We're building the world's most reliable autonomous vehicle software stack to solve the most difficult problem of all - delivering a solution that's safe in complex urban environments, without any driver involvement. That's defined as the top level in autonomy - 5.
Early approaches have required highly-accurate dense 3D prior maps and localization to solve this challenge. At FiveAI we're using stronger AI to solve the challenge, meaning we don't need to map 37.2 million kilometers of road across the planet to deliver a reliable solution.
Our solution will enable safe, cost effective urban mobility for all and we'll be announcing partnerships to deliver it to consumers in staged deployments across our cities and towns as we grow.
We use multiple sensors around the vehicle to provide a resilient and accurate view of our environment, including high resolution cameras, radar, ultrasound, LIDAR, IMS, audio and GNSS.
Using advanced computer vision combining deep neural networks with unsupervised learning technologies, our perception layers deliver a reliable scene segmentation, object classification, depth, state and position in 3D space. Our architecture will support the validation required for safe operation of a vehicle.
Predicting what happens next requires inference of the intentions of the actors in a 3D scene. Using a policy-based approach, our behavioural modelling layer assigns policy probabilities based on observation and solves the problem of how to iterate actor interactions forward in time within feasible computational limits.
A focus on high performance AI limits the need for dense prior maps to determine our motion; instead we determine tracks from point-to-point on a real-time basis, just like a human, using a much simpler prior map. Our vehicle is controlled on that path through tight control loops and multi-sensory feedback.