Waymo acquired Latent Logic, a UK start-up that uses imitation learning to simulate models of human behavior on the road. Financial details of the acquisition were not disclosed, but it marks the launch of Waymo’s first European engineering hub, which will be located in Oxford.
Spun out of Oxford University in 2017, Latent Logic develops simulations of motorist, cyclist, and pedestrian behavior using data collected from traffic cameras. This data is used to create simulation environments that might help autonomous vehicles interact more safely with human beings. Cyclists are considered by many to be the most difficult detection problem for autonomous vehicles.
An archived version of the Latent Logic website describes its technology: “There are two stages to our learning process: extraction of behaviours from raw data, and learning to imitate those behaviours. Computer Vision detects road users and tracks their motion. Imitation Learning, also known as Learning from Demonstration, then learns to create new, artificial behaviours which are indistinguishable from the ones used as the demonstration input, meaning our virtual humans are completely realistic. Our virtual humans integrate with our customers’ preferred simulator via a standard API.”
Latent Logic continues, “rather than hard-coding a specific set of behaviours, we apply machine learning to create agents that develop their own behaviours based on how humans actually behave. Because an a-priori description of human behaviour in all possible situations is almost impossible, agents with hard-coded behaviour cannot have the same flexibility and adaptability as our AI agents.”
The video below shows a short demonstration of the company’s technology:
Waymo has often touted its approach to autonomous vehicle simulation. In July 2019, it announced it had driven 10 billion miles in simulation. It also said simulation becomes more realistic as you gain more real-world driving miles, and no company has more autonomous driving experience than Waymo.
Apparently simulation software is in high demand these days. Just yesterday, The Information reported Uber is in talks to acquire Silicon Valley-based simulation startup Foresight. According to the report, Uber’s simulation software has suffered from “various deficiencies and still has trouble predicting how its self-driving car prototypes will handle the real world.”
Latent Logic tweeted that its team will remain in Oxford. And its two founders, Shimon Whiteson and João Messia, CEO Kirsty Lloyd-Jukes and other members of the engineering team will join Waymo.
“By joining Waymo, we are taking a big leap towards realizing our ambition of safe, self-driving vehicles,” said Latent Logic co-founder and chief scientist Shimon Whiteson. “In just two years, we have made significant progress in using imitation learning to simulate real human behaviors on the road. I’m excited by what we can now achieve in combining this expertise with the talent, resources and progress Waymo have already made in self-driving technology.”
Today, Latent Logic becomes part of Waymo. Their uniquely talented team, based in Oxford, UK, uses imitation learning to simulate realistic models of human behavior on the road—key to developing safe self-driving vehicles.
— Waymo (@Waymo) December 12, 2019
Waymo recently started to operate its self-driving taxi service near Phoenix, Arizona, called Waymo One, without human safety drivers. The program turned one year old on December 5, and to celebrate the company launched its ride-hailing app in Apple’s App Store. While this shows it is interested in eventually expanding its customer base, the self-driving service will certainly remain hyper-localized for some time.
In August 2019, the Waymo Open Dataset was released for autonomous vehicle researchers. Available for free, the dataset covers a wide variety of environments, from dense urban centers to suburban landscapes. It also includes data collected during day and night, at dawn and dusk, in sunshine and rain.
Largest Autonomous Vehicle Investments of 2019
|Company||Amt. ($M)||Lead Investor||Date|
|Argo AI||2600||Volkswagen AG||7/11|
|Cruise||1150||Honda Motor Corp.||5/7|
|Uber ATG||1000||Softbank Vision Fund||4/18|
|Nuro.ai||940||Softbank Vision Fund||2/11|
|Xpeng Motors||400||Xiaomi Corp.||11/15|
|Innoviz Technologies||170||China Merchants Capital||3/26|