面向空管模拟机培训的智能应答机长研究

针对当前日益增长的空中交通管制员(以下简称“管制员”)培训需求以及传统管制员培训模拟机面临的效率低等问题,设计了 1个面向空管模拟机培训的智能应答机长系统。该系统利用语音识别、指令提取、指令复诵、语音合成等技术,能够实现对管制模拟培训过程中管制员语音的智能识别和理解,并模拟飞行员自动输出复诵指令的功能。通过对真实空管对话语音模式的研究和分析,制定了 1套详细的复诵规则,以适应不同场景下的管制指令复诵模式。此外,集成了特情处理模块以支持管制员特情处理培训。在真实管制培训环境下进行实验验证,结果表明,所提出的智能应答机长系统综合复诵准确率为 88.6%,可以有效提升管制员培训质量和效率,显著降低了人力成本。并且,该系统可以作为子系统集成到现有的管制员培训模拟机系统中,具有较强的便捷性和兼容性。

Automatic Pilot Agent for ATC Simulator Training

Current Air Traffic Controllers(ATCO) training systems require human pseudo-pilot to complete the training procedure, which decreases the efficiency to controller simulation training. In this work, an automatic pilot agent(APA) framework is proposed to empower the controller simulation training systems. The proposed APA framework consists of four modules, including automatic speech recognition(ASR), controlling instruction understanding(CIU), controlling in-struction repetition(CIR), text-to-speech(TTS). Furthermore, a set of repeating rules are designed to adapt the various flight phase and ATC scenarios according to the ATC rules. In addition, the flight emergency handling module is integrat-ed in the proposed APA framework to improve the ability of the ATCO to handle the emergency situation intentionally and strategically. The experimental result shows that the proposed APA framework achieves desired performance, reaching 88.6% repetition accuracy, which can greatly improve controller training quality and efficiency, reduce human cost. In ad-dition, the proposed APA framework can integrate into the existing controller training simulator system as subsystem, which has strong convenience and compatibility.