Reconfigurable Approximate Computing in Autonomous Systems
Date:
This invited talk explored Reconfigurable Approximate Computing in Autonomous Systems, focusing on the balance between energy efficiency and trustworthiness.
Key themes included:
- Approximate Linear Algebra and FPGA-based reconfigurability
- Approximate accelerator generation and automation
- Algorithmic and hardware resilience under approximation
- Energy-efficient 3D depth reconstruction
- Approximate model predictive control and PID controller benchmarks
- Power scaling in fine-grained approximation
The talk highlighted both the opportunities and challenges of deploying approximation in autonomous systems under low-SWaP (Size, Weight, and Power) constraints, emphasizing reliability, scalability, and security.
