London, UK, May 29-31, 2012 ATACCS’2012 | DOCTORAL CONSORTIUM 229 Obstacle Tracking Results: Cartesian vs. Spherical Particle Filter Anna Elena Tirri, Domenico Accardo, Giancarmine Fasano, and Antonio Moccia Department of Aerospace Engineering (DIAS) Università degli Studi di Napoli “Federico II” Piazzale Tecchio 80, 80125 Naples Italy +39 0817682149 annaelena.tirri@unina.it ABSTRACT This paper focuses on test results from an Airborne Obstacle Tracking system for Unmanned Aerial System (UAS) See and Avoid applications that is based on Particle Filtering algorithm. It performs data fusion of airborne forward looking radar and electro-optical camera by exploiting data gathered during a Sense and Avoid flight experiment at Italian Aerospace Research Centre (CIRA). The developed model resulted adequate for tracking aircraft trajectories, thus overcoming the non-gaussian and non- linear form of the most widely adopted target dynamics models. Keywords Unmanned Aerial Systems, Sense and Avoid, Multi-Sensor Data Fusion, Particle Filtering. INTRODUCTION Recently, the growing interest in UAS has required the introduction of standards and specifications to allow UAS to fly in civilian, non-segregated airspace alongside manned aircraft. Regulatory agencies prescribe UAS to be equipped with a robust and reliable Detect, Sense and Avoid System (DS&A) in order to guarantee the safety of other objects in the flight path [4-13]. Several solutions have been proposed to attain this function [5,9,6]. In particular, an integrated radar/electro-optical configuration has been selected as the most adequate solution to attain this function within the framework of the CIRA/UNINA TECVOL project. Due to the presence of multi-sensor architecture and non-linearity of the considered dynamics, innovative methodologies are needed to overcome assessed data fusion limitations [8]. In fact, assessed EKF techniques have a series of drawbacks, such as: the linearization can produce highly unstable filters if the assumption of local linearity is violated, and the derivation of the Jacobian matrices are nontrivial in most applications and often lead to significant implementation difficulties [11]. Indeed, the relative flight dynamics is properly represented by a non-linear model in the maneuvering conditions. The EKF can be viewed as providing “first-order” approximations to the optimal terms and these approximations can introduce large errors in the true posterior mean and covariance of the transformed Gaussian random variables that represent the target state; this may lead to sub-optimal performance and sometimes divergence of the filter. In addition, EKF requires both a precise system model and the statistical property of the noise to achieve accurate performance. However, model uncertainty and incomplete statistical information are often encountered in real applications and make it difficult to precisely estimate the system states, leading to a very large estimation error. The paper focuses on test results from a Particle Filter (PF) obstacle tracking system. The software performance are analyzed in order to point out the potential impact of the technique under investigation on obstacle tracking capabilities in terms of accuracy and reliability with respect to an Extended Kalman Filter (EKF) that was developed in the initial stage of the project. A more detailed presentation of the developed EKF solution and the relevant flight test results can be found in Ref. no. 6. The developed PF tracking algorithm has been implemented in off-line simulations performed on real flight data. In particular, the software has been tested by exploiting both dynamics models in Cartesian and Spherical reference systems taking into account only radar measurements. AIRBORNE OBSTACLE TRACKING SYSTEM An Airborne Obstacle Tracking system is the core of a DS&A system. In fact, information about target’s position as well as target motion is considered mandatory for collision avoidance logic in order to decide whether or not an evasive maneuver must be performed. Flight regulations about mid air collision avoidance prescribe that the minimum distance between two planes must never be less than a bubble distance of about 170 m [14]. Thus, a suitable DS&A system must be able to estimate whether the Distance at Closest Point of Approach (DCPA) between the own aircraft and the intruder will be less than the bubble distance. In case this condition is Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Re-publication of material on this page requires permission by the copyright owners. ATACCS’2012, 29-31 May 2012, London, UK. Copyright 2012 IRIT PRESS, ISBN: 978-2-917490-20-4