Deep reinforcement learning for carrier-based aircraft flight deck operations scheduling problem.

Area of Science:

• Operations Research
• Artificial Intelligence
• Computer Science

Background:

• Flight deck operations scheduling is a complex, NP-hard problem.
• Traditional methods struggle with the efficiency-solution quality trade-off.

Purpose of the Study:

• To develop an advanced AI framework for optimizing flight deck scheduling.
• To overcome limitations of existing computational methods.

Main Methods:

• A deep reinforcement learning framework integrated with graph neural networks.
• Formulation as a Markov decision process for agent-based scheduling.
• Utilized a softmax exploration strategy with a discount factor of 1.0.

Main Results:

• The AI agent significantly improved solution quality over priority dispatching rules.
• Achieved competitive performance against meta-heuristics on small-scale problems.
• Demonstrated superior search capabilities on large-scale instances.
• Reduced scheduling decision time from minutes to seconds.

Conclusions:

• The proposed deep reinforcement learning framework offers a superior approach to flight deck scheduling.
• This AI solution meets real-time operational demands with high-quality, efficient schedules.