This paper presents an inexpensive path-planning capability for a quadrotor UAV operating in indoor environments using commercial-off-the-shelf (COTS) components. The path-planning is based on the Rapidly-Exploring Random Trees (RRT) algorithm, which allows the quadrotor to reach a target waypoint, or coordinate, by expanding a region-filling tree from its starting location. The flight route is implemented as a series of intermediate coordinates for an a-priori 2-D environment map and is transmitted individually from the ground control station during flight. Ultrasonic sensors are used in conjunction with the quadrotor's APM 2.6 autopilot for obstacle avoidance. The current software on the autopilot and ground station are limited to static obstacles, which include stationary objects like walls and furniture that are built into the environment map. Path-planning was successfully demonstrated through softvvare-in-the-loop simulation and human-assisted flight tests with the use of GPS. The APM 2.6 autopilot relies on the Kalman filter and GPS data to correct dead-reckoning error. Thus, initial tests show that a more accurate IMU is required to limit the drift, and that localization is necessary by means of accurate rangefinders, such as LIDAR, to replace the GPS input in the Kalman filter. Future efforts will focus on dynamic sense and avoid (S&A) capability, which will allow the quadrotor to adjust its flight path when a moving obstacle is detected.
展开▼
