The grant supported the company’s R&D on computer vision-based navigation for unmanned surface vessels.
BROOKLYN, NY
Manifold Robotics, a Brooklyn-based drone technology startup, completed an SBIR Phase I grant from the National Science Foundation (NSF) in October 2020, focused on advancing computer vision-based autonomy for aquatic surface drones, also known as unmanned surface vessels (USVs).
Under this SBIR project, the company started exploring computer vision to address the problems of USVs solely relying on GPS for autonomous navigation, which limits the ability to avoid hazards that may be not seen by operators. The Manifold Robotics team has successfully developed a deep learning framework to identify and classify potential obstacles encountered during USV survey operations. The framework was tested and refined on a small-scale commercial USVs for surface water surveys, which is one of the signature products of the company.
Computer vision has been embraced for use in autonomous vehicles such as cars and trucks with positive results, but the difficulty of existing frameworks in interpreting complex surface water environments has held back its widespread adoption for use with USVs. Manifold’s SBIR grant focused on addressing these challenges and opens the door for a cost-effective solution for small surface vessels in critical situations where current technologies fail.
Manifold’s research enables operations in areas with infrastructure or other obstacles, and when human intervention at a minimum is required for safe operation. The technology is expected to afford USVs to autonomously explore complex surface water environments that may contain a significant amount of obstacles, often found in urban waterways. The company has identified other potential outlets such as open-water navigation by vessels both unmanned and manned. “Although our computer vision framework is developed with small-scale vessels in mind, there’s no reason it can’t be applied to much larger vessels” Jeffrey Laut, CEO of Manifold Robotics and Principal Investigator on the grant said.
Pictured: a buoy's motion calculated by Manifold's computer vision algorithm.
The engineering team on the project took a novel approach to computer vision-based navigation. The USV is trained to assess the complexity of a potential obstacle’s motion in order to understand if it should be avoided by the vehicle or not. For example, light vegetation in a surface water environment may be classified as a non-water area by a commonly used deep-learning framework. By analyzing the vegetation’s motion using the proposed framework, it is possible to give the right decision to ignore and drive through, as this would pose a minimal hazard to a USV. On the other hand, a buoy exhibits less complex motion, and the USV would be instructed to avoid it as this would cause damage in a collision. Simply put, the USV doesn’t care what the object is, but instead how it could impact the USV during a survey.
While technology is still in a relatively early stage, the team acknowledges that their completion of the Phase I SBIR work is a significant milestone to bring an effective solution in USV autonomy. Manifold Robotics looks forward to continuing development of this technology and starting commercialization.
See the award details on NSF’s website here.
Stay tuned to our blog and subscribe for more exciting news about our computer vision work.
About America’s Seed Fund
America’s Seed Fund powered by the National Science Foundation (NSF) awards $200 million annually to startups and small businesses, transforming scientific discovery into products and services with commercial and societal impact. Startups working across almost all areas of science and technology can receive up to $1.75 million in non-dilutive funds to support research and development (R&D),helping de-risk technology for commercial success. America’s Seed Fund is congressionally mandated through the Small Business Innovation Research (SBIR) program. The NSF is an independent federal agency with a budget of about $8.1 billion that supports fundamental research and education across all fields of science and engineering. For more information, visit seedfund.nsf.gov.
About Manifold Robotics
Founded in 2016, Manifold Robotics Inc. was a spinoff from New York University Tandon School of Engineering with Jeffrey Laut as President and CEO. At that time, Manifold focused on robotics, human-machine interaction, and environmental science seeking to commercialize a small-scale autonomous robotic vehicle designed to collect data on water quality. The team was awarded a grant from PowerBridgeNY as well as an NSF SBIR Phase I award. The power line detection technology was first launched as a startup project funded by the National Security Innovation Network (NSIN), a Department of Defense innovation program office, to facilitate the transfer and transition of defense dual-use technology. The postdoctoral project, led by Jeffrey Laut and under the direction of Prof. Maurizio Porfiri, was incubated in the Dynamical Systems Lab at New York University’s Tandon School of Engineering, where the objective was to validate the technology and create a viable commercialization strategy.
