Show simple item record

dc.contributor.advisorKeshmiri, Shahriar
dc.contributor.authorEsposito, Julien Florian
dc.date.accessioned2013-09-29T13:33:58Z
dc.date.available2013-09-29T13:33:58Z
dc.date.issued2013-08-31
dc.date.submitted2013
dc.identifier.otherhttp://dissertations.umi.com/ku:12907
dc.identifier.urihttp://hdl.handle.net/1808/12164
dc.description.abstractThe motivation for the research presented in this dissertation is to provide a two-fold solution to the problem of non-cooperative reactive mid-air threat avoidance for fixed-wing unmanned aerial systems. The first phase is an offline UAS trajectory planning designed for an altitude-specific mission. The second phase leans on the results produced during the first phase to provide intelligent, real-time, reactive mid-air threat avoidance logic. That real-time operating logic provides a given fixed-wing UAS with local threat awareness so it can get a feel for the danger represented by a potential threat before using results produced during the first phase to require aircraft rerouting. The first original contribution of this research is the Advanced Mapping and Waypoint Generator (AMWG), a piece of software which processes publicly available elevation data in order to only retain the information necessary for a given altitude-specific flight mission. The AMWG is what makes systematic offline trajectory possible. The AMWG first creates altitude groups in order to discard elevations points which are not relevant to a specific mission because of the altitude flown at. Those groups referred to as altitude layers can in turn be reused if the original layer becomes unsafe for the altitude range in use, and the other layers are used for altitude re-scheduling in order to update the current altitude layer to a safer layer. Each layer is bounded by a lower and higher altitude, within which terrain contours are considered constant according to a conservative approach involving the principle of natural erosion. The AMWG then proceeds to obstacle contours extraction using threshold and edge detection vision algorithms. A simplification of those obstacle contours and their corresponding free space zones counterparts is performed using a fixed -tolerance Douglas-Peucker algorithm. This simplification allows free space zones to be described by vectors instead of point clouds, which enables UAS point location. The resulting geometry is then processed through a vertical trapezoidal decomposition where for each vertex defining a contour a vertical line is drawn, and the results of this decomposition is a set of trapezoidal cells. The cells corresponding to obstacle contours are then removed from the original trapezoidal decomposition in order to solely retain the obstacle-free trapezoidal cells. After decomposition, cells sharing part of a common edge are considered from a graph theory perspective so it becomes possible to list all acyclic paths between two cells by applying a depth first search (DFS) algorithm. The final product of the AWMG is a network of connected free space trapezoidal cells with embedded connectivity information referred to as the Synthetic Terrain Avoidance (STA network). The walls of the trapezoidal cells are then extruded as the AWMG essentially approximates a three-dimensional world by considering it as a stratification of two-dimensional layers, but the real-time phase needs 3D support. Using the graph conceptual view and the depth first search algorithm, all the connected cell sequences joining the departure to the arrival cell can be listed, a capability which is used during aircraft rerouting. By connecting two adjacent cells' centroids to their common midpoint located on the shared edge, the resulting flying legs remain within the two cells. The next step for paths between two cells is to be converted into flyable paths, and the conversion uses main and fallback methods to achieve that. The preferred method is the closed-form Dubins paths method involving the design of sequences of arc circle-straight line-arc circle (CLC) in order to account for the minimum radius turn constrain of the UAS. An additional geometric transformation is developed and applied to the initial waypoints used in the Dubins method so the flying leg directions are respected which is not possible by using the Dubins method alone. When consecutive waypoints are too close from one another, a condition called the Dubins condition cannot be respected, and the UAS trajectory design switches to the numerical integration of a system of ordinary differential equations accounting for the minimum turning constraint. Using the Dubins method and the ODE method makes it possible for the AWMG to design flyable offline trajectories accounting for the lateral dynamic of the fixed-wing UAS. The second original contribution of this research is the development and demonstration of the Double Dispersion reduction RRT (DDRRT), an algorithm which employs two new developed logic schemes respectively referred to as Punctual Dispersion Reduction (PDR), and Spatial Dispersion Reduction exploration (SDR). The DDRRT is employed during the real-time in-flight phase where it initially assumes a perfect terrain and no unpredictable threat, consequently following a 100% adaptive goal biasing toward the next waypoint in its list. When a threat such as an unpredicted obstacle is detected, the (PDR) acknowledges the fact that the DDRRT tree branches have met an obstacle and the its goal-biasing toward the next waypoint is decreased. If the PDR keeps decreasing, the DDRRT develops awareness of its surrounding obstacles by relaxing its PDR and switching to SDR which has the effect of increasing the dispersion of its branches, but keeping their extension bounded by the cell containing the last good position of the UAS, Csafe. If a number of branches reach a limit proportional to the Csafe and its relative area, then the STA network is queried for alternative rerouting. The two phases provide real-time reactive mid - air threat avoidance scenarios with the ability for a UAS to develop local and realistic threat awareness before considering intelligent rerouting. Either the local exploration of the DDRRT is successful before reaching a maximum number of points, or the STA Network is required to find another route.
dc.format.extent180 pages
dc.language.isoen
dc.publisherUniversity of Kansas
dc.rightsThis item is protected by copyright and unless otherwise specified the copyright of this thesis/dissertation is held by the author.
dc.subjectAerospace engineering
dc.subjectAircraft
dc.subjectAvoidance
dc.subjectCollision
dc.subjectReal time
dc.subjectUas
dc.subjectUnmanned
dc.titleReal-Time Obstacle and Collision Avoidance System for Fixed-Wing Unmanned Aerial Systems
dc.typeDissertation
dc.contributor.cmtememberTaghavi, Ray
dc.contributor.cmtememberEwing, Mark
dc.contributor.cmtememberDowning, David
dc.contributor.cmtememberAllen, Christopher
dc.thesis.degreeDisciplineAerospace Engineering
dc.thesis.degreeLevelPh.D.
kusw.oastatusna
kusw.oapolicyThis item does not meet KU Open Access policy criteria.
kusw.bibid8086033
dc.rights.accessrightsopenAccess


Files in this item

Thumbnail
Thumbnail
Video
Video

This item appears in the following Collection(s)

Show simple item record