Multi-modal, Multi-source Data Quality Models, Enhancement Mechanisms, and Quality-Aware Data Transformation for The Next-Generation Connected and Autonomous Vehicles (CAVs)

Figure 1. Multi-source and multi-modal data in CAV

Multi-modal, multi-source data in connected and autonomous vehicles (CAVs) refers to information obtained through different modes or channels, as shown in Figure 1. Each mode provides unique and complementary insights that, when combined, offer a more holistic understanding of the environment. Real-time decision-making in autonomous driving must integrate data in multiple modes from various sources dynamically to enhance the system’s situational awareness, reliability, and robustness.

The next-generation connected and autonomous vehicles (CAVs), embedded with frequent real-time decision-making, will rely heavily on a large volume of multi-modal and multi-source data. However, in real-world settings, the quality of different modalities and data sources usually varies due to unexpected environmental factors or sensor issues. This brings major challenges (but not limited to): (1) mitigate the underlying influence of arbitrary noise, bias, and discrepancy, (2) enhance data quality (DQ) in terms of different dimensions, (3) adapt the quality dynamically varying nature of multi-modal multi-source data. However, both AI researchers and practitioners overwhelmingly concentrate on models/ algorithms while undervaluing DQ, which is the backbone for assuring the performance, efficiency, reliability, and robustness of next-generation CAVs.

Figure 2. The Vase Framework for Task-centric, Data-driven, and Quality-aware CAVs

This project will develop DQ evaluation and enhancement mechanisms for multi-modal and multi-source data that fulfill the needs of CAVs with guarantees of functionality, efficiency, and trustworthiness. We will also design effective quality-aware data transformation frameworks that adapt to the heterogeneity and dynamic environment of CAVs.

Related papers:

Zhou, Y., Chen, H., & Sha, K. (2025). A Novel Multi-layer Task-centric and Data Quality Framework for Autonomous Driving. arXiv preprint arXiv:2506.17346.