Last week, automotive component supplier DENSO Corp. announced that it is working with Drishti Technologies Inc. to optimize production at several North American factories. DENSO said it is applying Drishti’s “action recognition” technologies to manufacturing processes to identify bottlenecks and provide feedback in real time.

Even as industrial automation spreads, people still perform 72% of production tasks, according to a survey of 100 manufacturers conducted by A.T. Kearney and Drishti. As a result, it’s important to monitor both the people and the machines in workcells, said Prasad Akella, founder and CEO of Drishti. Helping people be more effective in concert with robots produces better results than focusing on either in isolation, he told The Robot Report.

Drishti develops holistic view of production

Mountain View, Calif.-based Drishti combines computer vision, artificial intelligence, and continuous analytics to create data sets that it claimed “are several orders of magnitude larger than those produced by traditional time and motion studies.” It has been working with DENSO since its founding.

“I met Prasad at SRI International in 2017, when we realized that this concept had a huge potential,” recalled Raja Shembekar, vice president of DENSO’s North American Production Innovation Center. “Before Drishti, the concept of digitizing human motion to get insights didn’t exist.”

“I was working at SRI International, a 70-year-old R&D spinoff of Stanford University, when I had a conversation with DENSO about machines,” Akella said. “I asked, ‘Where is your data on manual processes?’ Raja came down, and he took one look at our science experiment and saw value.”

“DENSO opened up its facilities to us and worked through iterations of the new technology,” he said. “They’ve been phenomenal collaborators and helped us make this work at scale. I also credit our engineering team for finding clever ways to store data and metadata and use the two of them to bring life back to every bit of analysis. The beauty of AI is that it can narrowly focus engineers’ attention on what needs it.”

“We’re very excited that the World Economic forum chose Drishti as a technology pioneer for 2019,” Akella said. “Krishnendu Chaudhury, our CTO, is the brains behind one of the most advanced neural networks out there. Board member Ashish Gupta has been growing the business, and we’re very fortunate to have this team to build our technology and bring it to market.”

“Our vision is that we are part of the infrastructure,” he said. “This is important to integration. Robots used to be sold directly to OEMs, but we’d love to work with integrators to penetrate the market.”

Drishti AI designed to help line workers learn

“Even though DENSO is very lean, we still rely on humans in lower-cost countries, and people make mistakes,” Shembekar explained. “We want to make sure quality is top-notch, but an industrial engineer standing behind an associate with a stopwatch skews the data. It’s just a snapshot, and assembly is a dynamic process.”

“The single biggest lesson for [Drishti and DENSO] was that we needed to encapsulate this pretty sophisticated technology in an easy-to-consume manner so that front-line employees — line associates and line supervisors — could know what’s going on,” said Akella. “Let the engineer consume the data, and abstract it away from other users so the operator can understand it. How do you surface all the data, make it actionable, and hide the nuances of building the neural network?”

“The associates need feedback on consistency in real time,” he added. “At the heart of production is human variability, and the job can be monotonous. We can add a bit of gamification and provide a natural beat. Sometimes, there aren’t any natural beats a line associate can work with, not like on a conveyor belt. Feedback gives a more natural beat to follow.”

Immediate feedback boosts morale, retention

“I’m convinced that the way we’ve looked at people is very limited,” asserted Akella. “The fundamental assumption is that we’ve tapped out the capabilities of humans and that the only way to improve productivity is with machines.”

“Humans are still more dexterous and flexible than many robots, but machines can give guidance to reduce errors,” he said. “When we worked on cobots, we recognized that human cognition just needs power assistance. We expect people on the manufacturing floor for decades to come, and Drishti’s mission statement is ‘to extend human potential.'”

“Finding any deviations from standard work processes is valuable. As smartphone makers have learned, rework costs five times more than getting it right upfront,” said Akella. “Like when an NFL quarterback throws an interception and reviews the video to avoid it happening again, associates have the ability to review for quality control in the moment.”

“Drishti’s root-cause analysis brings search to the manufacturing floor,” he said. “As a result, cause and effect are clear, and there’s dynamic learning and less labor churn, which is important when there’s less than 3% unemployment.”

Longer-term goals

Drishti is looking at using its technology to monitor more manufacturing activities, said Akella.

“Equipment failures cost hundreds of thousands of dollars. We have patents filed, and we’re also looking at observing robots. It’s not just the internal data that they generate but the entire process,” he said. “We’ve already solved the harder problem — observing the variability of humans, who might be left-handed or right-handed, shorter or taller.”

Kariya, Japan-based DENSO, which has North American headquarters in Southfield, Mich., said the digitization efforts are in accordance with its Long-Term Policy 2030.

“As we look to 2025 and beyond, with robotics, IoT [the Internet of Things], cloud computing, and machine vision, the challenge is not to develop technology for the sake of technology but to solve complex manufacturing problems and improve the performance of DENSO and its associates,” said Shembekar.

Meanwhile, Drishti is working with other customers to understand their unique problems. “For example, training is a big issue,” said Akella. “With low unemployment, managers may not know who will show up for work on a given day, and they have to allocate people. Who’s more nimble or faster? We want to improve coaching to help the average person can play to their strengths and become as close to a master craftsman as possible in the shortest period of time.”

“Millennials are used to going to YouTube to solve problems,” Akella said. “Dojo stations in Japan are the digital version, with the physical version facing the operator. There’s a huge opportunity to not only transform manufacturing, but also how to train people.”