Energy control and user impatience

1
Energy control and user impatience

• Daniel Mosse

2
Power Model

• CPUs can vary frequency and voltage, screen
  brightness, etc.
• The strategy is if there is time, slow down
  processor to save energy (and it does!)
• How much, then, can we slow down tasks?
• Can we dim the screen to save energy?
• Can we slow down the network to save energy?
• Anything else?

3
User Impatience

• A smart phone runs several applications at a time
• When running the applications, not all compute
  power of the CPUs is needed
• A smart software platform (SSP) will decrease the
  CPU frequency to save energy, dim the screen,
  slowdown the network, etc.
• An impatient user will yell or shake the smart
  phone (input to the phone that it needs to
  improve)
• SSP logs the user interactions and create user
  profiles
• The SSP will then increase the frequency to make
  users happy again

4
User Impatience (II)

• The SSP youll build keeps track of
• The mix of applications running
• The screen brightness and/or CPU frequency at
  which the user is happy
• The rate of impatience demonstrations
• Stats about CPU frequency (increase/decrease)
• The SSP adjusts the frequency to maximize user
  happiness
• More battery life
• Less complaints (impatience demonstrations)

5
3 areas of projects

1. User interface creating/testing interfaces,
   developing/porting existing system to Android
   Framework, creating graphs that are user
   friendly.
2. System support android support, creating
   efficient data stores (DBs), making data
   available to others, etc.
3. Machine learning how to mine the data and learn
   user behavior wrt phone how much and how often
   to increase/decrease frequency, how to predict
   user impatience, how to avoid user becoming
   impatience!!

6
Student Profile

• Highly INDEPENDENT
• Need to know (or want to learn) Android and
  perhaps Android kernel programming
• Need HPT (High Pain Threshold)
• Need to test extensively and automatically (not
  manual testing)