Chinese
English
Research Fellow (Professor)  |  Hsiu, Pi-Cheng  
 
contact
vita
education
experience
interests
descriptions
activities
honors
publications
lab (New window)
 
 
 
 
 
Research Descriptions
 

Mobile computing has led to paradigm shifts in user behavior, application semantics, and device features. However, existing “hasty solutions” that borrow legacy designs directly from conventional operating systems cannot be applied seamlessly to mobile systems, particularly in conjunction with cloud computing. We conduct scientific and engineering research that would affect the future designs of mobile systems in the interconnection of smartphones, wearables, and the cloud. To give consideration to both academic value and practical use, we are devoted to not only original design concepts and system implementations but also theoretical derivation and analysis. Our research focuses include cloud-based energy saving services (2009-2014), user-centric mobile systems (2013-2018), self-powered intermittent IoT Systems (2016-Present), and embedded deep learning (2016-Present).

Current Research Focuses:

(1) Intermittent Operating System: We have recently developed an intermittent OS, which makes IoT devices operable in unstable power environments and intermittence transparent to programmers. With the OS installed, IoT devices will be endowed with the capability of (i) progress accumulation across power cycles and (ii) instant recovery from power failures. Our vanilla version is released under an open-source license and available at https://github.com/EMCLab-Sinica/Intermittent-OS. We are now extending the OS to add multi-core, I/O, and other supports.

(2) Intermittent Deep Inference: We have recently developed an inference engine and API to enable hardware accelerated intermittent DNN inference on battery-less IoT devices. The API helps AI developers to easily build low cost, intermittent-aware inference systems, and has been released at https://github.com/EMCLab-Sinica/HAWAII_Project. We are now extending the tool kits for more types of accelerators and neural networks.

 
 
bg