Purchase the AI Report
VSI is proud to offer a new body of work based on its research in automated vehicle technologies.
Learning to Drive: AI in Automotive is a new report from VSI that includes everything you need to know if you are designing, developing, or deploying automated vehicle systems. The purpose of this report is to explain AI’s applications in automotive, how AI works, the challenges with validation and testing of AI-trained algorithms, and a detailed examination of the AI ecosystem within the context of automotive.
Automated driving is the perfect use case for artificial intelligence because algorithms created through AI can handle situations that the vehicle has not been explicitly trained on. In other words, AI is the best solution for handling the limitless situations that an autonomous car will encounter.
AI-based systems are a logical way to design behavior models for automated driving because these models have the capacity to learn driving policy. In other words, cars programmed with recurrent neural networks can learn to negotiate with drivers and adapt to different driving styles.
AI is promising for highly automated driving (Levels 3-5). The reason is due to the way AI can handle unstructured environments where only cognitive reasoning can handle the limitless edge cases. Furthermore, the AI-based system is always learning – the more you drive the better the performance!
Despite the favorable outlook, AI-based systems must solve the challenges associated with the validation of algorithms created with deep learning. Since inference models created with AI are considered a “black box,” where traceability is not possible, the ability to meet Functional Safety standards is compromised. However, there are emerging approaches using Safety Monitors that can encase these AI inference models with deterministic code that can override the output if necessary.