While it is tempting to use formal methods ‘for everything’, it is more effective to use the right mix of methods at the required level of verification and validation for developing and deploying L3+ autonomous systems.
The Applied team discusses approaches taken by traditional and ML-driven companies to manage increasingly more complex requirements throughout the AV development lifecycle.
With large volumes of data generated in AV development from drives and simulations, making effective use of data can be challenging. The Applied team discusses how to effectively use data to track progress on known requirements and uncover gaps in coverage.
Annotating and curating datasets from the real-world driving data is a common approach for training AV algorithms, yet it is expensive, biased, prone to errors, and limited in scale. Learn how to use synthetic data to overcome these challenges.
No company is too early to implement simulation and automated CI testing regardless of their stage of autonomous vehicle development. The Applied team discusses best practices for setting up tests to speed up the deployment of autonomous algorithms.
Sensor simulation for autonomous driving systems is an exceedingly difficult task, requiring high-fidelity simulations that could be processed in real-time. There are techniques that could be used to accurately test your perception system.
With the limitations of requirements and scenario coverage approaches, domain coverage measured on ODDs may be a practical approach to evaluate progress in the autonomy roadmap.
We discuss how simulation fits into frameworks for building safety critical systems, recent standardization efforts around this technology, and the requirements for simulation tools to support this use case.
Toyota Research Institute - Advanced Development uses Applied’s synthetic data creation capabilities to develop realistic environments for the Lexus LF30 virtual reality experience at the Tokyo Motor Show.