Ambient Intelligence System for Saving Energy (aisse)


Lehigh’s  REU projects in the area of Ambient Intelligence (AMI) are linked together by a unified vision: the creation of a mixed-initiative system (human and machine) for controlling comfort and security conditions in a living space (e.g., a house). We model this as a task prediction (for security) and optimization problem with Pareto criteria balancing comfort and energy savings. Ambient Intelligence System for Saving Energy (AISSE) aims to  help people save energy by accurately pointing out when they are doing the opposite. Through motion sensor technology, we are able to develop a system that tracks people and the objects that they interact with and determine when they are wasting and when they are saving. Implementation of this system with further refinement would allow people to be more cost effective and take better care of their environments.

Advisor: Hector Munoz-Avila

Duration: 2.5 months

Team Size: A team of two B.S. Computer Science Engineering Students.

Skills Exercised: Research, Programming, and Rapid Prototyping


An Example of A Smart Space

In this smart home example, different motion sensors are placed around different rooms, such as motion sensors and heat sensors. We that appliances work autonomously, automatically knowing what the user wants. We intended for AISSE to be modeled after these type of behaviors and features.

The Energy Saving Recognition Problem

  • People are not energy efficient.
    • AC on and windows open? 
    • Lights on during the day?
  • How can we recognize the problems and find a solution?

AISSE in Everyday Life

  • Homes for the elderly
    • Prone to being forgetful
  • Moderate energy sources in a household
    • Televisions
    • Light bulbs
    • Air conditioning and heaters
    • Fans
    • Other appliances

Our Solution

  • Automatically detect if energy is being saved
    • Done with a Machine Learning Classifier and an Asus Xtion motion sensor
    • Given:
      • Usage History: previously recorded instances already classified as saving or not saving energy
      • A new instance
  • Obtain: A classification of whether or not a user is saving energy



1. Sense Motion

  • We use a motion sensor to track a user throughout a room.
  • We then take this information and determine what the user has interacted with.
  • This is my sketch of the mock room that my partner and I created. Each block houses a different object. (Fan, Light, Bed, etc.)
  • We designed it to be a triangle and not rectangle because the motion sensor has a limited range of view. This provides for more efficiency in terms of collecting data.
  • The actual testing space is defined using lines on the floor. It is an exact replica of the sketch.


2. Sensor Readings

The motion sensor returns tracks the user’s head and returns the X and Z coordinates.

  • This is a sample of the conversion code.  I used Python to understand what the motion sensor writes and output a file that actually says which block the user is in at a given time.
  • We then take this output and create an .arff version, which is used for classification.

3. Saving or Wasting Energy?

  • Data Mining to classify the answer: Is the user Wasting Energy? Is the user Saving Energy? 
  • We used Waikato Environment for Knowledge Analysis (WEKA) to determine whether or not user is saving energy.


Future work:

  • Take readings from multiple sensors to cover a larger area
  • Be able to turn on and off appliances based off of classifier's results


Click here to view the Technical Paper.