Optimum Airspace Partitioning for CenterSector Boundary Design Arash

Optimum Airspace Partitioning for Center/Sector Boundary Design Arash Yousefi George L. Donohue Research Sponsors: NASA ARC, FAA, Metron Aviation Inc. 1 st International Conference on Research in Air Transportation - ICRAT 2004, November 22 -24 2004, Zilina, Slovakia

Current Sectorization Has Historical – Not Analytical Origins

Traffic Is Not Uniformly Distributed Among ARTCCs – Productivity Overhead Concern Source: FAA Factbook, March 2004. URL: http: //www. atctraining. faa. gov/factbook

Given: Demand Profiles and Airport locations Desired: Optimum Center/sector Boundaries?

Optimization Parameter: ATC Workload (Modeling) Ø ATC workload is divided to 4 variables 1. 2. 3. 4. Ø Horizontal Movement Workload (WLHM), Conflict Detection and Resolution Workload (WLCDR), Coordination Workload (WLC), Altitude-Change Workload (WLAC). In each sector or volume of airspace during a given time-interval: More details: Yousefi, A. , Donohue, G. L. , and Qureshi, M. Q. , “Investigation of En route Metrics for Model Validation and Airspace Design”, Proceeding of the 5 th USA/Europe Air Traffic Management R&D Conference, Budapest, Hungary, June 2003.

Airspace Partitioning for Optimum Boundary Definition § § § Airspace of 20 CON US ARTCCs is divided to three altitude layers with 2, 566 cells. Disregarding the existing Center and sector boundaries. Hex-Cells are airspace elements and we compute complexity and workload metrics for each cell based on historic flight data and simulation. 1. Large enough to capture conflicts 2. Small enough for enough resolution 24 nm=0. 4 degree lat/long over FL 310 FL 210 -FL 310 below FL 210

Hexagonal Grid Selection Criteria § Common sides between hex-cells within a cluster. § Computationally less expensive than triangle. § Avoid the acute and right angles in triangle & rectangle that may result to short transit times for aircraft passing close to the edges.

Optimum Airspace Design Process Data Pre-processing Simulation/Optimization Post-processing & visualization Create hex-cell mesh § In 3 layers § 2, 566 in each layer Actual traffic from ETMS § Last Filed routes TAAM Simulation Traffic variables Airspace Complexity Visualizer (ACV) Defining design-period § ~45 K daily flights Create seeds for potential sectors WL calculation for each hex-cell for 10 min bins Optimization Hex-cell assignments Representation of new sector boundaries

TAAM Simulation § ~45 K Daily Flights from ETMS § Last Filed routes § Run Time=8. 5 hrs

WL Trend Throughout the Day Low altitude layer High altitude layer

Defining a Design-Period Design Period

Clustering Hex-cells to Construct sectors/Centers

Clustering Algorithm for ARTCC Boundary Design Given: Demand profile and location of current ARTCCs Ø Desired: What are the best ARTCCs to be opened and what is the best boundary? Ø MIN (variation of workload among ARTCCs) MIN (SUM of distances from each hex-cell to current Center locations) MIN (Maximum distance between the hex-cell and the seed) SUBJECT TO: § avoiding highly concave ARTCCS § number of ARTCCs are given § some other ordinary constraints (e. g. assignment of each hexcell to a single ARTCC, etc)

Locational Analysis & Facility Location Problems GIVEN: - I = {1, . . . , n} set of candidate locations for facilities - J = {1, . . . , m} set of demand points demand point Not opened Candidate location for facility

Clustering Algorithm for ARTCC Boundary Design d max d 4 d 1 d 2 Hex-cell center i d 3 d 5 Seed j

MINIMIZE (variation of workload among ARTCCs)

MINIMIZE (SUM of distances from each hex-cell to the seed)

MINIMIZE (Max distance between the hex-cell and the seed)

ARTCC Boundary Re-design (Keeping 20 Centers, Changing the boundaries)

ARTCC Boundary Re-design (Keeping 20 Centers, Changing the boundaries) ABQ

Reducing # of ARTCCs to 18

Reducing # of ARTCCs to 5 ABQ JFK, WL=58, 760 -Optimization 1 - MIN WL Variation & 2 - MIN SUM distance & 3 - MIN MAX distance

Reducing # of ARTCCs to 4 ABQ -Optimization 1 - MIN WL Variation & 2 - MIN SUM distance & 3 - MIN MAX distance

Clustering Algorithm For Sector Design § Given Optimum Center Boundaries, Find the Optimum Sector Boundaries § Similar to Center Boundary problems § Combinatorial minimization problem MIN (variation of workload among sectors) SUBJECT TO: § sector contiguity § avoiding highly concave sectors § number of sectors is limited § avoid extremely large sectors § some other ordinary constraints (e. g. assignment of each hex-cell to a single sector, etc)

Conclusion & Future Work Ø Clustering algorithms appear to produce reasonable results both for Center and Sector boundary design Ø Result is Formally an Optimum Solution for Chosen Object Function Optimization approach allows additional constraints (radar coverage, avoiding large airports close to boundaries, etc) Ø Cost - Benefit analysis for selection of best ARTCCs should be done (if goal is Overhead Reduction) Ø Extension of sectorization process for each altitude layer within each ARTCC Ø Using Com or Nav Aids as seeds or put the seeds along the major traffic flow paths Ø One could use RAMS or FACET instead of TAAM Ø NOTE: As an academic research, so far the intention has been to develop a partitioning METHODOLOGY. Future IV&V and cost benefit analysis are essential