Sub-cent processors, sensory computers, and space microdatacenters: Novel systems for ubiquitous computing

Computing devices and systems are pervasive and have an ever-increasing footprint — now over a trillion semiconductor integrated circuits are sold each year. From business machines to personal computers to mobile computing and the internet-of-things, the scope of applications supported by computing has continued to expand. Thus, it may sound surprising that a vast set of domains still have not seen much proliferation of computing. Broadly, the reasons which a domain may be unserved or underserved by computing are cost, capability, and need. For high-volume, low-margin industries (e.g., groceries, discount retail, agriculture, shipping and logistics, etc.), the cost of computer chips can be prohibitive. For some applications, the chips themselves lack required capabilities or characteristics — visually opaque electronics may be difficult for a surgeon to implant into a patient; worn rigid, non-conformal systems may cause physical discomfort; a system which meets a required form factor or power limit may not have the computational capability to meet an application’s requirement, etc., as is the case with legacy audio SoCs, which struggle to handle the intensive computation of new signal processing and machine learning based applications. Since the advent of the artificial satellite, Earth observation has been a key space application which, none-the-less, has historically not required significant on-satellite computing — data is transferred to Earth for processing. However, with the Cambrian explosion of satellites (from 120 satellites launched in 2010 to nearly 2700 in 2023), the need for in-space computing to reduce pressure on band-limited communication channels is becoming apparent. Similarly, only recently have chemical and odor sensors improved to the point where computation, and not sensing is an energy and power bottleneck for olfactory sensing systems — a need for computers targeting olfactory workloads is a recent development. The work presented in this thesis — which includes architectures, chip designs, and an automation tool — focuses on these unserved and underserved application domains. Part I focuses mainly on cost — through the use of novel architectures fabricated with printing and flexible electronics processes, chip bill-of-material may be lowered to support an expanded set of applications. These flexible electronics also meet conformality and thinness constraints. Part II focuses mainly on capability — novel reconfigurable and programmable hardware accelerators may provide the performance and energy efficiency needed to support sophisticated artificial intelligence and signal processing applications for wearable and internet-of-things electronics systems, such as earbuds, hearing aids, and olfactory & chemical processing systems. Part III focuses on need, specifically in the space domain. This part makes a case for the use of space microdatacenters — relatively large-scale, server-based computing systems used to support the large and growing number of Earth observation satellites in low-Earth orbit. This research will, hopefully, lead to the continued expansion of computing into new domains and new applications. Just as the low price-points of computers such as the Zilog Z80 and MOS 6502 ushered in a new era of home computing, sub-cent processors enabled by flexible and printed electronics processes may bring forth a new era of truly ubiquitous computing. New architectures which are both more performant and lower power than traditional architectures may pave the way for sophisticated artificial intelligence applications embedded in wearable and other small form-factor consumer electronics. And in-space computing, both server based and ‘edge’ computing looks primed to increase the amount of data we can generate in space, and reduce its processing latency.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Bleier, Nathaniel

Contributors dc:contributor

• Kumar, Rakesh
• Hanumolu, Pavan
• Kim, Nam Sung
• Mudge, Trevor
• Torrellas, Josep

• In-space Computing
• Flexible Electronics
• Printed Electronics
• Olfactory Computing
• Dataflow Architectures

Statement dc:rights

• Copyright 2024 Nathaniel Bleier

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/127131

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F127131