Over the coming decades our society will be revolutionised by Autonomous, Intelligent Machines and Systems (AIMS) which can learn, adapt and act independently of human control.
There is an exciting and timely opportunity to develop these technologies for sectors as diverse as energy, transport, environment, manufacturing and aerospace. In transport, for example, we will eventually have cars that can drive themselves, interacting safely with other road users and using roads efficiently; in energy we will have smart grids, intelligently acting to help maintain energy security for years to come; in manufacturing & industry, intelligent systems will cut waste in industrial processes; in aerospace we will have an increasing reliance on smart systems, from planning to energy efficient control.
The University of Oxford has a world-class reputation in the underpinning technologies of AIMS. Graduates have gone on to become entrepreneurs, taking leading roles in industry and commerce or continued their careers in academia.
Recognising the strength of our research and teaching in this area and the potential for future breakthroughs, the Departments of Engineering Science and Computer Science were awarded the Centre for Doctoral Training (CDT), which provides the focus for the development of these new technologies.
Our research combines academic training with industry-initiated projects and integrates fundamental theory with hands-on practice.
The first Key Skills Area is in enabling autonomous systems to identify and interpret complex scenes, from moving vehicles to human activity and form robust situation assessments to enable appropriate action and decision making. For example, robotic systems require such capabilities so that they can navigate in unknown environments; augmented reality systems require methods for scene perception and object identification. Our vision is to train a new generation of researchers that will be able to understand and embed such intelligent machines across sectors, from smart buildings to driverless cars.
The second Key Skills Area is in making machine autonomy and intelligence ubiquitous; allowing machines to discreetly pervade the world around us and assist us. At the heart of this is a scaling issue and the need to coordinate and harness the potential of ubiquitous computational agents to meet the challenges of sustainability, inclusion and safety and to enable effective & seamless machine-to-machine coordination and machine-to-human interaction. The CDT will promote a training foundation for students to inject machine intelligence into real-world applications, such as the critical domains of healthcare, smart grids and energy resources, big data analytics, disaster response, citizen science, human- in-the-loop systems and the environment.
The third Key Skills Area lies in developing effective techniques to monitor and control intelligent machines, such as those used in manufacturing, transportation and biosensing/healthcare systems, and to ensure their safety and dependability. For example, how do we ensure that the embedded software controller of the self- driving car does not crash, or that the implantable blood glucose monitor correctly identifies an abnormal range and raises an alarm? Verification via model checking provides automated methods to establish that given requirements are satisfied, but is challenged by the need to consider the complex interplay of discrete, continuous and probabilistic dynamics. Students will be challenged to apply this material to control and verification problems in diverse areas, such as automotive controllers, wireless security and coordination in rescue scenarios.
The fourth skills area will be to realise the vision of connecting intelligent devices seamlessly and everywhere and to allow them to share their sensing, monitoring and actuating capabilities. This is often referred to as “M2M” or the “Internet of Things”. Currently, there are key technical barriers in the widespread adoption of “intelligent networked” devices. First, machine interaction typically relies on context- awareness (e.g. location) which is problematic in indoor environments. Second, sensors and actuators are inherently unreliable, often lacking calibration, quality estimation, energy management and fault detection capabilities. This compromises their practical use. Third, most M2M solutions have been designed to meet functional requirements, ignoring security and privacy concerns, both in peer-to-peer ad-hoc networks and cellular networks.