Autonomous cars are among the most discussed and the most acknowledged technology is currently under development. However, as always happens with relatively new technology that has not reached its maturity, the existing terminology might be confusing. Currently one can face definitions, which are inconsistent both verbally and semantically including autonomous vehicles (AV), self-driving cars, autonomous cars, robot cars, driverless cars, automated vehicles, and others. Summarizing most of the available definition we will use the following (provided by SDC_Explained_2017)^[1]:

Self-driving cars are cars or trucks in which human drivers are never required to take control to safely operate the vehicle. They combine sensors and software to control, navigate and drive vehicles.

Unfortunately, currently, there are no legally operating, fully autonomous vehicles in the United States or other parts of the world. There are, however, partially autonomous vehicles—cars and trucks with varying amounts of self-automation, from conventional cars with brake and lane assistance to highly-independent, self-driving prototypes^[2].

Regardless of official announcements only a few of the companies are actually close enough to deliver a full-scale autonomous driving technology. At the time of writing this article, the most promising producers are: Waymo, GM Cruise, Argo AI, Tesla, Baidu^[3]

If autonomously driven kilometers and a number of vehicles deployed (tested) are used as a general measure, them far ahead is the Alphabet subsidiary Waymo (https://waymo.com/), which works on the technology since 2009, when the Google self-driving car project was launched. Currently, Waymo reports 32 million miles driven in autonomous mode, which is more than any other “builder” has done. In terms of technology, Waymo uses all of the available sensors – cameras, Lidars, radars, and even microphones to “hear” sirens of the emergency vehicles. The deployed autonomous cars are taxes in Phoenix (Arizona, USA). However, the “backup” driver can still be required due to safety reasons.

The technology behind includes the following main data processing steps:

• Mapping – the preprocessed map is required to build internal data structures and representation of the road infrastructure including traffic lights, sideroads, and other important objects. The map is built by the company staff in 3D;
• Real-time sensor data processing, which enables to recognize and map surrounding objects like pedestrians, other vehicles, traffic light status, and others;
• Modelling – this step enables to forecast object motion patterns, which is of very high importance for safe driving;
• Decision making – taking into consideration of the mentioned aspects the control software determines the exact way of actions;
• Execution – the made decision is being executed on the road.

The second-largest autonomous vehicles fleet consisting of more than 180 vehicles is deployed by General Motors’ Cruise division (https://www.getcruise.com/). The developing team puts a great emphasis on achievements in AI and robotics. However, a major part of the onboard hardware is made by the Cruise team as well.

Similarly, Waymo Cruise collects a lot of real-time data from Lidars, cameras, microphones, radars, and other sensors providing a rich information source to machine learning algorithms and safety mechanisms. According to the Cruise reports, the used robotics algorithms provide decision making on a millisecond scale enabling fast and proper response. For testing purposes data is being streamed to the development cloud and simulation toolset, which enables smooth access to data of the development team. The third-largest developer is the Ford Motor Company’s startup ArgoAI (https://www.argo.ai/), which runs over 100 testing vehicles in at least six cities in the US. While currently retrofitting some existing vehicle models, Argo AI's long-term goal is to develop their own cars and produce them in masses. However, before consumer deals, the company follows the B2B model for robot-taxis companies and other fleet management-related services. Like other companies Argo AI a fundamental emphasis puts on safety, which is ensured through simulations in a virtual world in multiple scenarios at once. The sensor systems, in general, are the same – lidars, cameras, radars, and microphone arrays. Among all others probably the Elon Musk’s Tesla (https://www.tesla.com/) is the most discussed on the playground. Besides its financial and venture activities, probably the most interesting are some of the aspects of the used technology.

• Tesla’s CPU (central processing unit – in a sense that this unit is responsible for the majority of data processing tasks and decision making) provides redundancy capabilities [CleanTechnica_2020]. Another important task is cross-referencing, which enables minimizing the impact of false decisions or miss-interpreted data.
• Another important feature is the lack of Lidars. The main emphasis is put on cameras (covering 360 around the vehicle), radars, and advanced sonars.
• A deep reliance on machine learning – this one of the stated distinctive features of Tesla’s technology at least as far as it is announced.
• Development is based on electric cars, not petrol inner combustion engines, which make the cars less effective and less controllable.

The latest but still being under development Tesla’s hardware version is HW4 based on NVIDIA’s systems. Despite bold promises of delivering fully autonomous cars by the end of 2020 at the moment of writing this page delivery are still on their way. However, still, Tesla’s technology is considered among the most promising. Last but not least is China’s Baidu (https://www.baidu.com/ one might think of Baidu like China’s Google), which has rolled out back in 2019 for public tests and currently is running over 300 vehicles. At the moment Baidu runs a robot-taxi service for test and advertisement purposes. Unfortunately, not many technical details are shared with the community, but some distinctive features are known, like vehicle-to-everything (V2X) technology as well as own hardware platform like Tesla has.

Besides the mentioned companies there are many more at different stages of development. However, the fundamental building blocks are the same:

• Self-awareness sensor systems like Lidars, cameras, and others, which provide data for decision-making in real-time under highly changing environmental conditions;
• High performance computing unit with redundancy and cross-check capabilities (not all of the developer ensures these capabilities yet);
• Simulation-based training before field tests, which reduces development time and increases safety;
• A great boosting effect could be smart environments like smart traffic lights, which through intensive communications with vehicles increases safety and through the output of traffic systems in general.

The main potential impacts of technology in the future is anticipated through the following main benefits ^[4]:

• Safety is the most anticipated with the potential to reduce the huge number of car crashes on a global scale. However, the main concerns are related to software security issues;
• Equity through enabling to mobilize people who currently because of different reasons cannot participate in mobile adventures. For instance, elderly people. However, this might have some negative aspects as well for instance, significantly increased traffic intensity, displaced employment structure and others;
• Environmental footprint which might shift to both – increased or decreased because of significant growth of total miles driven. On one case due to emissions, in the other due to the use of clean energy grids (for powering electric vehicles).

In the coming chapters, other types of autonomous vehicles are discussed.