GPS Independent Vehicle Localization
(September 2015 - December 2015)

• GNSS receivers suffer from the problems of GPS outage due to urban canyons. I worked on GPS independent vehicle localization using only MEMS IMU sensors. I fused the data from inertial sensors using Kalman Filter and employed Particle Filter to localize a vehicle with submeter accuracy. This work has been submitted to IEEE Vehicular Technology Conference, 2016.
• Attitude estimation from MEMS IMU under dynamic conditions suffers from large errors due to external acceleration. I created a novel Kalman Filter for determining the attitude of a land vehicle by adaptively estimating and compensating the external acceleration. This work has been submitted to IEEE Transactions on Intelligent Transportation Systems.
• I improved the results of model based external acceleration estimation from MEMS IMU sensors using SVD of noise covariance matrix in Kalman Filter. This work is under preperation for IEEE Communication Letters.
• Combining the above, we have formed a new localization technology capable of localizing vehicles with sub-meter accuracy using only low cost sensors. Currently, we are working on filing a Patent on this technology.

Word Sense Disambiguation Using Supervised Machine Learning
(May 2015 - August 2015)

• As machines are being instilled with state of the art artificial intelligence, new problems have been arising. One such problem is Word Sense Disambiguation in which a machine has to judge the correct meaning of a word with same spellings and pronunciation based on the context in which it occurs.
• I applied Supervised Machine Learning, specifically Naïve Bayes & Support Vector Machines to classify these multi-sense words and explored the effect of different data representations on the classification accuracy. This work was submitted to Scientia Iranica Journal.

Multi-Sensor Based Detection & Modelling of Abnormal Driving Behavior
(May 2014 - June 2015)

• Developed a wireless sensor network consisting of 7 sensors on the car: Accelerometer, Gyroscope, Magnetometer, Laser Range Finder, Sonar Range Finder, Camera and GPS and used 1 in-built sensor (speed sensor) of the car.
• Implemented a wireless data collection scheme to collect the values from all the sensors and fused them through Kalman Filter to obtain the parameters of interest: speed and heading of the car and the distance of surrounding traffic.
• Applied Machine Learning to predict lane changes, dangerous maneuvers, risk of accidents and developed an LCD plus GSM based notification system.
• This work has been submitted to IEEE Transactions on Intelligent Transportation Systems.