3D Obstacle Avoidance in Adversarial Environments for Unmanned Aerial Vehicles.
Abstract
As unmanned aerial vehicles (UAVs) are considered for a wider variety of military and
commercial applications, the ability to navigate autonomously in unknown and hazardous
environments is increasingly vital to the effectiveness of UAVs. Reliable and efficient
obstacle detection is a fundamental prerequisite to performing autonomous navigation in an
unknown environment. Traditional two-dimensional (planar) obstacle detection techniques,
though computationally friendly, are often insufficient for safe navigation through complex
environments in which commanded trajectories are simultaneously restricted vertically and horizontally by multiple buildings or by increases in terrain elevation. To this end, a pan/tilt-mounted
laser rangefinder is explored as a means of identifying and characterizing potential
obstacles in three dimensions (3D). From GPS position data and inertial sensor
measurements, the filtered laser rangefinder data are transformed into local inertial
coordinates and compiled into a dynamic three-dimensional grid-based mapping of the
specified domain. Utilizing the grid-based map data, path planning algorithms generate the
necessary obstacle avoidance trajectories. The Georgia Tech GTMax UAV helicopter and
simulation environment provide a suitable test-bed for verification of the proposed obstacle
detection and avoidance methodology.