Main theoretical material. Approx. 130 pages. The final version will be an open e-book with an ISSN.
Introduction
Computing has changed how people work, communicate, and use technology. Centralised computers first expanded scientific calculation and business administration. Personal computers, mobile devices, and networked services then brought computing into everyday life. Today, computers increasingly connect to sensors and machines that can move and act. Autonomy takes this development a step further: a vehicle can observe its surroundings, choose an action, and carry it out without a person directing every movement.
The foundations of autonomy reach back to research on feedback and automatic control in the mid-twentieth century. Shakey, developed at SRI between 1966 and 1972, brought together sensing, planning, and movement in a mobile robot. Later advances in sensors and computing enabled increasingly demanding tasks. The DARPA vehicle challenges showed how quickly progress could occur: no vehicle completed the desert course in 2004, while five finished in 2005. The 2007 urban challenge added interactions with other traffic. Machine learning has since opened new ways to recognise surroundings and make decisions.
Autonomous vehicles can perform repetitive work, reach remote places, and reduce human exposure to danger. This handbook considers vehicles on land, in the air, at sea, and in space. They share many underlying ideas, but their operating conditions and safety needs differ. Marine applications are considered only at a general level.
Making a vehicle perform a task is only part of the challenge. It must also behave acceptably when conditions change, information is incomplete, or something fails. A system that works in a demonstration may struggle with poor visibility, an unexpected obstacle, or a delayed response. Strong results in selected tests do not always carry over to more demanding conditions. This makes careful development and testing increasingly important as autonomous capabilities grow.
The handbook explains how autonomous vehicles work and how their behaviour can be checked. Its technical subjects include sensors and computing hardware, software, perception, mapping and localisation, planning, control, communication, and human interaction. The order follows the connections within a vehicle: its purpose and operating conditions shape the design; sensors provide information; software interprets that information; and planning and control turn it into action. People, safety responsibilities, and protection against cyberattacks remain part of the complete system.
Each main technical chapter combines an explanation of the technology with verification and validation, often shortened to V&V. Verification checks whether the system meets its stated requirements. Validation checks whether those requirements and the resulting behaviour are suitable for the intended use. This structure supports learning the technical foundations and their assessment together; readers already familiar with a topic can focus on its V&V.
A common development framework connects the subjects throughout the book. The V-model links the definition and design of a system to the checks needed to assess the result. The handbook applies it to autonomous vehicles by connecting mission objectives, operating limits, and human responsibilities to design decisions and test results. Assessment is planned from the beginning, and its findings can lead to changes in the design or requirements. Testing is therefore part of development rather than a final activity after construction.
Several methods support this approach. System models help describe how requirements, components, and behaviour fit together. A shared modelling language makes these relationships easier to record and communicate. Scenario-based testing examines situations such as an obstacle appearing or a sensor failing. Simulation allows many conditions to be explored, while physical tests check behaviour in the actual vehicle. Recording which requirements each result supports makes remaining gaps visible and helps identify what must be checked again after a change.
Artificial intelligence and end-to-end learning are changing how some vehicle functions are built. Instead of specifying every processing step separately, developers can train a model to connect sensor information to a planned movement. This makes the training data and the conditions used for testing especially important. Recent geopolitical conflicts also accelerate the development of ground, marine, and airborne drones, with growing emphasis on onboard autonomy and rapid adaptation. Faster progress increases the need to understand operating limits and retain clear human responsibility. This handbook connects an understanding of autonomous technology with practical methods for assessing its suitability for use.
