Understanding the distinction between automation and autonomy is crucial as these two levels of system control shape how AI approaches, risk assessments, and requirements are structured. Automated systems operate on predefined instructions, performing tasks within set boundaries, while autonomous systems dynamically adapt and learn, evolving with their environments. This chapter explores these differences, along with the standards, regulations, and safety considerations that govern both automated and autonomous systems across various applications.

FormalPara Objective

The EU regulation separates automatic systems from autonomous systems. Thus, it is important to understand the difference between the two types of systems since this will influence the AI approach, risk assessment, and system requirements in several ways.

FormalPara Introduction

As a starter, keep in mind that the decisions made or actions taken by an automated system are based on predefined heuristics. On the other hand, an autonomous system “learns” and adapts to dynamic environments and evolves as the environment around it changes. When its environment changes, an autonomous system evolves and improves its behaviour over time, making it capable of handling more complex and unpredictable situations. UNECE Regulation 156 (REG. 30) states the importance of managing software updates for advanced systems. It mandates that manufacturers ensure that updates do not compromise safety or functionality, especially in autonomous systems. This regulation includes requirements to protect against unauthorized changes and to secure the update process, which is essential for maintaining the reliability of autonomous vehicles as they learn and adapt. The ISO 24089:2023 standard focuses on software update engineering for road vehicles but can be adapted to other systems and domains, offering guidelines to ensure safe and effective software updates. It outlines how organizations should plan, develop, and implement updates for vehicles and electronic control units (ECUs), emphasizing the importance of maintaining vehicle safety and cybersecurity throughout the process. The standard provides steps for verifying, validating, and approving software updates before deployment. It also covers how to manage software update campaigns, including communication with users, handling update dependencies, and ensuring the integrity of the software.

The following diagram, from SAE International (2021), shows their suggested levels of automation (Fig. 17.1):

Fig. 17.1
Chart titled "SAE J3016 Levels of Driving Automation" by SAE International. It categorizes driving automation into six levels from 0 to 5. Levels 0 to 2 are labeled as "driver support features," requiring human supervision and intervention. Levels 3 to 5 are "automated driving features," where the vehicle can handle driving tasks under certain conditions, with level 5 being fully autonomous in all conditions. Example features for levels 0 to 2 include automatic emergency braking and adaptive cruise control. For levels 3 to 5, features include traffic jam chauffeur and local driverless taxi.

Levels of autonomy. © SAE International from SAE J3016™ Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles (2021-04-30), https://saemobilus.sae.org/content/J3016_202104/

Full size image

Levels 0–2 are just regular driving, maybe with a driver support system for automatically dimming the headlight or turning on the windscreen wipers. Levels 4 and 5 imply automatic driving, while Level 3 is something between levels 2 and 4. Level 3 capabilities could include autonomous highway cruising and lane-keeping, allowing drivers to engage in other activities while the vehicle handles the task at hand. While the SAE levels cover different automated driving features, autonomous driving is beyond these levels. An autonomous car will have abilities beyond predefined automated driving features and as such be able to observe the environment and learn how to behave in new environments and situations.

Sheridan et al. (1978) original and ground-breaking work “Scale of Human–Machine Interaction” defines eight levels of automation:

  1. 1.

    Whole task done by human except for actual operation by machine

  2. 2.

    Human asks computer to suggest options and selects from the options

  3. 3.

    Computer suggests options to human

  4. 4.

    Computer suggests options and proposes one of them

  5. 5.

    Computer chooses an action and performs it if human approves

  6. 6.

    Computer chooses an action and performs it unless human disapproves

  7. 7.

    Computer chooses an action, performs it, and informs human

  8. 8.

    Computer does everything autonomously

We see that there are several similarities with SAE levels and level 8 even mentions autonomy. Last but not least, the following table, which was developed for maritime use, is instructive (Myhre et al., 2019) and the approach and results are generally applicable (Table 17.1).

Table 17.1 Levels of autonomy
Full size table
FormalPara Requirements

The EU’s AI-Act “Whereas (12)” separates autonomy and automation as follows: “Moreover, the definition should be based on key characteristics of AI systems that distinguish it from simpler traditional software systems or programming approaches and should not cover systems that are based on the rules defined solely by natural persons to automatically execute operations. A key characteristic of AI systems is their capability to infer.” See Chap. 8 for more information.

The main questions when deciding the level of automation or autonomy are:

  • What is the users’ level of knowledge and experience?

  • How should the system be designed to ensure the situational awareness of operators?

  • What are the consequences of a mishap?

  • What are the planned environments and how well are they defined (ODD)?

Systems with SA levels 4 and 5 or automation levels 7 and 8 remove all control and responsibility from the users. Systems with SA levels 0–2 or automation levels 1–3 are support systems considered as “nice to have”. In both cases, the highest level will relieve the user of the possibility to take control. See also Chap. 6 on human aspects—controllability. Some standards, e.g. UL 4600:2023, include requirements both to autonomous equipment—e.g. self-driving cars—and to automatic equipment—e.g. automatic doors.

FormalPara Agile Adaptation

Changes to the planned system’s environment and behaviour will influence the needed training data. To quote the EU AI act (2024): “High-quality data sets for training, validation, and testing require the implementation of appropriate data governance and management practices. Data sets for training, validation and testing, including the labels, should be relevant, sufficiently representative, and to the best extent possible free of errors and complete in view of the system’s intended purpose.”

The most important issue that separates automation and autonomy is the autonomy’s need and requirements for high-quality training data.

FormalPara Safety Plan Issues

When planning the project, the following things are important:

  • Identify your customers—users, company marketing department, or others. For the developers, they are all “customers”

  • Differentiation between automatic and autonomous systems

  • Levels of automation

  • Frequent communication in the project—what have I done, why have I done it, how it will influence another project member’s decision, and so on.

  • Regulatory/standards Focus on Safety and Updates; see Chap. 8

  • Controllability; see Chap. 6