Sensor module volume forecast for AD (MUnits) 2014-2030 - Source: Yole Développement
It became clear to the Yole Développement team at the Consumer Electronics Show (CES) in Las Vegas, Nevada, at the start of 2016 that the Autonomous Vehicle industry is a battlefield. Thanks partly to our recently published report, “Sensors and Data Management for Autonomous Vehicles”, we can tell you where the front lines are.
In our report, we highlighted that the market for sensor modules for autonomous cars is expected to grow from $3B in 2015 to more than $35B in 2030. Car manufacturers and tech companies are therefore watching this business closely.
Source: Yole Développement
Peak hype was reached last week when car makers took over CES. People thought they were attending some kind of rebranded Automotive Show. This is not just a one-time thing – car makers’ presence grows every year. In our last report we called the automotive industry “the next consumer electronic market”.
Automotive is one of the most complicated markets. It combines low margins, long life cycles, complex industrial chains, risk, and high levels of regulation. It is torn by three revolutions at once: Electrical Vehicles (EVs); Autonomous Driving (AD); and the car-sharing movement. These revolutions are felt in three large industries: car manufacturing; Silicon Valley’s Internet business; and semiconductors.
Key car-sharing players are Uber, Lyft, and Didi Kuaidi, valued at over $50B, $5.5B and $16.5B respectively. Their movement changes the automotive industry landscape, fulfilling a new type of “on-demand” role. General Motors has invested $500M in Lyft. Didi Kuaidi, the Chinese service, has announced that it has provided a billion rides in China alone. The sharing economy is transforming the way we use transportation – and traditional car makers have clearly understood the threat.
Tesla is prominent in EVs, disrupting both the energy and automotive industries, and now targeting AD in its future vehicles. It has recently presented some features, including autopilot and cars that come when called, that should set the standard for the whole industry. Autonomous driving is also important for Silicon Valley companies like Google and possibly Apple. The promise of reducing the number of crashes, pollution and traffic, and allowing in-car entertainment and/or productivity is definitely exciting.
There are two different sensor-related strategies in AD. Almost all automotive sensor technologies are racing toward maturity, except Lidar, which is attracting large investments to accelerate its development. Yole’s research lays bare the technical choices made by traditional companies and newcomers related to the two main strategies among ultrasonic sensors, short-range or long-range radar, Lidar, dead reckoning sensors and data management.
The real bottleneck is currently in the upper layer of the system, namely embedded data processing and management. These functions are still under construction, with emerging players like MobilEye, and Nvidia providing advanced electronic control units (ECUs) that are challenging established players like Toshiba or Infineon.
Everywhere/Partial AD vs Somewhere/Full AD
- Everywhere/Partial AD is an incremental sensing approach embraced by traditional manufacturers. It relies on high sensitivity sensors combined with low-resolution maps, and then improves map quality using the sensors embedded on each vehicle. This solution allows partial automation in every situation as a ‘first step’.
- Somewhere/Full AD is exploited by Google and Baidu, based on 3D localized maps with high levels of detail, down to centimeter accuracy. The maps are combined with fewer sensors than Everywhere/Partial AD, often using Lidar. This solution allows full automation in specific environments and will work very rapidly, however it requires hard work to keep the maps updated.
In the Everywhere/Partial approach, the car’s numerous sensors attempt to provide complete external awareness. This is extremely complex and expensive, and so the automation level will only rise gradually.
In the Somewhere/Full approach, map-building sensors will be very precise and expensive but used in limited numbers. It is relatively easy to build a full automated vehicle by combining them with detailed original maps, although this constrains where AD can happen.
Both approaches will increase the number of sensors embedded in each vehicle sold. Yole projects that overall market volumes will expand ninefold, from 120M units in 2015 to over 1B units in 2030. This provides massive opportunities for sensor manufacturers and the semiconductor industry.
Some AD features will attract huge customer interest, and the market will evolve accordingly. Car manufacturers must now get closer to their customers to identify which AD functions will wow them.
Traditional car manufacturers must quit their comfort zone and many of them, like Toyota, Volkswagen, GM, Ford and Nissan, will have to compete with emerging actors like Tesla, Google, and probably Apple. Cars will no longer be traditional mechanical, fuel-powered, and high-powered electronic vehicles. They are evolving into electrical-powered, ultra-sensitive machines that can anticipate many critical situations thanks to their sensors, high-level processing units, deep learning software and high connectivity levels. The obstacles are not so much technical, but regulatory. Even though US companies are pushing their government, a global agreement is a very long time in the future.
AD will be a truly disruptive technology. Looking back at previous disruptions, radios did not replace theaters and TV did not kill radio. We cannot fully understand how AD will disrupt the automotive industry yet, but one thing is certain: This revolution will not kill the car we know. Instead, it will open new applications and increase mobility. Current car enthusiasts and dreamers of autonomous flying cars alike will have their share of the future.
Source : 
留言列表
生技醫療產業 (1)