Vision based navigation for an unmanned aerial vehicle

Abstract

We are developing a system for autonomous navigation of unmanned aerial vehicles (UAVs) based on computer vision. A UAV is equipped with an onboard cameras and each UAV is provided with noisy estimates of its own state, coming from GPS/INS. The mission of the UAV is low altitude navigation from an initial position to a final position in a partially known 3-D environment while avoiding obstacles and minimizing path length. We use a hierarchical approach to path planning. We distinguish between a global offline computation, based on a coarse known model of the environment and a local online computation, based on the information coming from the vision system. A UAV builds and updates a virtual 3-D model of the surrounding environment by processing image sequences and fusing them with sensor data. Based on such a model the UAV will plan a path from its current position to the terminal point. It will then follow such path, getting more data from the on-board cameras, and refining map and local path in real time.

Visual simulator for MavLink-protocol-based UAV, applied for search and analyze task

“…They are exceptionally critical when operating in urban environment and during inspection tasks. In such applications robot is required not only to avoid obstacles (even when flying on high altitude) but also to position itself precisely to execute given tasks [10], [19], [20] The scenario presented in section V require utilization of precise navigation as well. To support UAV with necessary capabilities the following methods were provided: 1) Calculating position of Points Of Interest (later referred to as POI) in the area photographed from high altitude.…”

Planning 3-D Path Networks in Unstructured Environments

“…Nikolos [12] proposes an off/on line planner but only for 2D-1/2 environments. Sinopoli [13] is interested in nap-of-the-earth navigation with the aircraft flying on top of the tree canopy instead of inside the canopy. Vision-based systems have also been considered for navigation in canyon-like structures such as urban environments [6], [5].…”

An Experimental Study of the Autonomous Helicopter Landing Problem

“…The Robotics Institute at Carnegie Mellon University has developed a "visual odometer" which can visually lock-on to ground objects and sense relative helicopter position in real time [2], but they have not integrated vision-based sensing with autonomous landing. The problem of autonomous landing is particularly difficult because the inherent instability of the helicopter near the ground [3]. [4] have demonstrated tracking of a landing pad based on vision but have not shown landing as such.…”

“…The overall landing strategy is best described as a simple finite state machine with three states 3 : search, track, and land. Initially the helicopter is in the search mode.…”