Startup Symbio Robotics Gets $30M Funding to Extend Factory Automation Using AI
The modern factory floor is increasingly automated, but manufacturers are looking for ways to make industrial robots more flexible. Symbio Robotics, a Bay Area startup, is aiming to solve those problems using its new AI-based platform that allows robots to learn, adapt and execute new assembly tasks.
The company emerged last week, announcing partnerships with automakers along with $30 million in venture funding to be used for what it calls “breathing new life into existing industrial robots.” The company’s early investors include Andreessen Horowitz.
Based in Emeryville, Calif., the company describes its SymbioDCS development platform as robot-agnostic, meaning it can integrate real-time AI software with existing automated manufacturing lines. The robotics middleware and Python programming framework are billed as easing the programming of industrial robots to add intelligence.
“Flexibility and automation have traditionally been at odds,” said Max Reynolds, the startup’s 29-year-old co-founder and CEO.
Rigid and inflexible block-box automation tends to work well for manufacturing tasks such as welding, but AI tools are so far used sparingly in complex assembly tasks. For automated manufacturing, “Final assembly is the final frontier,” and the focus of Symbio Robotics’ AI efforts, Reynolds added in an interview.
The platform leverages huge volumes of real-time sensor data in the form of streaming video, combining those sensor inputs with control software that allows re-programmed robots to learn and expand their capabilities beyond narrow assembly-line steps.
The software-defined approach uses Python libraries to build real-time automation models. Manufacturing robots installed on an auto assembly line, for example, could be retrained to execute new processes, which could then optimize those new functions through repetition.
Along with Python, the SymbioDCS platform features concurrent, real-time sensor processing and control running on Linux via Docker containers.
On the system front-end, sensors and cameras help the robot adapt to process variations. The opportunity, Reynolds added, is enhancing human-robot collaboration, then applying those improvements to specific workflows such as final assembly.
The software-defined approach also is promoted as the next step in smart manufacturing, creating what investors said are “intelligent, self-aware” robots capable of learning new assembly tasks.
Symbio also announced partnerships last week with Nissan Motor Corp. (OTCMKTS: NSANY) and Toyota Motor Corp. Toyota (NYSE: TM) is a pioneer in robotic auto manufacturing. Interestingly, electric car manufacturer Tesla, which relies heavily on robotic manufacturing, is just down the road from Symbio in Fremont, Calif.
Another early investor in Symbio is Eclipse Ventures. “Today's automakers need more nimble and nuanced robotics to perform increasingly sophisticated assemblies," said Greg Reichow, a partner at Eclipse Ventures and the former head of global manufacturing and automation engineering at Tesla (NASDAQ: TSLA).
Car makers are among the largest industrial adopters of robots. Rapid product cycles are forcing them to look beyond dedicated robots performing specific manufacturing tasks. The problem is many industrial robots run proprietary programming languages, further limiting their flexibility.
Symbio comes at the problem from a different angle: using AI to boost human-robot collaboration via a framework that supports programmers. SymbioDCS aims to help developers and domain experts in training machine learning models that allow robots to adapt to perform new tasks.
“To do this, they are building products that leverage AI strengths and human insight in a symbiotic way,” said Anca Dragan, assistant professor of electrical engineering and computer sciences at the University of California at Berkeley, and an advisor to Symbio. Dragan co-founded the university’s AI research lab.
Reynolds said the startup is currently deploying its first systems in production, then scaling its collaboration with Toyota and other customers.
Reynolds is part of a wave of mechanical engineers out of robotics hotbeds like the University of Southern California’s Viterbi School of Engineering who are seeking to marry AI technology with industrial applications. The emphasis is on closer collaboration between developers, domain experts and the robots they design. Human operators would provide feedback and design control policies for squeezing more performance out of robots. The machines could then move up the learning curve to tackle complex assembly tasks.
“It’s a really important design challenge,” Reynolds concluded. “AI can enable the best of human-machine interaction.”