How to get started with autonomous systems

An abstract image showing a manufacturing plant with an overlay of a person leaning over a control panel

Much like automation did many years ago, autonomous systems are transforming business for manufacturers. As green initiatives, infrastructure end-of-life cycles and shifting yield expectations continue to shape manufacturers’ future plans, engineering leaders explore new solutions to propel factories into the modern era. Autonomous systems elevate automated capabilities to equip manufacturers with intelligence for improving the quality and consistency of their outputs. Built, taught and deployed by your best engineers, autonomous systems enhance expert decision-making by making AI-powered optimization suggestions to create a more productive factory experience.

Autonomous systems can advise people or assist them, as well as work independently to execute designated optimization strategies. With the power to manage thousands of opportunities and inputs in the blink of an eye, AI transforms the factory floor.

In a previous blog, we shared the three reasons you might want to consider AI-powered automation. Here, we revisit those scenarios with more details and customer examples to help you understand  how to select the best use case to get started.

Complex systems with many variables

Systems with too many inputs to specify and/or multiple competing outputs to optimize are great candidates for AI-powered optimization. The larger the set of possible inputs and outputs, the more challenging it is to find optimal solutions using traditional engineering approaches, especially if those are changing constantly. Engineers and operators often turn to intuition and heuristics in such situations.

Autonomous systems can do a much better job of dealing with that complexity and optimizing for all of the variables. AI can learn from historical data or via simulations to analyze thousands of competing inputs and deliver an optimization strategy in little time. As humans, we might struggle to detect patterns from slight variations in things like temperature, air pressure or humidity, but these are no challenge for AI models. PepsiCo took variables like these into consideration when applying AI to make Cheetos puffs. By funneling these elements into the autonomous system, the AI can sort through each variable to consider every potential outcome.

Systems requiring hard-to-find expertise

As your previous generation of factory workers prepares to leave the workforce, new employees are expected to fill the void. The challenge for any manufacturer is passing factory expertise on to new workers. Many systems rely on complex calibrations that, if even slightly off, can result in extended timelines, wasted resources and out-of-spec product. That expert level of knowledge often takes years to develop, so any resource that manufacturers can apply to expedite the process is a boon to production.

Microsoft has pioneered a unique approach called machine teaching that allows the factory’s leading operators and engineers to teach the AI agents, allowing the resulting autonomous system to master complex strategies and methodologies much faster than traditional AI techniques.  The resulting AI agents can then guide new employees through that learning curve. NOV rapidly onboarded new operators by partnering them with an AI advisor. By offering workers assistance or advice, or even operating independently, expert-trained AI can improve performance outcomes far beyond inexperienced workers left on their own.

To learn more about partnering your employees with AI to support them, view our webinar, How to improve operator decision making with engineer-built AI.

Systems with time-sensitive decisions

Some optimization opportunities take place in a fraction of a second. Chemical manufacturers need precisely calibrated reactors to produce their products. Spend a split second too long at a higher temperature and, suddenly, the entire product is out-of-spec. Alternatively, even slight shifts in timing can reduce out-of-spec polymers and improve production.

Optimal control often requires considering a high number of inputs and reacting fast. While operators do the best they can to log, monitor and manage these variables, an autonomous system organizes a cacophony of inputs into a symphony of production. The AI processes inputs in real time to optimize and create efficiencies that were previously thought impossible. Bell is using autonomous systems to identify appropriate landing spots and perform precision landings with autonomous aircraft.

Any of these categories presents an excellent opportunity, and many real systems have a combination of the above – complex systems with unpredictable inputs requiring real-time decisions that only a handful of experts know how to control. To learn more about how autonomous systems can help you evolve your business, we recommend you start with our e-book: Get started with autonomous systems: A manufacturer’s use case selection guide. We have also created several infographics on common use cases in manufacturing, logistics, chemicals and climate systems: