Multimodal AI for location and route prediction
AURORA aims to predict locations and routes using artificial intelligence. Its approach combines historical GPS data with contextual information from voice recordings in vehicles. Movement patterns and spoken information are intended to be analyzed together to help estimate future trajectories and possible destinations.
The research approach includes machine learning and deep learning alongside speech recognition, speaker identification and semantic analysis. The project description names Named Entity Recognition as one method for extracting context from speech data. The central task is to connect this information with movement histories.
Combining the two data types is intended to support prediction even when movement histories are short. Adaptation to dynamic traffic situations is part of the described approach. Better predictions and more resource-efficient mobility are intended benefits, not demonstrated outcomes of a real-world traffic application presented here.
Research topics
- ML- and DL-based methods for location and route prediction from GPS data.
- Semantic analysis of speech context and its integration into prediction models.
- Multimodal learning methods for combining different data sources.
- Analysis of prediction accuracy, robustness and the trustworthiness of AI models.
