Control Architecture Design and Demonstration for Cooperative UAV's

Faculty contact: Jonathan How
Department of Aeronautics and Astronautics
Aerospace Controls Laboratory
Massachusetts Institute of Technology

Research funded by AFOSR Grant FA9550-04-1-0458

Research objective: develop new computational theory for the coordination of complex systems
comprised of many UAVs operating in dynamic and uncertain environments.

• Optimal coordination with heterogeneous UAVs/targets
• Trajectory design - Optimized but avoids SAM sites and no-fly zones

Focus on:

• Algorithms for cooperative activity and path planning.
• Develop feedback controls perspective to high-level planning
• Planner in the loop -> stability, disturbance rejection, performance robustness?
• Effect of noise and uncertainty on the closed-loop system?



 

Results in Phase 1:

• Extended cooperative path planning algorithm (RH-MILP) to 3D (Paper)
• Decentralized collision avoidance algorithm for multiple UAVs (Paper2)
• Robust Model Predictive Control (RMPC) to guarantee robust satisfaction of coupling constraints (Paper1)
• UAVs communicate, but solution process does not iterate -> scales well
• Extended Receding Horizon Task Assignment (RHTA) algorithms to add robustness to uncertainty in situational awareness
• Coupled reconnaissance and strike objectives (Paper)
• Filter-embedded task assignment (FETA) reduces impact of disturbances & uncertainty in cost estimates (Paper)
• New approximate solution to robust shortest path problems used to include environmental uncertainty/risk (Thesis)
• Completed design of DURIP multi-rover and multi-UAV testbeds
• Performed initial flight tests of path planning algorithms (Paper)

Recent Results in Phase 2:

• Developed new theoretical approach to optimal search that captures the uncertainty typically present in à priori descriptions of the environment
• Uses Beta function to include a distribution of these probabilities
• Yields robust search policies
• Developed new Robust Distributed Task Assignment (RDTA) algorithm
• Recognizes that UAVs can remove inconsistencies due to imperfect information by communicating their plans to resolve conflicts
• Similar to “closing a feedback loop” during the planning phase
• More efficient than just trying synchronizing full situational awareness
• Extensions to previous Robust Model Predictive Control Robust Safe but Knowledgeable (RSBK) algorithm
• Uses shorter planning horizons (faster computation), incorporates vehicle safety
  (handles pop-up threats) and includes effective cost-to-go (ensures completion)
• Compared RMPC with min-max approach – developed hybrid algorithm
• Embedded RHTA into multi-level search and tracking controller (with BAE)
• Hardware-in-the-loop demonstrations in preparation for summer flight tests.
• Augmented previous UAV fleet with 4 Monocoupe 90A’s