TU-TESL

• 22

Wave, a British start-up, raised $20 million (about 140 million yuan) in a round of financing led by eclipse ventures, a Palo Alto venture capital firm, with balderton capital, compound ventures and first minute capital participating, foreign media reported. Wave is developing artificial intelligence (AI) technology that uses enhanced learning, simulation and computer vision to teach cars to drive themselves. Zoubin ghahramani, Uber's chief scientist, and Pieter abbeel, Professor of robotics and pioneer of deep reinforcement learning at the University of California, Berkeley, and several famous angel investors also participated in the financing round. (photo source: venturebeat.com) Wayve was founded in Cambridge, UK in 2017. The core premise of the company is that the major breakthrough of the automatic driving vehicle lies in the better AI brain, rather than more sensors or "manual encoding" rules. The company says it will use the simulation environment to train the autopilot system, then transfer such knowledge to the real world, and simulate how people adapt to the environment in real time in the real world. The Wayve system will learn from what each safety driver intervenes to understand why drivers must intervene to bypass high-precision maps, lidar and other sensors that have become synonymous with self driving vehicles. It is worth noting that if a car manufacturer wishes to use hardware or sensors, Wayve's machine learning algorithm can also work with it. However, the core of Wayve is that self driving cars should be able to learn new environments like humans. Alex Kendall, co-founder and chief technology officer of wave, said: "our algorithm is learning to become Superman driver. We learn from careful driving behavior, which can eliminate 98.3% of human errors caused by negligence or invalid driving. Then, we feed back to the system to further improve the enhanced learning ability beyond human drivers. " Wayve insists that you can use end to end machine learning, basic camera and GPS navigation to create a safe and efficient automatic driving vehicle system. "As computing power and data continue to grow, learning based approaches are more inevitable, especially for mobile robots," said Amar Shah, co-founder and CEO of wave. The human brain has evolved for millions of years, while the computer has only evolved for decades, and now the computer is catching up. " In April, wave announced that its technology was "the world's first" when it demonstrated "using cameras and satellite navigation (satnav) to apply technology to cars in an unprecedented complex urban environment." So essentially, the car uses only machine-driven "eyes" and "brains" instead of relying on high-precision maps or sophisticated lidar sensors. Sure, wave's system seems to be a cheaper solution for autopilot technology, but why are companies like waymo, an autopilot subsidiary of alphabet, a parent company of Google, still investing a lot of money in the development of lidar sensors, while companies like TomTom are investing millions of dollars in high-quality maps? For autopilot cars and trucks, if we want to cross the busy streets at high speed, we must be able to recognize and understand the surrounding environment to avoid collisions, and the lidar sensor is to measure the distance between the vehicle and the object by emitting the laser to the object and measuring the reflected pulses. However, laser radar is not suitable for all scenes. Therefore, high precision map plays an important complementary role in the development of autopilot. It can help identify more accurately the lane, geometry and traffic signs, and also effectively enable the car to see the corners around. This is the function that lidar or computer vision can not achieve. Therefore, in essence, the basic satellite navigation and camera can only build a poor performance car, even though the system is trained, it can achieve 95% safe driving with the least hardware accessories. However, there will always be "extreme cases" and need to visit more data. Therefore, access to large amounts of data has become a selling point of autopilot, but Safety is also a big selling point for cars. (photo source: VentureBeat. Com) according to a blog post released by wave last year, the development of computer vision technology in the past few years has enabled it to "use only cameras to realize perception" and drive on the road. In 2015, based on the University of Cambridge, the company began to develop its own segnet deep learning system, and said that it has achieved advanced success, can understand semantics, geometry, depth and motion from a single deep learning model, and made a comparison. With Wayve's autopilot, the "eyes" of these cars become a 2 million 300 thousand pixel RGB camera with a high dynamic range and very good signals at night and in bright sunlight. (photo source: wave) wave, which recently acquired investment and moved its headquarters to London, has raised 3.1 million US dollars (about 2172.33 yuan) before. This time, it has obtained another 20 million US dollars investment. The company plans to launch a test site fleet composed of electric vehicles in central London, equipped with a large number of safe drivers. Kendall said: "we will send a team of eight i-packs from King's Cross Station headquarters in London to develop our technology. As soon as we can prove that such cars can run safely on London's complex roads, we will launch a commercial pilot. " It is also a fact that cannot be ignored that wave has successfully won some famous supporters from the field of venture capital and technology. And it's rare for Silicon Valley investment companies to invest in a self driving start-up in Europe.