Dallal, Ahmed
(2017)
Safety and Convergence Analysis of Intersecting Aircraft Flows under Decentralized Collision Avoidance.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Safety is an essential requirement for air traffic management and control systems. Aircraft are not allowed to get closer to each other than a specified safety distance, to avoid any conflicts and collisions between aircraft. Forecast analysis predicts a tremendous increase in the number of flights. Subsequently, automated tools are needed to help air traffic controllers resolve air born conflicts. In this dissertation, we consider the problem of conflict resolution of aircraft flows with the assumption that aircraft are flowing through a fixed specified control volume at a constant speed. In this regard, several centralized and decentralized resolution rules have been proposed for path planning and conflict avoidance. For the case of two intersecting flows, we introduce the concept of conflict touches, and a collaborative decentralized conflict resolution rule is then proposed and analyzed for two intersecting flows. The proposed rule is also able to resolved airborn conflicts that resulted from resolving another conflict via the domino effect. We study the safety conditions under the proposed conflict resolution and collision avoidance rule. Then, we use Lyapunov analysis to analytically prove the convergence of conflict resolution dynamics under the proposed rule. The analysis show that, under the proposed conflict resolution rule, the system of intersecting aircraft flows is guaranteed to converge to safe, conflict free, trajectories within a bounded time. Simulations are provided to verify the analytically derived conclusions and study the convergence of the conflict resolution dynamics at different encounter angles. Simulation results show that lateral deviations taken by aircraft in each flow, to resolve conflicts, are bounded, and aircraft converged to safe and conflict free trajectories, within a finite time.
The proposed rule is powerful when the pilots of the collaborating aircraft, resolving a potential conflict, are either humans or robots. However, when a human pilot is collaborating with a robot pilot for the first time, the robot control should optimize its control criteria to collaborate well with the human. Basically, the robot must adapt its output as it learns from the human. We study the situation of the Human-in-the-Loop, assuming that human will follow lateral maneuvers for conflict resolution. We model the human as an optimal controller that minimizes the risk of collision. Based on that model, we use differential game analysis to get an accurate estimate for the human control criteria. We propose a new algorithm, based on least square minimization, to estimate the Kalman gain of the human's model, and therefore accurately estimate his optimal control criteria. Simulations of this learned rule show that robot pilot can successfully learn from the human pilot actions, and both can cooperate successfully to resolve any conflicts between their aircraft.
Share
Citation/Export: |
|
Social Networking: |
|
Details
Item Type: |
University of Pittsburgh ETD
|
Status: |
Unpublished |
Creators/Authors: |
|
ETD Committee: |
|
Date: |
26 September 2017 |
Date Type: |
Publication |
Defense Date: |
13 July 2017 |
Approval Date: |
26 September 2017 |
Submission Date: |
19 July 2017 |
Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
Number of Pages: |
112 |
Institution: |
University of Pittsburgh |
Schools and Programs: |
Swanson School of Engineering > Electrical Engineering |
Degree: |
PhD - Doctor of Philosophy |
Thesis Type: |
Doctoral Dissertation |
Refereed: |
Yes |
Uncontrolled Keywords: |
Air traffic management, Conflict resolution, Decentralized control, Lyapunov analysis, Human-machine interaction |
Date Deposited: |
26 Sep 2017 20:28 |
Last Modified: |
26 Sep 2017 20:28 |
URI: |
http://d-scholarship.pitt.edu/id/eprint/32789 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
|
View Item |