Autonomy-level Classification for Robots in an IIoT World

The early industrial robots of the 1970s brought flexibility and precision of movement powered by advances in servos, motors and hydraulics coupled with early computer control. Robots were typically programmed and tasked with a single role in a production line. The aim was repetition and consistency of movement and output quality. An early automotive spot-welding arm may have appeared to operate with a degree of autonomy, but it was generally following the same path each time with little awareness of anything around it; hence, it might have operated in a safety cage because it lacked sensory awareness of the human workforce. Like many industrial plant machines, these robots are becoming increasingly instrumented and significant industrial IoT components. Standard robot arms are becoming more flexible and dynamic, and the robotics field is introducing new form factors and abilities as well.

The transition from spot-welding robots to fully autonomous rolling or walking devices that are still labeled 'robots' is littered with science fiction references that confuse artificial intelligence with the mechanics of movement. The term 'bot' has been appropriated by the software industry to refer to automated responses on helpdesks and with chatbots or in mass dispersal of questionable facts and opinions on social media. We cover robotic process automation (RPA), and you can find more information in our Automation and integration: Essential tech for digital business, 2019 Research Agenda.

In the evolving world of autonomous vehicles, a standard has emerged to speciate development in the industry. This classification seeks to help clarify the levels of autonomy as the industry progresses. The aim is for fully autonomous vehicles, known as level 5, to be able to transport people and goods with no driver intervention at all. Autonomous vehicles can be considered a specific use case within robotics that has been well described, but we currently lack any classification for the many other levels of robotic interaction. Figure 1 below shows the levels for vehicle automation; a fuller evaluation of this description can be found in our Technology & Business Insight report, The Changing Automotive Industry.

Source: Society of Automotive Engineers, January 2014

It is possible to similarly classify robotic applications along the following levels.

Source: 451 Research

Level 0 robots are the original industrial robots, along with other computer numerically controlled (CNC) tooling. They are placed at 0 as an anchor point to indicate the path toward autonomous robotics. These devices can be retrofitted as IoT-enabled devices, and the physical attributes of the devices can be made to work in the higher levels described, as is the case with all legacy industrial equipment across the shop floor. Brownfield retrofits and upgrades are the order of the day in the digital transformation of long-standing operation plants.

Level 1 brings in remote controlled devices, including bomb-disposal robots that provide a degree of telepresence, shielding operators from danger, and piloted drones. They are controlled via movement-translation devices such as a joystick. A key feature is the introduction of more advanced connectivity in both control and monitoring over that which could be expected in level 0. These can be considered the start of IoT device interactions.

Level 2 introduces smart tooling, complex devices that assist humans directly and where the level of human interaction involves gesture, voice and body movement. In particular, this level refers to exoskeleton systems (see below) but also higher-end telepresence including VR-based control systems using more human gestures to interact.

Level 3 brings a degree of autonomy onboard an industrial robot, with more processing power than level 0 and a degree of sense in responding to its environment. These devices may adjust as they engage with a task – i.e., realigning to pick up an object that is in a different orientation rather than assuming the same positioning each time.

Level 4 introduces mesh robotics, which are coordinated devices that use high-function code and machine learning to engage with one another to perform tasks with high variability in areas such as picking and assembly.

Level 5 – Cobots are an emerging segment of autonomous robots that are designed to work with humans but in a way that assists and understands the role that they are performing. We described many of the features of cobots in a previous report.

Level 6 robots are equivalent to SAE level 5 autonomous vehicles. The degree of autonomy in a self-driving vehicle that requires no human input makes these AVs high up on the autonomous robot scale. They deal with a known task (transport) in a complex but known domain (the road system). This also extends to the next wave of drones using flight for delivery and inspection tasks.

Level 7 robots move into areas previously reserved for science fiction, but they are starting to appear in demonstrations and videos. These fully autonomous, self-perambulating devices use a variety of ways to traverse the world and engage with tasks. They are still primarily high-end research devices but are starting to make their way out into the world. At this level, they are focused on movement to achieve a limited task.

Level 8 robots extend the movement and navigation abilities of the level 7 robots, but they are starting to have the ability to perform many different tasks and adjust to the situation or role they are given.

Level 9 represents robots that are autonomous, highly mobile and versatile that work in cooperation with others, combining elements of Level 4 robot-to-robot communication and Level 5 human cooperation as a cobot.

What most people consider 'robots' fall into levels 7-9. A fully functioning, AI-powered 'being' with a generic set of skills. Usually we assume these will have a humanoid shape matching our various social and religious creator myths. Sometimes they are the focus of dystopian science fiction as they turn against their makers. Research and development will continue to create these devices primarily as a front end for AI abilities, and the classification will eventually have to be extended to incorporate levels of intelligence, not just autonomy of movement.

Just as the level 0 original industrial robots can be integrated and adapted to be part of an IIoT implementation, other robots may advance up the levels, dependent on the needs of the task to perform. Our categorization will help in establishing where a particular robotic device currently is but may also help to explain how it may increase the level of autonomy as the field develops. An example of this can be found in the branch of robots related to exoskeletons.

Perhaps the most well-known exoskeleton use cases are in the industrial and medical verticals. Companies such as Ekso Bionics and suitX are developing wearable technology to help people with medical disabilities get back on their feet and to ease the burden of industrial workers by giving them a helping hand with heaving lifting. Outside of industrial and medical applications, military divisions are also starting to develop use cases for exoskeleton frames. Defense Advanced Research Projects Agency (DARPA) has been experimenting with a Warrior Web program since 2016 that is used to help soldiers carry heavy equipment in the field without exerting too much energy.

Because they assist human work rather than replacing it entirely, exoskeletons currently fall under the definition of Level 2 autonomy. However, over time, we expect them to reach Level 5 autonomy, at which point the exoskeleton will have a greater awareness of the task being performed to the extent that it may move the operator to a suitable position to perform it rather that only reacting to an impulse or movement. Similar to the deployment of collaborative robots, the next stage of development for exoskeletons is to enable sensor technology, computing power, machine learning and analytics to further enhance mobility and strength.

We have created a classification for the autonomous nature of robotics in the industrial sector. Clearly, there will be challenges with many of the domains in which robots will operate. There is less need for a surgical robot to move around a hospital, but its reactions and awareness will work at a much higher precision than many other robot types. Our broad definitions can be combined and drilled into as with the example of exoskeletons and the potential evolutionary path they may take. The actual intelligence level of a robot will be more related to classifications that will emerge from the AI industry over time.