If you see self-driving robots getting lost, or robot-minders feverishly trying to control them from smartphones, you can put it down to a buggy navigation system based on open source software.
Many robot makers skip building the autonomous navigation system because of its complexity. They rely on “black boxes" available from companies like Bluebotics in Switzerland or cheaper alternatives from China, instead of taking months to write their own code. The black boxes can be retrofitted into various machines to turn them into autonomous mobile robots. But the scope to create differentiating features or cater to new use cases becomes limited when black boxes drive the navigation.
This is where Puru Rastogi, a robotics engineer from Carnegie Mellon University, came up with a third alternative. His startup Mowito, founded last year in Bengaluru, builds the “intelligence layer" for mobile robots that takes inputs from LiDAR sensors or depth-sensing cameras to enable autonomous navigation. Robot makers get a plug-and-play system that’s easier to configure than a black box.
Inspired by movies
Ever since he saw the movie Flubber in the late nineties, featuring an eccentric professor and his sentient robot assistant Weebo, Rastogi knew what he wanted to do when he grew up. He was then in school in the NTPC township of Vidyut Nagar near Ghaziabad. “The theme of my childhood was to get robots into the real world. So all my career choices were easy to make."
He got into IIT Guwahati but his grades weren’t the best because he spent too much time on projects and too little on coursework. Still, for his MS, he applied only to top-end US universities with solid robotics programmes. As luck would have it, the only one that accepted him was Carnegie Mellon University (CMU), generally considered the best in robotics.
There, he worked with Professor Howie Choset, famous for his snake robot that can wrap itself around your neck. The Pittsburgh region around CMU is one of the top clusters for robotics startups. Rastogi joined CleanRobotics to co-invent a trashbot, which took him to the HAX accelerator in another leading robotics cluster, China’s Shenzhen. This was his introduction to real-world entrepreneurship.
Back at CMU, he switched to working with Professor Sebastien Scherer’s ‘airlab’ on a project to land drones on moving trucks. After completing his MS at CMU, where he also received the Swartz Entrepreneurial Fellowship, Rastogi joined Scherer’s startup Near Earth Autonomy in 2017 to develop navigation systems for aerial drones, an extension of his airlab work.
While on holiday at home in 2018 he felt a robotics startup out of India would have unique value to offer. He joined a leading robotics company in India as an intern to gauge the market and became aware of gaps in the ecosystem.
Robot makers in India were producing automated guided vehicles (AGVs), which go along paths laid out for them. Some were venturing into the more sophisticated autonomous mobile robots (AMRs), which could find their own way with computer vision. But even the best ones had rudimentary navigation systems. “I realized they were just putting open source navigation software into their mobile robots and selling them," says Rastogi. Or they were bunging in black boxes. Rastogi figured he had a value proposition in freeing robot makers from developing autonomous navigation systems.
Mowito’s software platform saves time for the robotics company to build and deploy robots for new use cases, such as the current spike in demand for cleaning and disinfecting robots because of the pandemic. Mowito is currently running a pilot for a California-based company making robots for cleaning parking spaces. These robots were guided by GPS earlier but the company wanted them to be able to move in sheltered spaces or indoors. Engineers in the company did not have the expertise to build the navigation system. Hiring engineers would be expensive and time-consuming, so Mowito entered the picture.
Covid-19 threw a spanner in the works. Running a robotics pilot in California from Bengaluru is complicated, considering the interplays between the “intelligence layer" and hardware on one side and user interfaces on the other. But it’s the new normal, and Rastogi is doing the pilot from India.
Earlier, the fledgling Indian startup joined the 2019 winter cohort of Axilor, a startup accelerator and seed fund by tech industry doyens from Infosys in Bengaluru. “From shop floors to commercial space housekeeping, autonomous mobile robots are becoming mainstream. Mowito is building the shovels for this gold rush," says Ganapathy Venugopal, co-founder and CEO of Axilor Ventures.
Mowito has four paying clients, with four more in the pipeline, which brings it close to being cash flow positive, says Rastogi. He aims to raise funds abroad once he has the initial test cases under his belt.
Most of the attention and big bucks have gone to autonomous vehicles, or self-driving cars. There are a number of third-party providers of navigation and other software systems for this sector. Now that’s starting to happen for self-driving robots too as they get into more complex use cases.
For example, an autonomous mobile robot in a warehouse is in a much more controlled environment than a floor-cleaning robot in an airport swarming with humans. A factory robot can just stop on encountering an obstacle and resume when the path is clear. But such a system would just freeze an airport robot. It has to constantly find new paths to avoid moving obstacles like humans, thus making navigation far more complex.
German robotics firm Magazino has a product similar to Mowito’s, Acros.ai, “a software toolkit for developing and operating advanced and cooperative robots in dynamic environments." San Diego-based Brain, which raised $36 million in April, has a BrainOS to convert “anything on wheels" into self-driving robots. Its business model differs from Mowito and Magazino as it gets into service contracts with original equipment manufacturers. It recently announced a partnership with Tennant, which wants to make cleaning machines autonomous.
Rastogi had also initially toyed with the idea of starting a firm to make autonomous floor-cleaning robots, given his early exposure to handling trash at CleanRobotics. But he realized it would take longer and more funding to build a full stack robotics startup. Then a meeting with a fellow robotics entrepreneur from CMU helped him make up his mind to go for a pure software play.
Rohit Dashrathi had started Rucha Yantra in Aurangabad to make automated guided vehicles (AGVs) for the automotive industry. He told Rastogi that Rucha Yantra would be his first client if he could develop a software layer to make its AGVs autonomous. That was music to a new entrepreneur’s ears—nothing like having a customer waiting to use your product even before you’ve built it.
Sumit Chakraberty is a consulting editor with Mint. Write to him at firstname.lastname@example.org.