Autonomous and coordinated field agents

Drone colonies embody a project that is innovative both in form and in substance. Since 2010, our research has aimed to model and test the behaviour of drones that have to communicate with one another, within the framework of various support missions, with little equipment embedded and with limited autonomy.

Initiated by the Regional Council of Aquitaine and led by Thales Systèmes Aéroportés, Aetos is a grouping dedicated to drone systems. It comprises over forty companies, including Sogeti High Tech.

The ability to detect and avoid stationary or mobile obstacles in real-time and automatically reconfigure a flightpath will reinforce autonomous capacities and expand the range of possible applications for drones. In time, they should be able to fly in squadrons comprising dozens of units. Each device is able to exchange information with its peers over secure communications. The aim is to relieve the pilot – i.e. the person responsible for flying them – and the operator – the person responsible for the mission – of some of the inherent mission constraints.

National mobilisation

Our various branches contribute to the project according to their expertise:

The Aix-en-Provence branch brings to the table its competencies in mechanical design acquired from Airbus Helicopters.

The branch in Bordeaux is developing the swarm simulator, the various potential models, their integration and behaviour.

At Ile-de-France, over 90 engineers have developed behavioural models amounting to over 1.6 million lines of code. They have also built the technical and operational drone simulator.

In Toulouse, a team is dedicated to learning algorithms and mission resolution simulation by cooperative multi-agent systems (CMASs). It is designing and developing the G-MASS (Generic Multi-Agent System Simulation), combined with the development of an experimental demonstrator that will put the algorithms into real-world functional conditions.

In practice

Sogeti High Tech is developing a simulation platform to test the various flight configurations for a swarm. This platform will serve to test or emulate the decision algorithms, flightpath generation, behaviour according to embedded sensors, energy consumption, transmissions, functional validation of ground stations, adequate ratio of operators to pilots according to the mission and number of drones, etc. This optimises integration of different modules, without costly flight testing.

The G-MASS simulation platform combined with the experimental demonstrator facilitates perfection of algorithms for drones made of up flying and wheeled robots, with intelligence distributed between embedded cards and Android mobile phones. The mobile phone is the command unit – the "brain" – whereas the execution unit is embodied by the frame supporting the sensors. The aim is for the drones to be able to locate and take control of a given area – such as a fire outbreak.

todo todo
  • Fabrice Robert
    Fabrice Robert
    Innovation Manager
    +33 (0)5 34 46 92 64