Realistic Simulation for Connected Autonomous Vehicles (CAVs)

The challenge

Connected Autonomous Vehicles (CAVs) rely on vehicle networks, advanced sensor technology, and cooperative perception to improve road safety, driving comfort, and energy efficiency. However, validating these systems under real-world conditions remains a major challenge. 

Real-world data collection and physical testing are costly, time-consuming, and difficult to scale. Existing simulation frameworks are often limited to single-agent setups, static scenarios, or simplified communication models, making them unsuitable for testing complex, safety-critical traffic situations involving multiple interacting vehicles. 

As a result, the performance of CAV systems in complex scenarios remains suboptimal, and potential communication failures or safety risks are often detected too late in the development process. 

How we will help

To achieve this, XITASO extends the established single-agent Autoware framework into a scalable multi-agent simulation environment and combines it with state-of-the-art AI and network simulation technologies: 

• AI-based generation of realistic traffic scenarios, trajectories, and maps 

• Generation of multimodal synthetic sensor data (LiDAR, Radar, RGB-T) with realistic characteristics 

• Integration of physical network simulators to accurately model real-world V2X communication 

• Scenario management for reproducible and complex test cases 

• An adaptive evaluator to iteratively reduce the Sim2Real gap 

• A seamless interaction between simulation and AI generation enables the evaluation of autonomous driving functions under realistic and safety-critical traffic conditions. 

Technologies & topics