Human-in-Loop Automation

Human-in-Loop Automation

By Hongbin Li

Robotics has been catching public attention with the recent fast development of drones, automatic driving, etc. They have appeared in numerous movies and TVs, such as the Terminator movie series started over 30 years ago. It is not a surprise to see robots as powerful machines with their own mind, act on their own with unparalleled power and capabilities. Robots are truly miracle of automation technologies.

Our Robots, Ourselves [1] is a book about the interaction between robots and humans. It tries to correct the common misunderstanding of the progress of automation and robotics: robots are not designed to replace human and to achieve full autonomy. Robots are developed to assist human to complete tasks that are difficult for human to finish alone. Humans are always involved with robotic actions, such as giving instructions at the scene with robots, remotely operating robots, monitoring and reviewing results provided by robots. Various examples of robots in the sea, air, and space, are presented in the book, which showed robots collaborating with human can accomplish tasks at lower cost and without danger to humans. In addition to the examples shown in the book, collaborative robots are one new type of robots that are specifically designed to work around humans and to share the same work place.

In contrast to robotics, Process Control is another automation field, less popular or interesting to the public, but essential to provide material goods for our daily life. It develops technologies to automate continuous manufacturing processes, most commonly seen in chemical manufacturing, oil and gas, water treatment, etc. In Process Control, automation is aimed to serve human operators for improving the product quality, manufacturing efficiency, and safety. Following the idea of the above book, we may ask: what is the role of human in Process Control?

Process Control and the Human Operator

In a modern manufacturing plant, a process with tens of thousands variables can be operated by a dozen human operators. Field measurements are continuously gathered by electronic controllers; status parameters are shown on Human Machine Interfaces (HMI) for monitoring; controller corrections and human commands are sent to the field for execution; any deviation from normal value ranges trigger alarms on HMI or other annunciators to alert operators; further deviation can cause action of the Safety Instrumented Systems (SIS) to shutdown equipment or process and bring it to a safe state. For these complicated systems, operability is at the center of process control. For example, the projects for installing or modifying process equipment must go through Hazard and Operability (HAZOP) study to identify any risk, hazard, or operability problems in early design stage. To identify and evaluate any hazard and operability problem in control systems, a dedicated Control System HAZOP (CHAZOP) can be carried out.

Human operators rely on HMI and alarms to monitor the status of the process under control. It is essential to account for humans’ cognitive characteristics and sensory limits in the design of HMI and alarms, often referred as Human Effect Engineering (HFE). For example, humans have limited attention bandwidth and are alert to changes, not static objects. The automation industry has established standards and best practices to take these human factors into account. For example, the International Society of Automation (ISA) has published alarm management standard (ISA 18.2) and human engineering best practice for control centers (ISA-RP60.3). Many manufacturing plants have further developed site-specific standards to be followed in the design of Process Control systems.

Although the human factors are given much consideration in the process industry, little is discussed in the college education of engineers and technical specialists. There are courses on control theories, instrumentation, process knowledge and equipment, but rarely touch on the subject of human effect engineering. Only when these young professionals join the work force, industrial training and work experience start to close this gap.

The Pitfalls of Ignoring Human Factors

Without the consideration of human factor, the designed Process Control system can be technically sound, but provide limited benefit to human operators. For example, modern control system software can generate colorful and highly animated HMI graphics and numerous alarms with relatively little effort. Operators can be overloaded or distracted by the excessive attention caused by the overly complicated graphics and flooded alarms. The increased functionality may confuse operators and cause extended troubleshooting time when process upset occurs. This problem is caused by the lack of consideration on the human attention bandwidth or sensory limit.

The Black-Box Problem

On the other end of the spectrum is the black-box type of design. It provides minimal status information and requires little interaction with human operators. This can lead to lack of trust from human operators who refuse to use the black-box in practice. Our Robots, Ourselves [1] provides an example on the automated landing (autoland) system in airplanes. Pilots just need to turn on autoland and bring airplane down with hands free. However, statistics show pilots prefer manual land and use the autoland only 2% of the time. Another solution is to provide landing assistance by integrating human pilots into the control loops. The heads-up display (HUD) fuses computer-generated display into see-through windows. HUD uses a “flight path vector” to indicate where the plane is going and leads pilots to follow the suggested landing path. Pilots actively operate the airplane with the HUD during the landing process. In 2009, the Flight Safety Foundation concluded that HUD may have prevented nearly 70% of takeoff and landing accidents. The less automated but more interactive solution turns out to be a big success.

In Process Control, advanced algorithms are often packaged into black-boxes. They are designed and commissioned by engineers and specialists with fine-tuned parameters. As process operating conditions change with time, the design parameters may become sub-optimal and produce less satisfactory performance. If the black-boxes make unexpected move that surprise operators, they are often turned off and left unused. In order for advanced technologies to be accepted by human operators, the methods need to be presented in an intuitive way and understood by operators.

Conclusion

Automation and robotics will continue to develop with the exponential advancement of information and computer technology. Correct understanding of their relationship with human will promote the acceptance by the public so that we can fully utilize them in every aspect of work and life.

Reference

[1] Our Robots, Ourselves. David A. Mindell, Viking, 2015.