A Communitech program that supports autonomous vehicle research has helped a Kitchener company set its sights a little higher — literally.
Clearpath Robotics has used its membership in the first cohort of the AVIN (Autonomous Vehicle Integration Network) Waterloo Ventures at Communitech program to develop 3D mapping that will extend the vertical perception capabilities of its OTTO Autonomous Mobile Robots (AMRs). It’s technology that can also be applied to consumer uses of autonomous vehicles.
Alain Francq, Senior Director of Government Affairs for Clearpath, says the investment of time, funds and staff resources in this research wouldn’t have happened as easily without the support of Communitech and the AVIN program.
Clearpath was founded by four University of Waterloo engineering students more than a decade ago, with the intention of using robots to “clear a path” through minefields. The mission grew to automate the world’s most dangerous, dirtiest and dullest jobs, such as continuously surveying hectares of crops and navigating underground mining operations. In 2015, Clearpath created a division of the company, OTTO Motors, to focus on using AMRs for materials transport for manufacturing.
Clearpath is now Canada’s largest manufacturer of autonomous vehicles, says Francq, but not the kind of driverless vehicles that would ferry people to and from work. OTTO Motors’ feature product is an intelligent, self-driving, material transport vehicle that can deliver payloads up to 1,500 kilograms to precise points in complex, expansive factories or across cavernous, dynamically changing warehouse operations.
This makes Clearpath different from other companies exploring vehicular autonomy and adds to the diversity of the first cohort of eight companies in the AVIN Ventures program. That first cohort graduated this year, and the call for the next cohort has gone out, with an application deadline of June 12.
Francq says, “Most AVIN projects are looking at outdoor environments, and we’re unique in that we’re looking at how to map and navigate around large-area indoor environments.” These are dynamic spaces: “We are navigating autonomous vehicles that can reach speeds of up to two metres per second driving down narrow aisleways and around corners, while keeping track of moving forklifts and people walking past the field of view.” Francq says OTTO vehicles face the same problems that autonomous vehicles face on roads. He noted that perception, autonomy and control are vital to the safe operation of any autonomous vehicle, and it was vehicle perception and mapping for advanced collision avoidance that brought Clearpath to the AVIN project.
That was the entry point for ‘why AVIN?’ In order to increase safety and speed, you must increase the vehicle’s ability to perceive (see) its environment to better map and recognize features and dynamically navigate around moving objects. And that’s where the need for some novel research and development comes in — “we wanted to create high definition 3D semantic maps for navigating complex indoor environments in real time”, says Francq.
The AVIN program supports such research, says Mauro Rossi, Director of Advanced Technology Platforms for Communitech, through funding (up to $50,000 to be matched by $75,000 from each participant) and through access to data growth coaches, expert panels and other assistance.
Rossi says a desired outcome for AVIN Ventures is the building of peer connections: “A really great aspect of the program is the connectivity that it has made between different companies – different companies within that cohort interacting amongst themselves. This program is designed to help build that ecosystem.”
And, says Rossi, “AVIN Ventures gives companies like Clearpath the ability to improve their technology, to try and bring it to the next level, to add a new feature, a new capability that they’re seeing customers asking for, either to help them along the way or help them explore the art of the possible.”
For Clearpath, it was all about developing a new 3D scanning device for its OTTO autonomous vehicles that could collect large, dense, data sets. These data sets could be used to develop a neural network for rich semantic labelling of industrial environments. OTTO robots primarily use 2D mapping for guidance and navigation, but there are many features above and below the 2D plane, such as raised forklift tines, or a drop-off at a loading dock, which require a more robust view of the 3D space around the vehicle.
The Ventures program gave Clearpath the funds, encouragement and resources to help the team build a “data collection appliance” featuring five cameras and a 3D LiDAR that was attached to a vehicle and could collect 10 gigabytes per minute of in bound data to paint a 3D map of the environment. Data collection was done at four unique industrial sites and sent away for annotation. The resulting data sets were used to create a neural network and prototype 3D HD maps.
The OTTO engineers had to overcome several challenges along the way, including handling huge amounts of streaming data, and recording and finding the correct balance of compression quality versus onboard resources. Data annotation was also a challenge with 15 classes of objects, where poor data would impact network learning.
Francq says, “This is not a project that we would have tackled at this level without Communitech taking a leadership role in the autonomous vehicle community.” That is precisely what Communitech, the Ventures program, and AVIN and OCE (Ontario Centres of Excellence) together have done. Being a part of the AVIN cohort exposed Clearpath to companies that were exploring similar ideas, but weren’t in direct competition, so collaboration was not impeded by fears of Intellectual Property conflicts. “We’re all working on the same kind of problems,” says Francq. “Every vehicle needs to perceive, control, and have autonomy to a high level, so we’re all trying to understand and advance the same issues for our organizations.
“Having Communitech pull us all together into a cohort of like-minded companies, working on the same problems, helps advance innovation in the community.”
Find out more about AVIN Ventures Cohort 2, and join Clearpath in the AVIN community.