Squadron leaders

A swarm of UAVs that can fight together as a group is being developed in a project headed by Cranfield University.

The research aims to create an airborne group of UAVs that can plan their flight trajectories during missions and carry them out safely. This will be achieved by each drone comparing its individual movements to the flight plan of others, within the guidelines of the main mission objective.

The three-year project, involving Cranfield and Imperial College researchers, has been awarded a £400,000 EPSRC grant and is due to start in October. Cranfield is to develop new algorithms for co-ordinated guidance and the integration of mission planning and guidance, while Imperial College researchers will look at issues including tracking control.

The use of multiple unmanned aerial vehicles (MUAVs) is attractive to the armed forces as they can provide significant reductions in manpower and reduce the risk to humans in critical security and defence roles.

They are also capable of surveying larger areas than soldiers on foot, while providing information that is more accurate than that gathered by traditional aircraft. It is hoped that the swarm will be capable of new tasks and operations due to its increased autonomy compared with singular remote controlled drones.

Single UAVs act as mobile sensor platforms. However, MUAVs offer a magnification of the sensing capability by creating an airborne, moveable multi-sensor, thus offering new opportunities for surveillance and reconnaissance.

The researchers must find a way to guarantee performance of the dynamic behaviour of MUAVs, particularly co-ordinating their flight dynamics in a predictable way so they can be relied upon to carry out safety-critical missions.

So the main focus of the programme is to develop and validate a rigorous, analytical framework for producing a guaranteed performance from the vehicles.

Their dynamic behaviour is fundamentally different to that of a single vehicle as they are interconnected in a system dependent on the individual dynamics of each aircraft and on the nature of the UAV-to-UAV and environment-to-vehicle interactions.

The modelling will use a mathematical technique known as Kripke modelling, which involves the programming of multiple scenarios or 'worlds' and possible outcomes depending on each choice. There is minimal communication between each drone.

Although there have recently been advances in swarm technology, an actual operation of a group of MUAVs has not yet been fully realised. The Cranfield researchers are therefore proposing to create a detailed set of parameters including each of the vehicles' many characteristics. It will also have to include a comprehensive analysis of their guidance and control systems.

'The main thrust of the project is to develop a rigorous, analytical framework that determines in detail the number and characteristics of the vehicles, the interactions between the vehicles and their embedded guidance and control schemes,' explained project leader Dr Antonios Tsourdos, senior lecturer in the guidance and control group at Cranfield's department of aerospace, power and sensors.

'This framework will be used to predict MUAV behaviour in two multi-task mission scenarios, and assess the resulting performance through simulations and ground vehicle experiments.'

Control of the UAVs will be hierarchical, consisting of three steps. First, co-operative decision-making will be carried out by UAVs acting as a group. They will co-operate according to common mission instructions or tasks to be carried out.

As a result, they will be able to generate co-operative trajectories, which will finally be acted upon by each individual unit, allowing the swarm to move safely as a group.

To ensure this occurs, the researchers must design each layer and its interaction with the next step so that the performance of both the individual UAV and the group can be guaranteed.

Work will involve the study of each of the three layers of co-operative control and the integration of them in two sample scenarios.