A Multi-Agent Simulation Method for Air Taxi Combining Double-DQN Network
By combining Double-DQN networks and hierarchical finite state machines, an air taxi scheduling model is constructed, which solves the shortcomings of existing air taxi simulation methods, realizes detailed simulation and optimized scheduling of individual air taxi behavior, and improves the realism of simulation results and system efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-06-03
- Publication Date
- 2026-06-30
AI Technical Summary
Existing multi-agent simulation methods for air taxis neglect the low-altitude environment and the individual behavior of air taxis, making it difficult to adaptively learn and optimize scheduling strategies. They also cannot effectively cope with dynamic order allocation and multi-vehicle collaboration, resulting in significant deviations between simulation results and reality.
An air taxi scheduling model is constructed using a Double-DQN network. Combined with a hierarchical finite state machine and a composite urban air taxi traffic network model, the model simulates the full lifecycle behavior of air taxis and passenger behavior at stations in detail. The model is then trained and optimized using a multi-agent simulation system.
It significantly improves the realism and credibility of simulation results, can autonomously learn and optimize scheduling strategies, improve system operating efficiency, and provides multi-dimensional statistical indicators and visualization results to support decision-makers in analysis and planning.
Smart Images

Figure CN122311985A_ABST