|Wayve has a different approach to vehicle automation / Photo Credit: TierneyMJ (via Shutterstock)|
UK-based startup Wayve has amassed $20 million in a series A round of funding led by Palo Alto venture capital (VC) firm Eclipse Ventures, along with Balderton Capital, Compound Ventures, Fly Ventures, and First Minute Capital, according to technology news platform VentureBeat. Several investors like Uber’s chief scientist Zoubin Ghahramani and UC Berkeley robotics professor and pioneer of deep reinforcement learning Pieter Abbeel also participated in the round.
Wayve’s premise on making self-driving cars mainstream will come from leveraging AI brains rather than more sensor or “hand coded” rules. The firms said that it trains its autonomous driving system through simulated environments, transferring the knowledge from the simulation into the real world, much like how humans “adapt to conditions in real-time.” The company’s systems learn “from each safety driver intervention,” allowing them to understand “why the driver had to intervene.” Wayve’s machine learning algorithms can work hand-in-hand with any hardware or sensors, but the firm’s want autonomous vehicles to learn from new environments.
Wayve co-founder and CTO Alex Kendall told VentureBeat explained, “Our algorithms are learning to become super-human drivers. We learn from attentive human driving, which already eliminates the 98.3% of human road errors due to inattention / ineffective driving.” Wayve further improves a human’s capability with using reinforcement learning by providing feedback to the system.
Wayve asserted that it can develop a safe and effective driving system using GPS navigation, end-to-end machine learning, and basic cameras. Learning-based approaches will become more inevitable in mobile robotics considering that computational data and power will grow, Wayve co-founder and CEO Amar Shah explained. Kendall said, “We’re observing massive progress in computer vision, including with some of our own work, allowing us to perceive depth very accurately from cameras.”
The problem that the autonomous vehicle is facing is not perception, but in deciding how to act, Kendall emphasized. Decision-making is a complex process, especially in London. Hence, this is where the power of end-to-end learning will be leveraged. Eclipse Ventures partner Seth Winterroth noted that Wayve’s approach to autonomy is grounded on timely advancements in the fields of computer vision, reinforcement learning, and simulation.