Qualcomm is funding a University of Colorado Denver project, “Accelerating Visual Computing Algorithms for Enhanced 3D Sensing,” lead by Dan Connors, assistant professor of electrical engineering. The $30,000 grant from Qualcomm will support student research in designing next-generation mobile processor architectures.
The core objective of the project is to explore energy efficient algorithms that can overcome the power dissipation issues of mobile platforms. Traditional mobile processors scale poorly when attempting to emulate human vision functions, and thus a novel smart architecture is necessary to enable real-time analysis with limited constraints. The proposed work enables the investigation of smart architectures that use all silicon resources (CPU, GPU and DSP) to enable human-like understanding. A component of the work will extend the group’s current framework for approximate parallel computing (APC) that allows a programmer to dynamically govern the execution of parallel tasks within processor resources with respect to specific algorithm, data, and machine characteristics.
The funded project builds on existing momentum and success of Connors’ research group in embedded systems and computer engineering. Specifically, in May, Connors lead a team of electrical engineering students to win first place in the Intel Cornell Cup Embedded Design Competition with a 3D vision system capable of executing a 3D Simultaneous Localization And Mapping (SLAM) algorithm for unmanned vehicles. Qualcomm also hired Skyler Saleh, BS ’15, as a full-time engineer based on his work on the project.