Page 31 - First special issue on The impact of Artificial Intelligence on communication networks and services

ITU JOURNAL: ICT Discoveries, Vol. 1(1), March 2018

RECONFIGURABLE PROCESSOR FOR DEEP LEARNING
IN AUTONOMOUS VEHICLES

3
2
2
2
Yu Wang 1 2 , Shuang Liang , Song Yao ,Yi Shan , Song Han 2 4 , Jinzhang Peng and Hong Luo 2
1
Department of Electronic Engineering, Tsinghua University, Beijing, China
2
Deephi Tech, Beijing, China
3
Institute of Microelectronics, Tsinghua University, Beijing, China
4
Department of Electrical Engineering, Stanford University, Stanford CA, USA

"CTUSBDU The rapid growth of civilian vehicles has stimulated the development of advanced driver assistance systems
(ADASs) to be equipped in-car. Real-time autonomous vision (RTAV) is an essential part of the overall system, and the
emergence of deep learning methods has greatly improved the system quality, which also requires the processor to offer a
computing speed of tera operations per second (TOPS) and a power consumption of no more than 30 W with
programmability. This article gives an overview of the trends of RTAV algorithms and different hardware solutions, and
proposes a development route for the reconfigurable RTAV accelerator. We propose our field programmable gate array
(FPGA) based system Aristotle, together with an all-stack software-hardware co design workflow including
compression, compilation, and customized hardware architecture. Evaluation shows that our FPGA system can realize
real-time processing on modern RTAV algorithms with a higher efficiency than peer CPU and GPU platforms. Our
outlook based on the ASIC-based system design and the ongoing implementation of next generation memory would target a
100 TOPS performance with around 20 W power.
Keywords - Advanced driver assistance system (ADAS), autonomous vehicles, computer vision, deep learning, reconfig-
urable processor

1. INTRODUCTION

If you have seen the cartoon movie WALL-E, you will re-
member when WALL-E enters the starliner Axiom following
Eve, he sees a completely automated world with obese and
feeble human passengers laying in their auto driven chairs,
drinking beverages and watching TV. The movie describes
a pathetic future of human beings in the year of 2805 and
warns people to get up from their chairs and take some exer-
cise. However, the inside laziness has always been motivat-
ing geniuses to build auto driven cars or chairs, whatever it
takes to get rid of being a bored driver stuck in traffic jams.
At least for now, people find machines genuinely helpful for Figure 1. The market pattern of automotive cars.
our driving experience and sometimes they can even save
peoples lives. It has been nearly 30 years since the first In current ADASs, machine vision is an essential part; it is
successful demonstrations of ADAS [1][2][3], and the rapid also called autonomous vision [10]. Since the conditions of
development of sensors, computing hardware and related weather, roads and the shapes of captured objects are com-
algorithms has brought the conceptual system into reality. plex and variable with little concern for safety, the anticipa-
Modern cars are being equipped with ADAS and the num- tion for high recognition accuracy and rapid system reaction
bers are increasing. According to McKinseys estimation to these is urgent. For state-of-the-art algorithms, the number
[4], auto-driven cars will form a 1.9 trillion dollars mar- of operations has already increased to tens and hundreds of
ket in 2025. Many governments like those in the USA [5], giga-operations (GOPs). This has set a great challenge for
Japan [6] and Europe [7][8][9] have proposed their intelli- real time processing, and correspondingly we need to find a
gent transportation system (ITS) strategic plans, which have powerful processing platform to deal with it.
drawn up timetables for the commercialization of related
technologies.

© International Telecommunication Union, 2018 9