Uncertainty-aware and safety-enhanced management of CAVs for safer mixed traffic

Purpose and goal:

Safety of connected and autonomous vehicles (CAVs) in mixed traffic environments (CAVs and human-driven vehicles (HDVs)), requires an advanced approach that takes into account the multifaceted uncertainties inherent in real-world traffic scenarios. Designing operational control mechanisms that can detect and counter safety issues is critical to ensuring the safety of CAVs in mixed traffic environments. This project aims to improve the safe operation of CAVs in real and complex traffic environments.

Expected results and effects:

Detection and quantification of uncertainty in mixed traffic scenarios. This objective aims to develop advanced perception and quantification models that can predict and quantify uncertainties that may result in safety concerns for CAVs.

Safety-enhanced and uncertainty-aware operational control of CAVs in mixed traffic environments. This objective aims to develop adaptive and robust methodologies for the operational control of CAVs in mixed traffic based on perceived/anticipated uncertainty. Approach and implementation:

Bayesian probabilistic models with Monte Carlo simulation will be employed to quantify uncertainty based on data from multiple sensors using Kalman filtering and semantic data fusion. Multi-agent reinforcement learning and model predictive control for advanced uncertainty-aware and cooperative control, so that CAVs can timely react to uncertainty in mixed traffic and cooperate effectively with each other. These algorithms should focus on ensuring safe operation control of CAVs