DeepNNCar is a low-cost research testbed for designing, training and testing autonomous driving pipelines and assurance components
System-level Assurance Case has become a critical requirement of the regulatory acceptance. However, problems such as human relaiance, lack of automation and design procedure makes it challenging to design a robust assurance case.
Though machine learning components have shown remarkable performance for challenging tasks such as autonomous driving ([NVIDIA DAVE-II](https://www.youtube.com/watch?v=NJU9ULQUwng&ab_channel=IProgrammerTV)), they have shown to be susceptible to slight shifts in the operating contexts, popularly known as out-of-distribution (OOD) data.
Cyber-Physical Systems need a contingency plan once it encounters high risk because of operational hazards.
Proactive risk assessment is required for runtime safety assessment of autonomous systems. It needs to utilize the detection results from the anomaly detectors to estimate the risk of the system
Scene generation has recently gained interest for testing AVs in simulation.