A low-cost analog computing system, developed at Massachusetts Institute of Technology Lincoln Laboratory (MIT LL) to solve problems that are too complicated for current digital computing technology, is being commercialized by a start-up founded by former MIT LL scientists.
Specifically, the new system is designed to solve combinatorial optimization problems—those involving so many variables that sifting through all the possible combinations to arrive at an optimal solution cannot be done efficiently by traditional digital computers.
Combinatorial optimization problems span various industries. The solutions can be used, for example, to design efficient supply chain logistics, operate telecommunications systems with minimal interference, and discover new drugs.
MIT LL’s Optimization Processing Unit (OPU) employs what the developers call “an algorithm in hardware form”—a free-space optical coupling scheme that transmits information between oscillators without the need for wires. Inspired by nature, the OPU has been designed to find a path to optimal energy use for each problem it encounters. That path reflects the best solution to the problem.
The technology transfer story began in January 2019, when MIT LL scientists Jeffrey Chou, Bill Herzog, Suraj Bramhavar, and Siddhartha Ghosh filed an invention disclosure for the OPU technology, titled “All-to-All Connected Ising Machines with Free-Space Optical Interconnects.”
A provisional patent application was subsequently filed in March 2019. Just before leaving MIT LL, Bramhavar and Chou filed another invention disclosure in October 2019 titled “All Electrical Fully Connected Coupled Oscillator Ising Machine” and a second provisional patent application was filed shortly thereafter.
A Cooperative Research and Development Agreement (CRADA) is currently being established with Sync Computing as part of the Defense Advanced Research Projects Agency (DARPA) Activate Fellowship program.
Bramhavar and Chou created the startup Sync Computing to license the relevant intellectual property. Like MIT, Sync Computing is based in Cambridge, Massachusetts. The company received venture capital support from The Engine (an MIT spinoff dedicated to commercializing “tough tech”) and Underscore.vc.
A nondisclosure agreement with Sync Computing and MIT LL was signed in November 2019, immediately after the company’s formation. A collaboration for engineering services for “Coupled Oscillator Computer Support” was issued November 2019. Sync Computing was able to further debug the board and refine the prototype for MIT LL, which was mutually beneficial.
A Cooperative Research and Development Agreement (CRADA) is currently being established with Sync Computing as part of the Defense Advanced Research Projects Agency (DARPA) Activate Fellowship program. MIT LL serves as an Activate anchor partner under a CRADA where fellows embed to develop new solutions related to microelectronics, from fabrication techniques to sensor and chip architectures. Sync Computing founder Bramhavar is anticipated to begin this fellowship as a DARPA Entrepreneurial Fellow in early FY21.
The Sync Computing team is currently working on building and scaling the OPU and assessing the market verticals in which it will be most useful. MIT LL has continued work for the Department of Defense (DoD) on maturing this technology, and continues to do research on computing with coupled oscillators for government sponsors.
The transfer to Sync Computing provides a pipeline into the commercial sector. Potential commercial applications include largescale materials simulations; drug discovery; satellite, vehicle, and logistics optimization; financial portfolio optimization; accelerated deep learning; and unsupervised machine learning. A higher functioning prototype and scaled-up manufacturing are also being explored for DoD and national security applications.