Impact of ITS advances on the industry Chapter | 9 109technology guarantees continuous learning, not only from a single user but from
all the users of this technology (Shalev-Shwartz, Shammah, & Shashua, 2016).
The correct estimation of the behaviors of surrounding vehicles will provide
the autonomous vehicle controller the information needed to understand the
situation better, act preventively and provide a smoother and more comfortable
reaction in case of an incident. It can also help in reducing energy consumption
by avoiding unnecessary throttling.
In the case of intelligent multimodal transportation and smart mobility in
the context of smart cities, the ITS must guarantee informed and justified deci-
sions and gradually prove to the users that all the recommendations help them
in gaining time or reaching their destinations in a convenient manner. These
directly noticeable advantages will increase people’s trust and will accelerate
user acceptance.
Future vehicles will have multiple energy sources and sinks for propulsion.
These need to form a multi-redundant, safe, and reliable systems. Requirements
increase even more with the integration of cognitive intelligence: In order to be-
have like a human-driven vehicle, future automated vehicles will need to “look
ahead” not just for potential obstacles, but also on weather, terrain, and other
parameters. In this context, information from external of the vehicle is to be
integrated, requiring safe and secure communication. Of course, all need to be
realized in a cost- and power-efficient way. We must not need several kilowatts
of power just to run the signal processing of an automated vehicle. To handle
all these requirements, ITS puts emphasis on the development of AI-optimized
hardware (also called silicon-born AI) and on the realization of powerful, safe,
reliable, and secure hardware platforms.
In terms of the underlying technology and the manufacturers behind them,
AI has attracted interest from every corner of the technology world. This has
ranged from graphical processor unit (GPU) and CPU companies to FPGA
firms, custom ASIC markers, and more. There is a need for inference at the
edge, inference at the cloud, and training in the cloud—AI processing at every
level, served by a variety of processors. The importance of embedded hardware
and mainly microprocessors is obvious for the AI-powered vehicles’ industry.
Since the continuous training of machine learning models relies on the fast pro-
cessing of heterogeneous data and requires significant computing power, the
leading tech companies and AI research institutions invest lots of money in
researching for high-performance processors that can handle the large compu-
tation load at the edge, thus avoiding bandwidth consumption and processing
bottleneck on the cloud. Typical examples are:
• The deployment of low-energy consumption, but powerful, GPUs, which
can be embedded in autonomous vehicles. NVidia, a major graphics hard-
ware accelerator developer, is currently developing AI accelerator chips that
can be embedded on autonomy-level five vehicles that can process camera
and other sensor input data and employ pretrained models in order to take
decisions in real time.