School of Aeronautics and Astronautics Aircraft Conflict Resolution: Aircraft Conflict Resolution: A Stochastic Optimal Control Approach Inseok Hwang (with Wei Liu) Fli ht D Flight Dynamics and Control/Hybrid System Laboratory i d C t l/H b id S t L b t School of Aeronautics and Astronautics Purdue University Workshop for Aerospace Decision and Control, Georgia Tech Atlanta, GA , June 12, 2012 Air Traffic to and from Atlanta Airport movie Outline • Background and motivations k d d • Aircraft Aircraft dynamics and wind dynamics: Stochastic Differential dynamics and wind dynamics: Stochastic Differential Equations (SDEs) • Problem formulation: Stochastic Optimal Control • Solving the optimal control problem using approximating Markov Chain. • Simulations Background • Current Air Traffic Control (ATC) – Centralized control by ground ATC systems – Rigid and hierarchical structure – Not well scaled up to ever increasing air traffic demand • Next Generation Air Transportation System (NextGen) – Collaborative Collaborative decision making: flat, net decision making: flat net centric control – Aircraft are allowed to coordinate with each other – Safety enhancement by safe separation assurance • Conflict Detection and Resolution ( (CD&R) ) [1] FACET: NASA national aerospace modeling and simulation software [2] http://innopedia.wikidot.com/nextgen‐defined Aerospace Related Research Topics • Air Traffic Control Air Traffic Control – Air traffic surveillance • Multi‐aircraft tracking and identity management • Multi‐sensor data fusion and fault detection – – – – – – Separation assurance: conflict detection and resolution Trajectory based operation: trajectory prediction and ETA computation Safety through conformance monitoring (airspace and airport) in ATC y g g( p p ) Dynamic Airspace Configuration (en‐route and terminal airspace) Arrival flight scheduling and sequencing Performance metrics for NAS • UAS applications – Modeling and simulations using open source environment (Mixed Airspace) – Autonomous navigation and control of UAS (e.g., sense Autonomous navigation and control of UAS (e g sense‐and and avoid) avoid) – Cyber‐secure UAS design and control • Other applications – Mode Mode awareness (mode confusion) using FDIR for (Stochastic) Hybrid Systems awareness (mode confusion) using FDIR for (Stochastic) Hybrid Systems – Space applications: non‐Keplerian spacecraft tracking and low‐thrust orbit transfer Aerospace Related Research Topics • Air Traffic Control Air Traffic Control – Air traffic surveillance • Multi‐aircraft tracking and identity management • Multi‐sensor data fusion and fault detection – – – – – – Separation assurance: conflict detection and resolution Trajectory based operation: trajectory prediction and ETA computation Safety through conformance monitoring (airspace and airport) in ATC y g g( p p ) Dynamic Airspace Configuration (en‐route and terminal airspace) Arrival flight scheduling and sequencing Performance metrics for NAS • UAS applications – Modeling and simulations using open source environment (Mixed Airspace) – Autonomous navigation and control of UAS (e.g., sense Autonomous navigation and control of UAS (e g sense‐and and avoid) avoid) – Cyber‐secure UAS design and control • Other applications – Mode Mode awareness (mode confusion) using intent inference awareness (mode confusion) using intent inference – Space applications: non‐Keplerian spacecraft tracking and low‐thrust orbit transfer Aircraft Conflict Resolution: Overview • Two Two kinds of midair conflict: Aircraft kinds of midair conflict: Aircraft‐weather weather and Aircraft and Aircraft‐aircraft aircraft conflict • Most conflict resolution algorithms are to d fl l l h design conflict‐free trajectories for aircraft to follow (open‐loop control) – Easy to understand and compute – Easy to implement – Could be inefficient and/or unsafe if aircraft deviate from the optimal if aircraft deviate from the optimal Conflict‐free trajectories Closed‐loop Optimal Control Algorithm [1] Figure source: THALES Proposed Aircraft Conflict Resolution • We We propose to use a closed‐loop propose to use a closed loop technique to solve this problem which technique to solve this problem which can explicitly account for uncertainties (such as wind) Cockpit display of Traffic Collision Avoidance System (TCAS) Nearby aircraft positions Suggested climb gg rate Stochastic System Dynamics • Aircraft and Wind dynamics in the horizontal plane Aircraft and Wind dynamics in the horizontal plane – Aircraft: : aircraft position; : aircraft heading (control input); : cruise speed; : wind velocity vector – Wind: : deterministic wind velocity; : intensity of uncertainties in wind dynamics; : Brownian motion y ; • Combined Dynamics (Stochastic Diff. Eq.) Deterministic wind field Aircraft‐Weather Conflict Resolution • Aircraft‐weather conflict resolution as an optimal control problem Aircraft weather conflict resolution as an optimal control problem – Given aircraft and wind dynamics – We want to drive the aircraft to the target set without entering the target set without entering the forbidden or exit the whole area. – Cost 1 Running cost Exit cost where is a stopping time Design to minimize the cost , subject to the aircraft and wind dynamics r(x) takes high value on boundaries of the forbidden sets and low value on boundary of the and low value on boundary of the target set. Aircraft dynamics Feedback control Feedback Stochastic Optimal Control Problem • Optimal control problem O ti l t l bl – Dynamics: – Minimize: • Hamilton‐Jacobi‐Bellman (HJB) equation – Cost‐to‐go: starting from a point , Cost to go: starting from a point – HJB equation: Approximating Markov Chain • SSolving the stochastic optimal control: a Markov Chain (MC) l i h h i i l l M k Ch i (MC) approach – Discretize the state space using grids: – Design the transition probabilities – If the convergence relations hold, the MC converges to the SDE N(q) : Neighbors of q Jacobi Iteration for Optimal Control of Markov Chain – Optimal control problem: • Minimize: Cost-to-go C tt off th the Markov chain First time when the Markov chain exit the domain Running cost – HJB equation (discrete version): Exit cost This HJB can be solved numerically! – Jacobi iteration (similar to gradient descent): J bi it ti ( i il t di t d t) Convergence relation Numerical Simulations • Aircraft‐weather avoidance scenario Ai ft th id i Cost‐to‐go computed by Jacobi iteration Optimal control input (aircraft heading) and 50 aircraft trajectories Aircraft‐Aircraft Conflict Resolution • Two‐aircraft dynamical model • Forbidden set: • Optimal control problem: Optimal control problem: – Choose to minimize: – takes high value on the boundary of Aircraft‐Aircraft Conflict Resolution Numerical Simulation Results • Optimized cost‐to‐go and optimal control Optimized cost to go and optimal control input to Aircraft 1 when Aircraft 2 is in different locations When Aircraft 2 is outside S0, Aircraft 1’s y control is not affected by Aircraft 2. When Aircraft 2 is inside S0, Aircraft 1’s control is affected by Aircraft 2 control is affected by Aircraft 2. Numerical Simulation Results • Optimal cost function and optimal control input Optimal cost function and optimal control input Multi‐aircraft Conflict Resolution • 4‐aircraft conflict resolution scenario f fl l Computational Complexity Analysis • The computational complexity of the proposed algorithm increases h l l f h d l h rapidly as the number of aircraft involved increases (curse of dimensionality) • To overcome the rapid glowing computation time, a decomposition method is proposed method is proposed • Comparison of the computation time (sec) of the conflict resolution algorithm with and without decomposition (using MATLAB) Summary • The aircraft and wind dynamics are modeled by Stochastic Differential Equations (SDE) • The conflict resolution problem is formulated as the optimal control of SDE • To solve the optimal control problem, a Markov chain is constructed to approximate the SDE approximate the SDE • The performance of the proposed conflict resolution algorithm has been demonstrated with illustrative conflict scenarios • Future work: – Using Differential Transformation to further improve computational complexity to better accommodate high fidelity models and complex scenarios I. II. Hwang, J. Li, and D. Du, “Differential Transformation and Its Application to Nonlinear Optimal Control,” ASME Journal of Dynamic Systems, Measurement, and Control, Vol. 131 (5), September 2009 R. Huang, I. Hwang, and M. Corless "A Nonlinear Algorithm for Tracking Interplanetary Low‐Thrust Trajectories," AIAA Journal of Guidance, Control and Dynamics, To appear
© Copyright 2026 Paperzz