Data Science Talk Series

Title: Building Predictable and Energy-Efficient Computing Systems for Autonomous Vehicles/Robots

Abstract:

Owing to their high safety and efficiency, autonomous vehicles (AV) have attracted colossal investment and studies from industrial and academic communities. Since the primary goal is to understand the driving environment and drive the vehicle safely, building a reliable and efficient computing stack becomes one of the fundamental challenges for autonomous driving. In this talk, I will cover two works. The first work is "Prophet: Realizing a Predictable Real-time Perception Pipeline for Autonomous Vehicles", which addresses the DNN inference time variations in AV's perception. The second work is "An Open Approach to Energy-Efficient Autonomous Mobile Robots", which improves the energy efficiency of autonomous mobile robots.