Presentation
Context-Aware Runtime Model Reconfiguration for Energy-Efficient Autonomous Vehicle Perception
DescriptionAutonomous vehicles (AVs) require multiple sensors, deep-learning models, and powerful hardware to perceive and safely drive in real-time. However, in many contexts, some sensing modalities negatively impact perception while increasing the overall energy consumption of the system. Since AVs are energy-constrained devices, we propose a sensor fusion approach that uses context to dynamically select the computation path on an FPGA during inference using partial model reconfiguration. We show that our hardware-aware approach significantly reduces the energy used by the overall system compared to both algorithm-only approaches and traditional methods. Besides, we evaluate several system-wide optimizations encompassing sensors and hardware.
Event Type
Work-in-Progress Poster
TimeWednesday, July 12th6:00pm - 7:00pm PDT
LocationLevel 2 Lobby
AI
Autonomous Systems
Cloud
Design
EDA
Embedded Systems
RISC-V
Security