Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles (AVs) require exhaustive validation prior to production release to ensure the safety of each new vehicle. Validation teams are required to drive millions of actual road-miles while recording the data from on-board cameras and sensors to determine if the systems behave as designed under real-world conditions. This process generates a massive amount of data (petabytes) which must be annotated and analyzed to understand system performance and identify areas for improvement. With increasing levels of driver assistance in each model-year, this problem is becoming untenable and a major source of project delays, cost overruns and potential warranty and legal liability.
The Ottometric platform automatically reviews thousands of hours of raw endurance test and training data, distilling it down by an order of magnitude. Employing proprietary AI and mathematical models, Ottometric’s technology analyzes sensor and system data to verify correct operation of ADAS systems. It also generates a prioritized list of scenarios that warrant manual review, thereby guiding engineers to the “needles in the haystack”. This speeds up required certifications and greatly improves identification of long-tail issues, which can otherwise be missed.
Our team includes industry veterans from companies like GM, FCA, Autoliv, NVIDIA and Optimus Ride who have been instrumental in the evolution of ADAS over the past decade, and have a deep understanding of the systems and the training/validation process. Our unique combination of expertise in automotive electronics, AI, computer vision and big data analytics has enabled us to develop a breakthrough technology that will accelerate the availability of increasingly advanced driver assistance systems.
We are backed by investments from Trucks Venture Capital and Berkeley SKYDECK, and have a world-class group of advisors from the automotive industry, investment community and leading academic institutions.
Contact us at firstname.lastname@example.org