Aviation Management
The handling airport and flight traffic, and ensuring smooth flux of passengers and operations
Aviation Processes
Aviation is composed of several processes that work in tandem, which are
- Flight Scheduling
- Fleet Assignment
- Aircraft Routing
- Crew Scheduling
- Gate Scheduling
- Bus Planning
- Personnel Planning All of these are complex problems
Analytical Components of Aviation Management
From an analytical perspective, there are three components of aviation management
flowchart LR
A["Prediction"] --> B["Optimization"]
B["Optimization"] --> C["Control"]
Think of these components as abstractions of the processes mentioned above. Almost all of the aviation processes fall under one or more of these abstract categories
Prediction
Finding patterns in past data and using them to make predictions about the future. Algorithms are used to fit models to historical data and trends. The models predict and estimate future trends. Prediction is explicitly data and model-based
Optimization
Finding values for variables that optimize a function under constraints. The idea is to select a cost function (often, a real-world problem that requires solving of some sort), model it mathematically, and then attempt to dial the “knobs” of the function just the right way to ensure the best outcome
| Standard Methods | Heuristics |
|---|---|
| Mathematical Optimization | Genetic Algorithms |
| Linear Programming | Greedy Heuristics |
| Convex Programming | |
| Quadratic Programming | |
| Mixed-Integer Programming | |
| Standard methods are mathematically defined and efficient while heuristics aren’t. However, heuristics are more flexible and general. One can apply heuristics where fixed mathematical approaches may not work (for example, when the problem statement itself isn’t rigidly defined or modelled) | |
| Delay Management is a good application domain of optimization |
Control
Given the current state of an aviation process and the expected state, control algorithms ensure optimal control of the system, keeping it constantly and automatically maintained
Reinforcement Learning, LQG Design, Agent-based Simulation, and Dynamic Programming are examples of various control methods
See UAM - Flight Trajectory Control and Optimization
How to do all this well?
- Proper visualization is needed
- Abstract the problems and challenges
- Use different perspectives and approaches
- Prioritize accessibility