Unmanned Aerial Vehicle Dynamic-Target Pursuit by Using Probabilistic Threat Exposure Map

Abstract

A strategy is presented for an unmanned aerial vehicle (UAV) to follow a moving target in an area the probabilistic threat exposure map of which is assumed to be known based on a priori data. A probabilistic threat exposure map is defined to be the risk of exposure to multiple sources of threat as a function of position. The strategy generates speed and heading angle commands within the dynamic constraints of the UAV. There are three main objectives in order of priority: 1) All of the restricted areas are avoided. 2) The UAV stays within the proximity of the target by a prespecified distance. 3) The total threat exposure level is minimized. During the pursuit, the heading and speed of the target and their time variations are not directly measured but are estimated from the measurements of the target positions. If, for any reason, the sensor can no longer measure the current position of the target, the strategy starts using the predicted target states based on the past measurements to guide the UAV toward the proximity of the target until the UAV detects the target again. I. Introduction I N this paper, a target following strategy is introduced for an unmanned aerial vehicle (UAV) flying through an area of multiple sources of threat that are modeled as a probabilistic threat exposure map (PTEM). The PTEM is a map that indicates the threat level of an area due to different types of static threat sources using probability density functions. Recently, there has been an increasing interest in probabilistic approaches in mission planning for the UAVs. This is because the probabilistic approaches are inherently very suitable to handle the uncertainty in the information, such as the locations of the threats. There are several papers in the literature using various probabilistic approaches to deal with the path-planning problem of the UAV applications based on the probabilistic map of the area of operation. 1,2 In Refs. 1 and 2, various path-planning strategies are proposed to minimize the level of threat exposure while flying to a stationary target through an area of multiple threats. This risk of exposure to a source of threat is a function of position, defined to be the probability of becoming disabled by the source of threat at a given position. The probability is assumed to have a Gaussian distribution over the area of operation. The probabilistic threat exposure map is constructed from the probability distribution functions of all of the sources of threat in the area. In Ref. 3, a graph-based probabilistic approach is developed to use the probabilistic map of the area. Unlike the Voronoi graph-based approaches, the nodes and links of the graph are based directly on the probabilistic map. The region of operation is divided into cells whose occupancy value is determined based on the sensor readings. By the application of the conditional probability of occupancy using Bayes rule and the Bellman‐Ford algorithm, the shortest path is found. Because the path-planning strategies rely on probabilistic maps, the construction and online update of the maps are very crucial. In Ref. 4, a probabilistic map of an area with multiple distinguishable moving obstacles is built by using Bayesian estimation, and