult to predict with any accuracy what the impact of AI will be. The following are our aims in terms of safety:

    Social Benefit - Serve global human and environmental good by solving large scale engineering problems, thereby improving people’s lives
  • International Standards - Comply with international standards on AI safety as they are agreed. We are committed to AI being for the benefit of all
  • Diversity - We aim to recruit the broadest possible range of talent in our work and encourage different viewpoints


Currently we are working with the water industry to survey their tunnels to prevent leaks and identify potential structural failures before they happen. Structural failures such as tunnels or sewers collapsing are obviously costly to deal with. Leaks alone cost water companies 9.6 billion USD every year. We can help reduce these costs.

Current methods, such as using CCTV surveys, don’t work properly in larger structures, nor do they give accurate measurements that can be compared effectively over time.

Using low-cost sensors, we create accurate (< 5 mm accuracy) 3D maps of underground structures. The result is higher quality 3D maps, without having to use expensive, survey-grade LiDAR sensors. Our AI-enabled software is able to automatically identify defects in structures so that they can be repaired before expensive leaks or catastrophic damage occurs.

We provide a new LiDAR survey method that is more accurate than CCTV and cost effective.

ADAS (advanced driver assistance systems) are wholly reliant on sensors. On any vehicle multiple sensors will need to interact and learn from each other to deliver the safest and smoothest ride. Existing sensors used in ADAS and autonomous vehicles are not able to cope well in poor visibility: fog, snow, heavy rain, even with the most advanced LiDAR, radar and cameras on the market. They also generate vast amounts of data, 400 GB every hour (from cameras, sonar, GPS, LiDAR) that needs real-time processing.

Our models address these challenges by predicting the environmental variations from radar, LiDAR and other data. This information is then fed back to the sensors, enabling them to adapt to the conditions of their surroundings in real-time and see beyond the fog, rain, or snow.

Our algorithms can compress the data so that it requires less processing, saving:

Power - so the car can drive for longer.
Weight - fewer batteries, processors needed.
Space - more room for passengers and luggage.