Deval Shah

Ph.D. Candidate, Department of Electrical and Computer Engineering, University of British Columbia   (Expected graduation date: January 2024)

Thesis Advisor: Prof. Tor M. Aamodt

Thesis Title: Energy-Efficient Acceleration for Autonomous Robotics

Summary: Autonomous robots have great potential to improve our day-to-day lives. Autonomous robots are complex systems that perform several latency-sensitive and energy-consuming computational tasks to sense their environment and act toward a goal. In this dissertation, I focus on improving the energy efficiency of robot perception and planning tasks.

The first part of my research focuses on reducing redundant computation in robot motion planning using application-specific characteristics for algorithm-hardware co-design. Further, I explore an application-specific reliability metric for cost-effective soft-error mitigation in motion planning hardware accelerators. The second part of my research focuses on deep regression networks and proposes a framework and taxonomy for regression by binary classification. The proposed approach improves accuracy for a wide range of regression problems and enables the use of sparser and smaller neural networks for energy-efficient perception and

planning in robotics.

Work experience: Qualcomm India Pvt. Ltd., Bangalore

M.Tech.: Microelectronics and VLSI, IIT Bombay

B.Eng.: Electronics Engineering, M. S. Univerisity