期刊名称：International Journal on Electrical Engineering and Informatics
印刷版ISSN：2085-6830
出版年度：2018
卷号：10
期号：3
页码：542-579
DOI：10.15676/ijeei.2018.10.3.9
出版社：School of Electrical Engineering and Informatics
摘要：Two approaches of collision avoidance (CA) control design are discussed: thestochastic process -based CA and the reactive CA. We believe that the well-establishedstochastic process -based CA that has been widely used in general aviation flights may not workwell in UAV flights for at least two reasons: the difference in encounter characteristics, and thedifference in available resource provisions. Problems on reactive CA and search-based CA aremostly simplified to two-dimensional flight cases and depend heavily on non-partisan observerin providing motion data to the onboard controller, thus resulting CA control systems that areever-dependent to external sensory resources. This shortcoming can potentially be solved usingkinematic-based collision threat situation (CTS) model. Existing CTS models using ‘collisioncone’ or velocity obstacle (VO) approach are discussed. These models reveal the existence of aninfinite number of evasion planes for a given initial threat situation, which requires the CAcontroller using such approach to search for a plane that provides the most efficient evasivemaneuver, which in turn requires more computing power and time. To overcome theseshortcomings, we propose a CTS model that is based on the kinematic relation between pair ofbodies involved in a CTS. We also define the state of the CTS model and construct the CAcontroller to reduce the value of the state of the CTS. Since the CTS is evaluated in relativemotion context between the bodies, the resulting model is readily compatible with output datafrom onboard sensors, eliminating the need to perform coordinate transformation that will in turnimproving the computational efficiency of the whole CA control system. Furthermore, the modelalso provides us a deterministic evasion plane, thus eliminating the need to perform a searchprocedure. The performance of our CA control design is evaluated using energy-based functionand a series of simulations.
