CAS765: Wireless Networking & Mobile Computing

Instructors: Rong Zheng

Email: rzheng

Lectures: M 9am - 12pm,

Office Hours: Tue 4:30pm - 6:30pm, ITB 121

Class web site:

Prerequisites: Matlab, Probability (understanding of Bayesian rules, total probability etc.)


With the prevalence of wireless technologies and broadband access, untethered mobile and wearable devices have become an essential part of everyday life. This class aims to cover data acquisition, sensor signal processing, and machine learning techniques relevant to mobile computing as well as state-of-the-art research in this area.

Reference books & materials: (not required)

Tentative Course schedule:



Additional Readings/notes

Sept. 14

Introduction (Slides)

Sept. 21



Sensors and sensor data processing I (Slides)

Sept. 28

Sensors and sensor data processing II (Slides)

HW1 (description, dataset) due

Oct. 5th


Sensors and sensor data processing III (Slides)

Oct. 12

Midterm recess
HW2 (description, dataset) due

Oct. 19

Data Acquisition

Android Sensing Subsystem (Slides)

[Final project pre-proposal]

Oct. 26


Bayesian filters and SLAM basics (Slides)

HW3 (description, dataset)

Nov. 2

[Final project proposal]

Nov. 9


Machine learning 101 (Slides)

HW4 (description, dataset)

Nov. 16

Advanced topics

Feature extraction (Slides)

Nov. 23

Student Paper Presentation

Nov. 30

No class (students are welcome to stop by my office to discuss final projects)

Dec. 7

[Final project demo]



  1. WiFi-based trilateration
  2. Step counting and phone attitude estimation
  3. Particular filter with IMU for indoor localization
  4. Room tag recognition

Suggested Final Project Topics (1 or 2 students per group)

  1. Robust step counting: the peak detection based step counting method introduced in the class is sensitive to the choice of filter parameters. In reality, gait patterns are person dependent. This is especially true for people with movement difficulties. The project aims to develop a robust step counting method that requires zero-configuration and is adaptive to individual's gaits. (See [JSPJ09] for robust features that can be used for step counting.)
  2. Indoor-or-Outdoor detection: one issue in indoor positioning is to determine whether a person is inside a building or outside. For example, a person is inside a building, one may swtich from using GPS to a different localization solution that works better indoor both in accuracy as well as power consumption. Thus, a naive solution that uses GPS directly will consume to much power. (See [ZZLLS12][RKSM14] on the reading list and some follow-up work).
  3. Reliable floor level detection: detecting which floor a person is at is another problem in indoor localization. Several approaches have been investigated in literatures, namely, detecting stair asending/descending and use of elevators, use of barometer sensors for air conditioned buildings.
  4. Integration of visual cues in indoor localization: visual cues such as arrows, signates, labels are abundant in indoor spaces. Extracting these visual cues using smart glasses and incorporate them in providing better localization solutions is promising(whether as part of observation models, or to label indoor space, ...).
  5. Gamification of indoor localization campaign: Fingerprinting using WiFi, magnetic fields or light intensity is shown to be useful in indoor localization. However, extensive site survey to collect fingerprints is labor intensive. Gamification on the other hand has the potential to make the process fun and attractive. The goal of this project is to design a game to attract users to collect fingerprint measurements with location tags. An alterative topic would be to design games for floor map construction or correcting existing floor maps.
  6. WiFi + IMU SLAM In practice, location of WiFi APs are often not known. This makes it challenging to apply triagulation based localization solutions. The goal of this project is to design and implement a solution that combines WiFi readings (as part of observations) and IMU sensor readings (as part of motion model) for SLAM for the estimation of the locations of WiFi APs and user locations (See [JMQSTA11]).
  7. Comparative study of power profiling/accounting approaches on mobile devices: Lightweight solutions to profiling the power usage of mobile applications are essential to drive decisions for mobile offloading and computation partitioning across different platforms (e.g., wearables, phones and cloud). This project aims to provide a quantative study of existing methods in terms of overhead and accuracy. See [DZ11], [PHZ12], [XLLZ13].

Reading list