Romain Lefez

Development of perception solution for autonomous logistic vehicles.

• Prototyping and development of a harsh weather filter for lidar sensors.

• Development of full performance test method to assert safety and automatic tuning.

Demonstrator (Dragonfly) of auto emergency capabilities of an Airbus A350 to return on ground safely after pilot incapacitation. Work mainly on landing part.

• Technical lead of computer vision areas (sensors, VBN algorithms including calibration measure ment and computation).

• Architect of final approach solution (CV apps, navigation fusion filter (EKF) + database, guidance laws, IHM).

• Main integrator of landing solution into the embedded system (bay of several computers set into the flight test aircraft running on a main hand DDS middle ware).

• Development of auto massive testing tools and performance generation tools to demonstrate landing solution readiness for closing the loop.

Innovation department of vision based navigation field mainly for space prototypes.

• Development of an online image stabilization algorithm for a long range camera fixed to an (Airbus) helicopter. Prototyping of an unusual but generic intrinsic calibration optimization based on a field of splines.

• Adaptation of a model based tracking solution on RemoveDebris images captured during its first test in low space orbit.

• Prototyping and delivering an integrated solution (ROS) to perform global image navigation based on satellite geo-referenced rasters and stream of the camera looking forward in the cockpit of an Airbus flight test aircraft to automate taxi phase (demonstrated in flight). Prototyping a pla nar SLAM (bundle adjustment, levenberg-marquardt) to perform the same objective for taxi and landing phase.

• Development of an enhanced Mosse tracker giving invariance regarding scale and rotation change. Prototyping a SLAM algorithm (bundle adjustment, levenberg-marquardt) to perform navigation through an auto computed 3D map updated at each flight above the planned path.