A Dyson vacuum cleaner that maps your living room is just the beginning—this project aims to give robots the ability to recognise, track, and physically interact with individual objects in a cluttered, changing home. Today’s mass-market robots can only handle one task: floor cleaning. That is because more complex chores—tidying a table, fetching an item, or loading a dishwasher—require a robot to understand its 3D surroundings at the level of individual objects, not just open space. Current algorithms cannot reliably do this in unstructured environments. This partnership between Imperial College and Dyson targets that gap, combining state-based estimation and machine learning to build next-generation object-based SLAM systems. If successful, the research could unlock a new category of domestic robots that perform real physical tasks beyond cleaning. The core work is fundamental—advancing the science of visual scene understanding and vision-guided manipulation—but it is guided throughout by Dyson’s product roadmap. The algorithms developed here are intended to transfer directly into commercial products, potentially reshaping what consumers expect from an autonomous home device.
View original technical description
"Perceiving, Modelling and Interacting Autonomously in a Dynamic Object-Based World" The Dyson Robotics Lab at Imperial College was founded in 2014 as a collaboration between Dyson Technology Ltd and Imperial College. It is the culmination of a thirteen-year partnership between Professor Andrew Davison and Dyson to bring his Simultaneous Localisation and Mapping (SLAM) algorithms out of the laboratory and into commercial robots, resulting in Dyson's 360 Eye vision-based vacuum cleaning robot in 2015 which can map its surroundings, localise and plan systematic cleaning pattern. Our success in working together made it clear that computer vision is a key enabling technology for future robots. This proposal aims to fund the Lab to push the forefront of visual scene understanding and vision-enabled robotic manipulation into new and more demanding application areas. The research activity we are outlining in this Prosperity Partnership complements the large internal R&D investment that Dyson is making to to created advanced robotic products. The aims of this partnership are to invent and prototype the breakthrough robot vision algorithms which could truly take us to next generation capability for advanced robotics working in unstructured environments, and to transfer this technology into the long-term product pipeline of Dyson as they aim to open up new product categories. Dyson has now been working on robotics for nearly 20 years, a period during which the emergence of real consumer robotic products has happened alongside astounding progress in academic research in the broad field of AI. At the present time, floor cleaners are still the only category of mass-market robot which have achieved significant commercial success. This can be put down simply to the greater difficulty of the other more complex tasks and chores that a consumer might want an autonomous product to achieve. These tasks place much larger demands on a robotic system to understand and interact with its complicated 3D surroundings and the objects they contain. This programme will focus on creating the research breakthroughs needed to enable this next generation capability. There are scene perception and modelling competences which underly all of these use cases, and these will be our research focus as we develop the algorithms behind next-generation object-based SLAM systems by combining all of our knowledge in state-based estimation and machine learning. We will also work more specifically on the methods for training learning systems; methods for advanced vision-guided manipulation; and the frameworks needed for practical, contextual human-robot interaction. The core scientific work will be forward-looking and academic, but always with a strong guidance from our partners at Dyson.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know