CDC’24 in Milan

I attended the 2024 Conference on Decision and Control in Milan. The photo above is a panoramic view of the city from the roof of the Duomo. This year, I had one paper, which was led and presented by my former student Matthew Wallace, who recently graduated.

The paper with Matthew, Brett Streetman, and myself was titled “Model predictive planning: Trajectory planning in obstruction-dense environments for low-agility aircraft“. In short, the work presented a full-stack motion planning and control strategy for low-agility aircraft. Unlike drones, which are highly maneuverable, low-agility aircraft such as fixed-wing airplanes, sacrifice maneuverability for the benefit of extend range. Low-agility aircraft cannot rapidly slow-down, accelerate, or change altitude. They must maintain a relatively fast airspeed otherwise they will stall and drop from the sky. For such aircraft, avoiding obstacles is uniquely challenging. In our paper, we considered a fixed-wing aircraft navigating through a cluttered field of obstacles. Our approach, which we call “Model Predictive Planning” (MPP) has several parts:

  • A modification of the classical RRT (rapidly exploring random tree) algorithm adapted to finding multiple distinct paths around obstacles. These are geometric paths; we can generate them quickly, but they are not guaranteed to be feasible with respect to the limited maneuverability of the aircraft.
  • A raytracing step where each candidate path is broken up into segments and a sequence of polyhedral sets are created that form a “safe set” surrounding the path.
  • A convex quadratic program is solved whereby the candidate path is perturbed within its safe set while satisfying the aircraft dynamics (including actuator and state constraints), so it becomes a feasible trajectory for the aircraft. Paths that cannot be transformed (infeasible) are eliminated. We then select a trajectory among those we computed.
  • This process is repeated at regular intervals, in a receding horizon fashion, as in model-predictive control.
We added some extra bells and whistles to complete the GNC stack: an Extended Kalman filter to fuse sensor measurements and an LQR trajectory-tracking high-frequency controller. We tested MPP in simulation on a realistic longitudinal aircraft model to verify its effectiveness. Overall, this was a fun project, and best of all, it worked!

Thanks and congrats to the organizing committee and everybody else that came together to make CDC a great success once again!