An Evaluation of Stereo and Laser-Based Range Sensing for Rotorcraft Unmanned Aerial Vehicle Obstacle Avoidance • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • Stefan Hrabar Autonomous Systems Laboratory, CSIRO ICT Centre, Australian Research Centre for Aerospace Automation (ARCAA), Brisbane, QLD, 4069, Australia e-mail: stefan.hrabar@csiro.au Received 15 June 2011; accepted 18 October 2011 We present an evaluation of stereo vision and laser-based range sensing for rotorcraft unmanned aerial vehicle (RUAV) obstacle avoidance. Our focus is on sensors that are suitable for mini-RUAV class vehicles in terms of weight and power consumption. The study is limited to the avoidance of large static obstacles such as trees. We compare two commercially available devices that are representative of the state of the art in two-dimensional scanning laser and stereo-based sensing. Stereo is evaluated with three different focal length lenses to assess the tradeoff between range resolution and field of view (FOV). The devices are evaluated in the context of obstacle avoidance through extensive flight trials with an RUAV. We discuss the merits and limitations of each sensor type, including sensing range, FOV, accuracy, and susceptibility to lighting conditions. We show that the stereo device fitted with 8-mm lenses has a better sensing range and vertical FOV than the laser device; however, it relies on careful calibration and is affected by high-contrast outdoor lighting conditions. The laser has a wider horizontal FOV and is more reliable at detecting obstacles that are within a 20-m range. Overall the laser produced superior obstacle avoidance performance, with a success rate of 84% compared to 42% for 8-mm stereo. C  2012 Wiley Periodicals, Inc. 1. INTRODUCTION With their ability to hover in place and climb or descend vertically, rotorcraft unmanned aerial vehicles (RUAVs) are well suited to a number of applications. These include in- specting structures such as power lines, bridges and cool- ing towers, surveillance, search and rescue, and package delivery. Many of these applications require (or would ben- efit from) close-range (up to 35 m) obstacle detection and avoidance capability. We differentiate this from the desired “see and avoid” capability for fixed-wing UAVs, where the goal is to detect other aircraft out to a few kilometers. The sensors and sensing techniques for close-range obstacle de- tection are more closely related to those used on ground- based vehicles. Both stereo vision and scanning laser range finders or LiDARs (light detection and ranging) have been widely used for mapping and obstacle avoidance on ground-based robots for a number of years (henceforth we use the term “laser” or “scanning laser” to refer to these scanning laser range finders). Because cameras are lightweight and rel- atively power-efficient, stereo has been a viable option for use on mini-UAVs for some time. Until recently, how- ever, lasers have been too heavy and power-hungry for use on mini-UAVs. With the launch of products such as the Hokuyo UTM-30LX, it is now feasible to use lasers on this class of UAV. The penalty for the reduced weight and power of these smaller devices is their reduced sens- ing range. The UTM-30LX has a theoretical sensing range of 30m, for example. In many RUAV scenarios this sens- ing range is adequate, as RUAVs can fly slowly or hover in place. It is also comparable to the sensing range of a com- pact stereo pair. Given that these two range-sensing options are now available for use on this class of RUAV, we have investigated the relative merits of each. This has been done through extensive flight testing of obstacle avoidance sce- narios with these sensors mounted on an autonomous heli- copter platform. The flight trials were conducted with a number of goals in mind. We hoped to establish a level of confidence in the ability of each sensor type to detect a variety of ob- stacles under real-world conditions. The published specifi- cations for scanning lasers do not take into account factors such as vibration and lower-frequency motion of an aerial platform, reflective properties, geometry, or sparseness of the obstacles. The natural attitude and altitude variations of an aerial platform effectively give the sensor a larger “vir- tual” FOV, for example, and we hoped to quantify this ef- fect. Although it is possible to produce three-dimensional (3D) laser scans by actuating a two-dimensional (2D) scan- ner (rotating, nodding, etc.), we limit our investigation to a rigidly mounted 2D scanner. The actuation mechanisms add additional weight, which could make the devices too heavy for mini-RUAVs. Journal of Field Robotics 29(2), 215–239 (2012) C  2012 Wiley Periodicals, Inc. View this article online at wileyonlinelibrary.com • DOI: 10.1002/rob.21404