AI’s Practical Impact on Warehouse Robotics
Unlock the potential of AI in warehouse systems— It holds the key to a future of productivity and safety.
In warehouse robotics, the interplay between AI and practical applications takes center stage. Daniel Theobald, founder and chairman of Vecna Robotics, challenges the prevailing narrative surrounding AI’s transformative role.
The high-profile rise of generative artificial intelligence (AI) models like ChatGPT over the past year has helped feed the prevailing narrative that AI will soon be a panacea in various fields, including warehouse automation. But this overhyped view, grown out of fascination with a powerful new tool, ignores AI’s current limitations. AI does offer a lot of value to warehouse operators and manufacturers, but it must be applied where it’s most capable of doing good.
It’s important to understand what AI brings to the table—and what it does not. The hype around AI, particularly on the heels of recent advancements in large language models (LLMs), often features descriptors like intelligence, learning, and sometimes even sentience. Whether people are touting AI’s revolutionary potential benefits or warning of an extinction-level event, the assumption is that AI will act with an intelligence surpassing that of humans.
But this anthropomorphizing of AI is irresponsible and just wrong. What AI does is compress data in a way that provides better predictions. Predictive analysis of cybersecurity threats doesn’t see the future; it states probabilities based on past behavior and current conditions. ChatGPT can spin out an email, legal brief or love poem in the blink of an eye, but it only predicts the next word based on all the words and usage it has access to. There is no intelligence, no knowledge, and no real learning. It’s just compression, a mathematical function.
The hyperbolic descriptions currently being thrown around about AI’s intelligence create anxiety and a false sense of authority. And it can lead to people using the technology thoughtlessly.
For warehouse robotics, AI and machine learning (AI/ML) can produce significant improvements, but anyone ready to let the technologies run warehouse operations has gotten far ahead of themselves.
See more: How AI Empowers SMBs in Navigating the Digital Era
Can AI Guarantee a Safe Workplace?
Safety is a big reason to limit AI’s use in warehouse automation. The technology is not close to the point where it should control valves and motors on heavy mobile equipment, for instance, because of the vital need for safety-rated base system controls. We must prove with five-nines certainty that robots will fail safely, and AI alone cannot deliver that.
Robots with low safety impact, like small order-picking AGVs and AMRs navigating the warehouse floor, can afford to make mistakes, like bumping into a worker. Large warehouse robots and those that support flexible manufacturing, such as counterbalanced forklifts and pick-and-place robots with articulated arms, cannot. They need a physical safety system tested by a person and extra layers of safety controls.
The possibility of leveraging AI to improve the real-time performance of warehouse robots also raises practical concerns. Even if AI systems could satisfy safety concerns well enough to be deployed into production in real-time systems, would the juice be worth the squeeze?
Putting cutting-edge AI systems safely into operations would require enormous expense and effort and would likely not produce any revolutionary breakthroughs in productivity. Even if warehouse operators thought it could be worth it, who has the time, budget, and patience to experiment with AI-driven systems while disrupting current operations to get those systems working? Large, well-capitalized logistics players with ample research and development budgets might be able to develop such systems in a vacuum, but trying to develop them live on the warehouse floor isn’t viable.
In a nutshell, when the answer doesn’t matter, AI is great. But when you must be right, there are too many undefined parameters in warehouses—such as dust, people, floor conditions, humidity, weight, and site variability, to name a few—to deploy AI-driven automation safely and with the appropriate near-term ROI.
AI/ML’s Immediate Benefit for Simulations
Those shortcomings don’t mean AI/ML has no place in warehouses. It should be at the foundation of plans for improving operations, as long as warehouse operators use it for what it does best.
The need for better robotics and automation is clear. Warehouses have a labor shortage, with most of the market being 10% to 25% understaffed, according to Vecna Robotics’ 2023 warehouse automation statistics survey. And although supply chain professionals see the benefits of warehouse automation, most have yet to take full advantage of automation. AI substantially benefits warehouse automation systems in navigation, pallet detection, and mobile fleet management. And because of safety concerns, disruptions to operations that can affect productivity, and the need to verify outcomes for specific site variations, the simulation will likely be the next big breakthrough for the technology.
In particular, AI/ML can optimize algorithms to make simulations more impactful, improving the speed, scale, and insights of simulations. Nobody will operate a warehouse without a full digital twin in the next few years because digital twins can help avoid many potential problems.
For instance, an AI-fueled system can monitor everything in the warehouse. With the data collected and analyzed, decisions related to automated systems will stop being guesses. Some problems are inevitable in a warehouse based on initial conditions and usage. A finely tuned algorithm used in a digital twin can predict, with a high degree of certainty, that a certain problem will likely crop up in, say, four hours, which no human can predict. Predictive analysis applied in that way can save warehouses and manufacturing plants from significant downtime costs. For example, production line stoppages at automotive plants cost an estimated average of $22,000 per minute or about $1.3 million per hour.
Using AI to optimize simulations presents huge opportunities. It can help operators avoid disruptions and inform decisions on using less energy, making better use of resources, taking care of workers, or more efficiently serving customers. AI in warehouse robotics can enable more consistent, efficient operations while allowing operators to use their human employees best. Because of all that, simulations represent one area where you can expect much progress over the next decade.
The Key to Improving Automation Decisions
We won’t fully capture the value that robotics can provide if we can’t effectively direct them to the most valuable work needed and how to perform it best. The key is to collect good data, fully understand it, and then use that information effectively to drive automation decisions.
The good news for warehouse operators and manufacturers is that some sectors, such as automotive manufacturing, have paved the way for robotics over the past decade. They have already borne the high costs of innovation to an extent. Industries that are beginning to make greater use of automation, such as agriculture and construction, could take a page or two from the industries that have already pioneered some solid successes.
Some projects will fail, as they always do while making progress. But warehousing and manufacturing now have a broad set of tools, including AI/ML, to help them make the most of the robotics and automation. They just need to be sure to apply them in the best way.
How do you envision the role of AI shaping the future of work in warehouses? Share your thoughts with us on Facebook, Twitter, and LinkedIn. We’d love to hear from you!
Image Source: Shutterstock
MORE IN AI WAREHOUSE
