JOURNAL OFAEROSPACE COMPUTING, INFORMATION,AND COMMUNICATION Vol. 8, February 2011 Evolving Philosophies on Autonomous Obstacle/Collision Avoidance of UnmannedAerial Vehicles Anusha Mujumdar ∗ and Radhakant Padhi † Indian Institute of Science, Bangalore 560012, India DOI: 10.2514/1.49985 Much of the benefits of deploying unmanned aerial vehicles can be derived from autonomous missions. For such missions, however, sense-and-avoid capability (i.e., the ability to detect potential collisions and avoid them) is a critical requirement. Collision avoidance can be broadly classified into global and local path-planning algorithms, both of which need to be addressed in a successful mission. Whereas global path planning (which is mainly done offline) broadly lays out a path that reaches the goal point, local collision-avoidance algo- rithms, which are usually fast, reactive, and carried out online, ensure safety of the vehicle from unexpected and unforeseen obstacles/collisions. Even though many techniques for both global and local collision avoidance have been proposed in the recent literature, there is a great interest around the globe to solve this important problem comprehensively and effi- ciently and such techniques are still evolving. This paper presents a brief overview of a few promising and evolving ideas on collision avoidance for unmanned aerial vehicles, with a preferential bias toward local collision avoidance. I. Introduction U NMANNED aerial vehicles (UAVs) are expected to be ubiquitous in the near future [1], autonomously perform- ing complex military and civilian missions such as reconnaissance, environmental monitoring, border patrol, search and rescue operations, disaster relief, traffic monitoring, etc. Many of these applications require the UAVs to fly at low altitudes in proximity with man-made and natural structures. A collision with a stationary structure or another UAV could prove to be potentially fatal, and might even result in mission failure.Therefore, it is required that the UAV must be able to successfully sense and avoid obstacles and at the same time look ahead to pursue its goal. This requires robust and computationally feasible collision-avoidance algorithms to be implemented onboard the UAV. The problem of avoiding collision with obstacles online can be perceived in two ways: as a path-planning problem, and a collision-avoidance maneuvering problem. Both of these problems have been heavily researched upon in recent years. This section describes the path-planning problem and the collision-avoidance problem. A conventional path-planning problem deals with finding a path that connects a specified start configuration to a desired goal configuration and avoids any obstacles along the way. For UAV flight in urban terrain, the obstacles would be buildings, trees, and other such known and static structures. An algorithm that can find the solution to such a path-planning problem is called a planning algorithm, path planner, or guidance algorithm. Consider the environment in Fig. 1. The black solid line represents an obstacle-free path planned beforehand by a planning algorithm. The UAV must now track this path. While following this path, a sensor onboard the UAV detects a new “pop-up” obstacle Received 22 March 2010; revision received 2 December 2010; accepted for publication 6 December 2010. Copyright © 2010 by Anusha Mujumdar and Radhakant Padhi. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission. Copies of this paper may be made for personal or internal use, on condition that the copier pay the $10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code 1542-9423/10 $10.00 in correspondence with the CCC. ∗ Former Project Assistant, Department of Aerospace Engineering, anushamujumdar87@gmail.com. † Associate Professor, Department of Aerospace Engineering, padhi@aero.iisc.ernet.in. 17 Downloaded by INDIAN INSTITUTE OF SCIENCE on June 17, 2014 | http://arc.aiaa.org | DOI: 10.2514/1.49985