Context-aware mobile devices

Master thesis : a learning control approach for context aware applications, where reinforcement learning is used for a mobile device system to autonomously learn to adaptive to user’s behavior pattern, and thus be able to pro-actively attend to user’s needs in a least intrusive manor.

I have created a series of utilities on the Sharp Zaurus 5500L PDA that monitors user pattern of usage for the system, and actively provides assistance when needed. (C++)Sharp Zaurus

  • Read-It: a news reader that uses a voice synthesizer to read to the user daily news which he selected before leaving home.
  • Track-me: an in-door position tracking system on a PDA that keeps track of the where-about of the user. This data is used in conjunctions with other software to make decisions about how to actively assist the user.
  • App-mon: an application monitor that keeps track of software usuage of user within applications
  • File-assist: an assistant services that uses many modality of inputs including user’s schedule, positions, and file access activity history such that it automatically retrieves the appropriate notes, slides or other related documents before the user sits down for a class or meeting.
  • Advanced power-mon: a context-aware power-management services that adapts to the user activity patterns such it can learn the appropriate power management policy that both minimizes power consumption as well as user annoyance.

Publications

  • Application of Reinforcement Learning in Multi-Sensor Fusion Problems with Conflicting Control Objectives, AutoSoft Journal, to appear, 2007. S. Ou et. al. [pdf]

Leave a Reply