AcListant Validation Plan

Abstract

The project AcListant will show for Düsseldorf airport and its airspace, how automatic speech recognition (ASR) and an Arrival Manager system may benefit from each other. The Arrival Manager gets an additional input sensor. ASR receives context information to significantly reduce the word and concept error rate. A first prototype with limited functionality was already demonstrated at the ATC global fair in Amsterdam in 2011. According to the six levels of maturity of the E-OCVM methodology [2] The validation project AcListant will validate whether the prototype can be classified as V4 (Pre-operational), i.e. an industrial prototype has been built up, which is ready for implementation. The current document presents the validation plan for these validation exercises. For this purpose 5 high-level validation goals are identified: 1. The usage of the AMAN 4D-CARMA improves approach control in terms of arrival predictability and optimal landing sequence. 2. An additional sensor (mouse or keyboard input) speeds up the plan adaptation process of the AMAN. 3. Automatic speech recognition enhances both AMAN performance and the quality of controller work regarding performance improvement (e.g. throughput, vertical flight path) and workload reduction. 4. Context information created by the assistant (AMAN) improves ASR performance. 5. Adaptation of ASR components to the ATC context (e.g. model for non-native speakers, digit recognizer, robustness concerning out grammar utterances) improves ASR performance