As anyone who has tried to navigate through cities using a smartphone can attest, GPS-based location services can be extremely inaccurate in dense urban environments, where satellite signals are often blocked by high-rise buildings and other tall structures. This phenomenon, commonly referred to as “shadowing”, can lead to location errors of over 100 meters in the most densely populated urban areas, such as New York and San Francisco, making accurate navigation difficult. The ability to precisely pinpoint GPS locations in urban environments would have a profound impact on a wide array of activities, including transportation services, delivery services, emergency e911 responses, location-based advertisement and fitness tracking.
Dr. Upamanyu Madhow, a professor of Electrical and Computer Engineering at UCSB, together with his research group, stepped up to the challenge of developing an algorithm that would compensate for shadowing. The initial idea was an unexpected outgrowth of a project on RF-enabled drones for situational awareness, involving research groups led by Dr. Joao Hespanha and Dr. Madhow, and funded by the Institute for Collaborative Biotechnologies (ICB). Dr. Madhow and his students soon realized that the blocked signals themselves could be used to correct location information by “shadow matching” the blockage against 3D maps of the environment, and potentially provide a low-cost, real-time solution to correct the inaccuracies resulting from shadowing.
The UCSB Technology Management Program provided an opportunity for early exploration of potential business opportunities for shadow matching: a team formed by graduate student Andrew Irish won in the Grand Prize and Tech Push categories in the 2014 New Venture Competition. On the academic front, graduate students Andrew Irish and Danny Iland, and postdoctoral researcher Jason Isaacs, guided by Drs. Madhow and Hespanha, won several awards at Mobicom 2014, a flagship computer science conference, with a real-time positioning prototype. In addition to Irish, Isaacs, and Iland, the prototyping effort also involved undergraduate researcher Brian Sandler (funded by an National Science Foundation Research Experience for Undergraduates grant).
Shadow Matching in San Francisco illustrating likely location on right side of the street (true path is yellow line). Satellite rays drawn from true location and color coded by signal strength.
After encouraging feedback from potential customers and partners, as well as guidance from seasoned entrepreneurs and technologists serving as advisors, ShadowMaps was formally incorporated in February 2015 with Irish, Iland, Isaacs and Dr. Madhow as co-founders. The goal was to transition research on shadow matching into software that could be used in daily life. An alpha version of its cloud-based localization technology, designed to accept location data from mobile devices and provide improved location estimates back via an Android app or an Application Program Interface (API), was deployed in San Francisco by the summer of 2015. ShadowMaps’ approach combines machine learning algorithms, 3D maps and GPS satellite signal data to correct for the shadow effect. It can be achieved through software updates without the need for new hardware, making the product easy to deploy and incorporate into existing applications.
ShadowMaps has been successful in reducing average cross-street errors by 60% and decreasing errors associated with GPS localization by 50%! The company was acquired by Uber in 2016.