Breaking the automation risk and reluctance cycle

Even with the traditional advantage of low labour costs, Chinese manufacturers have invested heavily into automation and robotics. By comparison, the UK – with relatively high labour costs – has lagged behind. A 2019 study by the International Federation of Robotics found the UK was ranked outside the top 20 in the world for the number of robots per 10,000 workers. Professor Phil Webb, Royal Academy of Engineering and Airbus Chair in Industrial Robotics and Assembly, Centre for Robotics and Assembly, Cranfield University explains more.

With the ongoing labour and skills shortages in the manufacturing sector and industry, there’s no logic to the lack of investment into A&R – no logic, but instead an ongoing and difficult situation where businesses, especially SMEs, have been unwilling to risk the major investments needed to make the leap into A&R-based operations. And that’s a critical barrier for UK manufacturing, given that SMEs make up 99% of the sector, and 58% of its jobs.

Finding practical ways for more businesses to adopt A&R will create a virtuous circle, not just in terms of competitiveness and encouraging market innovation, but more widely for levels of productivity and economic growth for the UK as a whole. In 2023, the Department of Business and Trade estimated that upskilling the workforce around the use of A&R alone would lead to a 3.2% growth in the UK’s GDP.

The recent Catapult report (2050 vision for automation and robotics in UK manufacturing), spelt out the nature and extent of the opportunity – how we already have a foundation of academic expertise and research in A&R, especially in the automotive, aerospace and defence sectors; a wealth of start-up enterprises with innovative tech; and how increased onshoring has meant more scope and potential need for extending A&R use.

The problems are confirmed in the report as being the low levels of technology adoption, inadequate capital investment and significant skills shortages; a focus on short-term goals among businesses, and no government-led strategy to stimulate the investment needed for transformation.

Research led by Cranfield University in collaboration with Loughborough and Birmingham Universities, alongside the University of Sheffield has led to a practical means of breaking the cycle of risk and reluctance. Funded by the Engineering and Physical Sciences Research Council, the ‘Reconfigurable robotics for responsive manufacture’ project has been delivered by the university consortium with input from end-users such as Airbus, Cosworth Racing, Loop Technology and BAE Systems.

The major challenge involved with implementing A&R is the set-up phase, given the complexity and costs of establishing IT systems and re-programming any necessary changes as and when they’re needed. The units of hardware in themselves can be made available, such as robots and controllers, but the code that gives them their functionality has to be pieced together and tested manually. Consequently, automation cells have usually only been created to take on a fixed, large volume, long-term production role, where return on investment can be readily predicted.

The new software platform being proposed, R3M, provides a ‘no-code’ alternative. In addition to providing a low cost means of initially setting up an A&R system, there is the basis for using A&R as a more flexible and responsive resource that can adapt to changing needs.

A manufacturing operation should be in a position to adapt to meet evolving demand and variable types of product rather than be continually restricted by its fixed capacity. With the reconfigurable robotics software environment, a manufacturer needs only to outline the tasks required from A&R, along with the hardware available, using CAD and process data — the robots, cameras and grippers etc — and the software encodes the tasks, configuring a cell of activity and a basis of integration and communications.

Some aspects of the software are modular, drawing on blocks of standard operations, but the overall system is tailored to meet the specific needs and conditions of each business. The system ensures the configuration created is safe and legally compliant, and has the built-in adaptability to allow for the intervention of human operators when necessary.

There is a clear need to establish a flexible or reconfigurable manufacturing systems to occupy the middle-ground between bespoke production and mass production. Traditional manufacturing systems such as dedicated manufacturing systems (DMS), job shops, flow lines, cellular manufacturing systems (CMS) and flexible manufacturing systems (FMS) all have their own limitations and difficulties.

DMS produce core products at a high production rate with low flexibility — but product features are expected to be constant during the system lifetime, customisation is costly and difficult to implement, and modifications rarely occur. FMS consist of automated workstations managed through a central control unit; the throughput of these systems is lower than for DMS and the dedicate equipment increases the part full cost. CMS overcome some limitations of the previous systems through the use of multiple independent working cells dedicated to product families with similar processing requirements, but are essentially still only designed to produce a specific set of products with stable demand level and a sufficiently long lifecycle.

The core characteristics needed for a reconfigurable manufacturing system are – modularity (the ability to design hardware and software components with modular attributes; able to identify as many common modules as possible to be shared among the product families); integrability (including and integrating new components, modules, resources and technologies within the existing system); convertibility (adjust and evolve functionality over time in response to the dynamic changes); diagnosability (self-checking its current state and diagnose rapidly); customisation (the variety of the required system modules/components according to the flexibility of processing a product family); and, scalability (able to adapt production capacity to cope with changes in the system/process).

R3M provides this kind of adaptive system, designed to automatically respond to the dynamic landscape of product variations and demand fluctuations. It aims to work by meticulously capturing vital data relating to assembly products alongside the capabilities of equipment, ensuring a harmonious data flow and integration across the different aspects of engineering involved; a precise alignment of product, process, and equipment domains.

A pivotal role is played by the use of skill-based concepts such as ‘skill requirements’ and ‘skill recipe’, establishing a coherent relationship between domains, facilitating automated programming and enhancing efficiency. The system intricately maps out product descriptions, including the parts and skills involved, and set out the conditions and connections needed in terms of requirements, PPR (product, processes, resources), and precedence: turning product needs into a string of executable tasks.

This kind of automatic programme generation involves a process of three phases. ‘Recipe generation’, where product requirements are mapped to the corresponding equipment skills — offering a range of recipes, each with its own approach to optimising resources — and fine-tuned by expert insight to ensure the best, most accurate option is used.

The second phase is ‘reinforcement learning’, using advanced software and hardware to ‘train’ an agent capable of autonomous task execution and ensure there is the reinforcement of desirable behaviours in terms of the most efficient ways to complete tasks based on parameters taken from the process domain to guarantee focused training.

The final stage, Code Generation, translates the agent’s trained policy into the actual executable code, tailor-made for specific product and equipment setups. The code is provided in a form that can be continually updated and ‘retrained’ to reflect any evolution in technologies, learning and new requirements.

Perception capabilities, backed up by computational models for sensor fusion, are used as the critical foundation for robots to adjust to varied environmental conditions and task requirements. The key attributes of this system include high-resolution sensing for precise component recognition, real-time performance for minimising delays where it comes to reconfigurations, adaptive algorithms for learning new task specifics, and built-in robustness in terms of dealing with changes in the working environment.

The resulting system provides automatic interoperability. The demonstration software is to be tested in a live assembly facility at the University of Sheffield’s Advanced Manufacturing Research Centre’s flagship Factory 2050 facility from February 2025, and will then be available for free use by systems developers, manufacturing and industry businesses.

Introducing a reconfigurable robotics tool will be a powerful means of building resilience into UK manufacturing as a whole, particularly among the mass of smaller businesses: opening up access to the rapid uptake of A&R, increased competitiveness and lower costs, as well as reduced reliance on specialist skills in the workforce.