Advanced robotics in tomorrow's factory

Addressing the production challenges of complexity, customisation and openness.

Mass production was developed well over 100 years ago. Manual labourers worked on static assembly lines where they built products by hand from various parts and subassemblies. The introduction of automation to manufacturing freed humans from some of the most tedious and repetitive assembly-line tasks, which are ultimately faster and more reliable. These new technologies significantly improved the speed of production. Today, however, the gains made with automation have begun to plateau due to growing product complexity.

A growing trend in industrial and consumer products is an emphasis on customisation, adding to the complexity of products. Because of this, many of today’s products are too complicated to be produced with established automation technologies alone, forcing manufacturers to augment traditional robotics with manual assembly by human labourers. People are valued for their ability to rapidly understand and account for changes in a process. This value is evident for electronic and automotive manufacturers as well as logistics and warehouse operations. Compared to people, who can rapidly make changes and generate their own procedures for a new process, robots are not currently able to efficiently adapt to changes.

A flexible and automated (even autonomous) production system is the Holy Grail for many manufacturers that wish to simultaneously overcome the challenges of growing product complexity and greater customisation. The ability to rapidly switch production from one product to another will be a defining feature of businesses on the path to lot sizes of one and the highly customisable products of tomorrow.

A changing robotics market

Customisation that requires high-mix production and the complexity it brings are not the only novel requirements in a new era of manufacturing. Manufacturers are looking to strip out waste from their processes to produce high-quality, customised products at a competitive price faster than ever before. This often entails saving materials and energy in production, minimising delays when product changes are required, moving to flexible production with a less experienced workforce and adapting existing production lines to minimise upfront investments and maximise future capabilities. Among these alternatives, flexible automation is considered the key element in addressing the needs of the manufacturing industry of tomorrow.

Based on projections by the International Federation of Robotics (IFR), industrial machine growth is expected to exceed 10% annually over the next three years. This growth is anticipated across many industries, including automotive, electronics and other process-oriented industries like plastics and chemical production. This projected growth is driven by inquiry and investment in new technologies.

The goal of advanced robotics

Implementing advanced robotics in tomorrow’s manufacturing is a challenge, but also an exciting journey. To understand why, it may be helpful to compare the applications and engineering paradigms of traditional industrial robotics with advanced robotics.

Traditional industrial robots operate in a highly deterministic environment, with fixed robot cells, detailed shop floor layouts and no human co-workers. Any operations completed by the robots are designed to have a negligible or zero fault tolerance. This situation led to predetermined, offline programming based on a detailed simulation environment. Without concurrent operation and planning, iterating the manufacturing process for traditional industrial robotics takes time and resources.

The goal of advanced robotics is to shed this rigidity, enabling operations in an unknown or dynamic environment with frequently changing tasks to produce highly customised products. Advanced robotics is meant to save time and optimise production not only between robots, but with human collaborators. Doing all of this requires runtime decision-making and reactive programming for unforeseen events in the process with concurrent control of the open hardware standards and interfaces. And just like their human counterparts, these advanced robots need to be able to adapt and improve. But instead of using human senses and logic, they will continuously improve from data collected by the Industrial Internet of Things (IIoT) processed with artificial intelligence (AI).

Four stages of adoption

Industry 4.0 and, in turn, advanced robotics are positioned to improve visualisation, enable early validation of designs and facilitate collaboration across a business. To meet these goals, advanced robotics processes focus on implementing a comprehensive digital twin, providing modern and adaptable solutions based on sensing technology and doing all of this while retaining a flexible and open ecosystem.

Every company that is considering or already on the path to implementing advanced robotics will want to proceed in phases to achieve the final goal of advanced robotics. Companies can reside in one of four phases — entrants, veterans, pioneers and visionaries — with most companies today in the first two phases. To reduce overall cost and complexity, these phases are often spread across several years depending on the business and how much investment they can make to progress to the next phase.

Entrants

The first stage on the journey to advanced robotics is the adoption of widely available fixed automation robotics or similar technology. Businesses within this phase are considered entrants and most of their operations are performed manually. Process plans are designed and assigned by humans and tasks handled by robots are generated based on a specific location and time. Although these machines can be completely reprogrammed for new tasks within the software environment, downloading the new program to the robot involves machine downtime and reduces the overall facility efficiency. Once a machine is running again with a new process, calibration is performed manually based on fixed locations and parts.

Simulation of robotic motion and validation of production processes is also done in this phase, but it is limited to predominately physical aspects of the machines. Mechanical interferences can be checked to ensure full motion, including the validation of the native robot programming, to ensure desired motion. But these capabilities have room for improvement, especially in a world of highly customised products.

With so many labour-intensive tasks required to get a fixed automation robot running a new task, it is easy to see why a facility that handles many different products may want to automate the entire process.

Veterans

In the veteran phase, manufacturers begin adding some higher-level features. This phase is where many using advanced robotics sit, and it requires implementing a digital twin to validate the entire robotic production process, including mechanics and control algorithms. Each of the manual tasks can be simulated within the digital twin to optimise the motion paths of the robots and gain a greater understanding of the overall system. Digitalising this workflow is key to making improvements later in the transition to advanced robotics.

Process planning is still designed and assigned by humans and robot motion is still manually generated, but the functionality based on the digital twin allows for some higher-level input and display when working with these machines. Robot motion is generated according to system signals and the logic within the PLC on top of what was available in the entrant phase (location and time), a process usually referred to as virtual commissioning.

Now it is possible to see the spacing and interconnectivity of the facility before a machine is physically installed onsite. It is also possible to validate the software running on the PLCs and HMI programs in the virtual environment. Simulating the desired operation at the automation level on down allows for manually optimising the machines so they run as efficiently as possible. It also means even less downtime for the facility as shorter lead times are needed for implementing new or replacement units.

If a large group of robots are completing the same task, the changes can be downloaded to all robots simultaneously instead of manually programming each robot to work properly with its respective task and environment.

This digitalisation also offers the ability to experience the complete facility using virtual reality (VR). In this environment, engineers can evaluate or commission the manufacturing process while immersed in its digital twin. It also allows the comprehensive validation of mechanics and automation, PLC programing and human machine interface programming in VR.

Figure 1: Flexibility vs automation.

Pioneers

Progressing into phase three, these pioneers of their industries are beginning to automate their production systems. As with the two previous phases, process planning is still done by hand, but that is where the similarities diverge. Robot motion can still be generated based on previously available information, but now there is access to sensor data for feedback optimisation. This could be vision data, force readouts or information from other specialised sensor arrays for the relevant process. The sheer volume of data requires these robots be designed and programmed for adaptive planning based on location, time, signal, PLC logic and sensor data. This results in a higher fidelity simulation and implementation of the production process in non-deterministic working conditions.

Production process simulation can now be performed with varied or flexible parts and a dynamic environment. With easy-to-use, task-based programming for feedback control, the manufacturer can create a connected download for the entire shop floor. Semiautomatic programming on the part of the machine frees more resources for other optimisation tasks within the facility.

As an example, a facility may be assembling 5–10 dishwashing machine models. Of these, there could be variable sizes and a few different form factors. It is difficult to assume every product would have the same dimensions and different units may have different hole spacings. Products like these may also be far less rigid than products assembled with traditional robotics that can contribute to suboptimal spacing and a loss of alignment. Instead of programming for every possible array of holes, general definitions are given to the robot. The holes are known to be approximately so far apart, but with the sensor feedback the robot can readjust the positioning to correctly place a fastener in each hole. This reduces the number of instructions, which in turn reduces the time needed to get these machines operational for new product lines or small changes to the original design.

Visionaries

The companies that implement advanced robotics and reach phase four are set to become the visionaries in their field. Phase four is the complete automation of robotic function; it reaches the upper right corner of the flexibility/automation diagram in Figure 1. Now comprehensive design, simulation and deployment of a flexible production system is possible. Automated process planning and implementation is done by relying on machine learning. Optimisation algorithms integrated with physics simulation are used to train and validate the advanced robotics control scheme. The motion paths are generated automatically based on real-time system data implemented in the previous stages and the system can respond immediately to changes in production.

Machines are updated with automatic program generation while remaining online and as with phase three, each machine implements online closed-loop calibration for both dynamic cells and product variations. By bundling each robot with edge computing devices, the robots can learn and improve their processes with machine learning. Including 5G communication in the robot edge device systems can also enable a completely mobile, secure and reconfigurable facility. AI drives the manufacturing process and best practices are driven by data to self-optimise and produce the best results.

In the future, automated mobile robots (AMRs) could be used to deliver parts and completed assemblies around the manufacturing floor, replacing traditional conveyor belts. The AMRs could then be included in the complete simulation of the workshop floor to optimise delivery based on robot load or obstructed paths due to humans on the floor.

Conclusion

Mass production has been evolving for over a century and will continue to evolve into the future, but advanced robotics is where today meets tomorrow. By implementing adaptive machines that can not only sense what they are working on but react to that information, factories and other facilities that rely on automation are taking a step towards complete customisation and openness. Advanced robotics takes the complexity that has grown since the inception of mass production and leverages it to develop a highly connected and efficient workflow that can meet the changing needs of a diverse landscape of high-mix production output.

Top image credit: iStock.com/kynny