Development of a low-cost full-stack Spatial AI system

点击翻译按钮获取中文摘要

Robotics and drones are set to change the world we live in. While industrial applications such as robotic assembly, painting, welding, and materials handling have historically dominated the market, today, the growth in the number of robots and drones is driven largely by non-industrial applications, such as consumer or enterprise robots. The new applications of robotics and drones span a diverse range of sectors, including autonomous farming, medical sample deliveries, and personal assistants. Of the 76,000 drones predicted in the UK skies by 2030, one third will deliver improvements to public services; for example, supporting emergency services with search and rescue, surveying incident and crime scenes, and supporting the fire brigade in investigating burning buildings (PwC Drones Report, 2018; Flying High, Nesta 2018).All robotics and autonomous systems require sensors to understand where they are in the real world and safely move around in their surroundings. Laser-based techniques such as LiDAR are used in Google's self-driving cars, in combination with bumper-mounted radar, GPS, and other sensors; however, while these systems offer state-of-the-art performance, they are too expensive, bulky, and power hungry for many robotics applications. Cameras and other lower cost sensors such as accelerometers can be attached to smaller robots and drones that have application in a wider range of markets, including environments without reliable GPS; for example, in indoor household products. These sensors enable the robot or drone to view its surroundings and determine its location, in a process known as visual simultaneous localisation and mapping (SLAM).SLAMcore is a London-based start-up founded by world-leading computer vision and SLAM experts from Imperial College London. Our cofounder Professor Andrew Davison designed the algorithms at the heart of Dyson's first robotic product, the Dyson 360 Eye robot vacuum. We are developing systems that can not only position and map robots and drones in their environment, but use artificial intelligence to identify and respond to the objects around them; for example, recognising and tracking people.With funding from Innovate UK, SLAMcore will develop a system that relies on lower cost sensors and computer (Raspberry Pi), consumes less power, and has a smaller size, supporting the creation of new markets including service robots for retail, hospitality, and healthcare settings. The project will deliver a new technical capability for the UK and the global robotics and drones industry (predicted to reach USD 226billion by 2021; Tractica, 2016).