Hardware Accelerators for Autonomous Cars: A Review

TL;DR

This review paper discusses hardware accelerators and machine vision systems in autonomous vehicles, analyzing current technologies, challenges, and future directions to improve reliability and performance.

Contribution

It provides a comprehensive analysis of recent hardware and algorithmic developments in AV perception systems, highlighting pros, cons, and potential future research avenues.

Findings

Current AV systems lack reliability due to accidents and licensing issues.

Various hardware technologies have been evaluated with their respective advantages and disadvantages.

The paper suggests pathways to enhance hardware and algorithmic robustness in AV perception.

Abstract

Autonomous Vehicles (AVs) redefine transportation with sophisticated technology, integrating sensors, cameras, and intricate algorithms. Implementing machine learning in AV perception demands robust hardware accelerators to achieve real-time performance at reasonable power consumption and footprint. Lot of research and development efforts using different technologies are still being conducted to achieve the goal of getting a fully AV and some cars manufactures offer commercially available systems. Unfortunately, they still lack reliability because of the repeated accidents they have encountered such as the recent one which happened in California and for which the Cruise company had its license suspended by the state of California for an undetermined period [1]. This paper critically reviews the most recent findings of machine vision systems used in AVs from both hardware and algorithmic points of view. It discusses the technologies used in commercial cars with their pros and cons and suggests possible ways forward. Thus, the paper can be a tangible reference for researchers who have the opportunity to get involved in designing machine vision systems targeting AV