Intelligent Control of UAVs

Abstract:
Uninhabited Air Vehicles (UAVs) and Uninhabited Combat Air Vehicles (UCAVs) operate in environments with a high likelihood of damage and hardware failure, thus removing life-threatening situations for pilots. BAI is designing the intelligent controls that retain a high level of sophistication and performance and minimize development costs.

Problem:
Lowering production costs on UAVs and UCAVs might include removing non-flight-critical systems such as countermeasures and highly-redundant actuation. There is also a strong requirement to reduce the autopilot development costs for these vehicles. UAVs will have to make up for their lack of sophistication in hardware with more complex software, and in particular control software. The need for intelligent control of UAVs / UCAVs becomes evident.

Intelligent control algorithms are expected to provide: autonomous trajectory generation given waypoints or high level tasks, control reconfiguration for command tracking despite damage and control effector failures and learning mechanisms by which the controller compensates for mis-modeled or changed vehicle dynamics online.

Solution:
BAI has assembled a team who collectively have decades of experience in the development, implementation and flight testing of both production controller and advanced research flight control systems. The key technical features of the team's algorithms are:

• Learning: the stability and control loops use function approximators with local support, which learn the stability and control derivatives of the aircraft as it flies. This is a significant advantage over other adaptive control methods which do not retain useful (or any) information about the aircraft.
  
  
• Online trajectory generation: a hybrid automaton generates trajectories for way point (along with velocity, flight path, and heading) following. The initial libraries used by the automaton are constructed off-line, but mechanisms for on-line modification of the databases are included to reconfigure the trajectory generator as necessary.
  
  
• Reconfiguration: provided both in the control algorithms and the trajectory generation.
  
  
• Provable Stability: development of both components of the intelligent controller are based on formal stability analysis.
  
  
• Reduced Development Cost: the control design is first performed in a low-cost, medium-fidelity simulation, but is still applicable to the true air vehicle. This technique is validated by applying the control laws to a high-fidelity simulation of a UAV.