See the original story in Japanese.
Tokyo-based Ascent Robotics has been developing autonomous driving software for Level 4 capable-vehicle (fully autonomous driving including steering, accelerating and braking without human input) in the Japanese road environment. The firm launched a beta version of AI (Artificial Intelligence) learning environment named Atlas last week.
Atlas is a learning environment for AI used in simulator-based vehicle / robot, integrating VR (Virtual Reality) human interface and 3D simulation environment, in addition to deep reinforcement learning algorithm. By using both real and pseudo data for learning AI, the learning efficiency is increased to more than 50 times compared to that when using only real data.
Utilizing the superiority of Atlas, Ascent Robotics aims the realization of Level 4 autonomy by 2020 when market competition is expected to be intensified. Since the Japanese or Asian road environment is considerably different from Western ones that DeepMind, Waymo, and Uber deal with, Ascent Robotics focuses on the Level 4 autonomy applicable for narrow streets having much traffic as the conventional Japanese streets normally have, and the team expects to lead the global market in this field.
Ascent Robotics was founded in September 2016 by Fred Almeida, the Canadian engineer who had successively held high-level positions in Salesforce or Pasona Tquila. According to Nikkei Robotics, the firm has already concluded a business contract with one of the major Japanese automakers. Ken Kutaragi, known as the father of PlayStation, has also joined the firm as Outside Director.
Looking at the trend of the major Japanese automakers, Toyota, Mazda, Honda and others have targeted about 2025 as the goal for establishment of level 4 autonomous technology. If the practical application of Ascent Robotics’ technology is realized, schedules of the major automakers’ research and development may be pushed forward a few years and more.
Translated by Taijiro Takeda
Edited by “Tex” Pomeroy