The Wise, Above-Average Use Of Autonomy In Auto Manufacturing

“Embrace your average,” quips comedian Dustin Nickerson in a viral TikTok video. “I’m very moderate, I’m very in the middle, I’m very average … I consider myself to be the Toyota Corolla of people.”

However, there’s nothing average about how Toyota approaches manufacturing. And and with the assistance of Symbio, a software-enabling manufacturing provider, they have taken originally product-focused, improvement concepts and applied it to manufacturing: autonomy and artificial intelligence (AI). Yes, Industry 4.0 was imagined years ago with the concept of self-learning based upon applying performance data for continuous improvement, but the newest implementation goes further towards the goal of 42% of manufacturing executives: “to improve operations efficiency via connected manufacturing.”

“Historically, if you think about industrial robotics, the ‘killer app’ has always been welding robots,” states Max Reynolds, CEO of Symbio.“We have all seen the picture of a sea of robots through the welding line, and in core segments like that it’s almost 100% automated. But when you walk downstream like the final assembly, it’s only 5% automated, which is true of other verticals like aerospace, white goods, heavy industrials, etc as well.”

Reynolds goes on to explain that real-time control, real-time machine learning and autonomy instead of automation will create exponential value and at a grander scale than traditional implementations. “We are striving for more flexible production,” explains Reynolds. “Typically, the number one thing you’re optimizing against is cycle time of the individual task of the robot. We optimize around a more holistic value proposition around Overall Equipment Effectiveness or OEE, which includes cycle time, availability (the amount of uptime for systems, particularly during changeovers) and quality. But when we talk about AI more broadly, we’re typically talking about replicating the senses. And so we also provide capabilities like force feedback, which is the equivalent of touch for a robot. And computer vision with a 3D spatial awareness of the vehicle as it moves down the production line so the robot can perform assembly tasks in that moving reference frame as opposed to requiring a stop station or mechanical fixturing.”

Just like the autonomy being development for embedded (a.k.a., in-vehicle) technology, a valuable aspect is a flexible, extensible interface for sensors with a central controller stitching those multiple inputs. In other words, robots from various suppliers can send and receive information via a quasi-port to the core software, which then oversees the entire system.

This generates a lot more information, a high-speed feedback loop, traceability of actions to efficiency, and a lot of return on investment. “For one plant where we’ve gotten feedback from the customer, we’ve seen an overall increase of efficiency of 15%,” states Reynolds.

“Symbio’s AI-based software gives Toyota’s team real-time control of our industrial robots and provides even more flexibility to help meet changing customer and market demand,” said Pascal Renouil, General Manager of Advanced Technology at Toyota North America.

Author’s Note

It’s rare for the right hand and the left hand of an organization to work on similar research and development and do so in coordinated fashion. How shall we quantify quality goals for machine learning or autonomy? How shall the ‘better product’ be defined and where is that definition stored, tested, etc.? How does cybersecurity get assured?

Many of these questions get asked within product development and validated via assessments and Quality Assurance departments. Many manufacturing facilities have considered portions of traditionally product-focused standards for engineering rigor (e.g., Automotive SPICE) and have been applying them as the once-clear-boundary between advanced product development and manufacturing have been blurred by ongoing, product enhancements like real-time calibrations and Over-The-Air (OTA) updates.

But the very rare is to ask these engineering rigor questions across the entire end-to-end way of working simultaneously to find the better solution.

That’s nowhere near average.