Machine/Deep Learning Controls Model Based Design HIL ADAS

***** ******** **** **, *** *** Lopamudra Baruah 906-***-****

Northville, MI 48167 www.linkedin.com/in/lopamudrabaruah ac0oel@r.postjobfree.com SUMMARY Graduate research student with two internships, & two RA experience; Strength in deep learning, machine learning, optimization, embedded controls, computer architecture. Looking for full time position starting June 2017. EDUCATION

Michigan Technological University MS Electrical Engineering GPA 3.30 May 2017 Houghton, MI University of Petroleum and Energy Studies B. Tech Electronics Engineering GPA 3.09 May 2015 Dehradun, India ENGINEERING PROJECT EXPERIENCE

Hardware-in-loop Simulation: HEV Control Design and Validation Jan 2016-April 2016

• Developed a control system for a configurable HEV using Stateflow and Simulink in MATLAB.

• Designed modules for driving logic, engine start/stop control, motor start/stop control, torque split strategy, unipolar and bipolar stepper motor diver logic.

• Validated the sub-models in Simulink using signal builder, and also using the MotoTron system, where the input signals were calibrated using MotoTune, and the outputs were displayed in Mototune display window and also the ECU.

• Implemented several improvements to get better performance of the hybrid vehicle such as modification of blend factor, Hysteresis loop Implementation in Engine Start/Stop and E-Motor Start/Stop, and Regenerative Braking and Battery Charging/ Discharging.

Establishing Remote Electronic Throttle control using CAN Mar 2016- Apr 2016

• Designed a PID controller to set up a feedback position control of Bosch DV-E5 throttle using two ECU, CAN tools, and Mototron systems to establish the CAN communication, and thus the remote Electronic throttle control was set up.

• Received the information by the other ECU and passed through a PID controller to reduce the error. The information of duty cycle was calculated and then transferred to the ECU 1.

• Tuned the parameters of the PID controller using real time calibration system.

• Defined baud rate, bus channel, Message ID, ID type for the communication.

Development of a Fuel Injection Control using sensors and actuators Feb 2016- Mar 2016

• Developed a model to control the Electronic Fuel Injector using Simulink.

• Calculated parameters like start of ignition of engine using the RPM and the crank angle position data was obtained using the Hall-Effect sensor.

• Improved the performance of the controller by adding features like Regenerative braking at higher speed by implementing the control logic in Stateflow.

• Validated the system using DC motors, fuel injector, Desktop Simulator, Hall Effect sensor, and solenoids.

Development of Hybrid Electric Powertrain Controller Using Stateflow Jan 2016 – Feb 2016

• Developed a powertrain controller for a series-parallel HEV using Stateflow using system inputs like Accelerator Pedal Position, Brake Pedal Position, State of Charge, and Vehicle Velocity to control the system outputs such as Brake Torque Request, Motor Torque Request, Regenerative Brake Torque Request, and Friction Brake Torque Request.

• Four parallel states were inside the Stateflow controller: Charging state, Battery protection State, Blending Strategy, and Braking state. The control logic of the controller decided the flow between the states.

• Improved the overall performance of the controller by reducing the load on Engine, and using more Regenerative Braking.

Cooperative Localization of Mobile nodes in a Wireless Sensor Network Oct 2015 – Dec 2015

• Simulated three nodes with different sensor accuracy which can measure their own pose and their neighbor’s pose.

• Proved using Kalman Filter that results using cooperative pose estimation are more accurate than those without information from other sensors.

Comparison of performance of Binarized Neural network (BNN) and Convolutional Neural Network. Sep 2016-May 2017

• Reducing computations complexity by setting binary weights in neural network instead of floating point.

• The aim is to show that BNN performs well and can be compared with the performance of traditional CNN.

• Worked on cloud computing services such as AWS to utilize the large computing resources that are available online.

Implementation of the J1772 control pilot's PEV charging current and duty cycle relation Mar 2016 – April 2016

• Implemented J1772 control pilot, which is used to establish the communication between the electric vehicle and the electric vehicle supply equipment. J1772 is an SAE standard for electric vehicle charging purpose.

• Established the relation between duty cycle and current using Beaglebone Black single board computer, and OpenADR software. The duty cycle determines the maximum current that is available for the electric vehicle.

• Added Beaglebone Black as a Virtual End Node (VEN) to the OpenADR. Provided the current signal using OpenADR.

• Coded in Python to convert current to corresponding duty cycle inside the VEN. Displayed the duty cycle on oscilloscope. SPECIALITY PROJECT: MACHINE LEARNING AUG – DEC 2016

Developing a regression algorithm for learning a predictive model using Matlab Aug 2016 - Sep 2016

• Used three different data sets from the UCI Machine Learning repository to test the Matlab code.

• The code performed regression and learnt a linear regression model. The results proved that Gaussian basis function gave much better results than polynomial basis function.

• Implemented 3-fold cross validation in all the data sets.

• Calculated the squared error for all the datasets and the average was calculated of the squared error of the predictions.

Implementation of a Naïve Bayes classifier in MATLAB Sep 2016 - Oct 2016

• A large text data set consisting of newspaper articles was provided with different topics.

• Training data and labels, and testing data and labels were given, where the training data contained word histograms of all documents. The labels contained the class assignment information.

• Trained a Naïve Bayes classifier using the training data, and then the learned classifier was used to predict the class labels for the test documents.

• Obtained a classification accuracy of 80% on the test documents.

Implementation of the dual form of Support Vector Machine (SVM) classifier in MATLAB Oct 2016 - Nov 2016

• Three different datasets were provided each containing the training data kernel and the training labels.

• Chose the box constraint C for each dataset using cross validation method to get an optimum value.

• Used MATLAB’s quadprog solver to implement the dual form of the SVM classifier, and the built-in SVM solver in MATLAB was used to compare the results.

• Proved the essential role that the bias term b plays in the correct classification by the SVM. The prediction equation uses the bias value to find the predicted variable.

• Classified by checking the sign of the predicted variable and thus assigning to the appropriate class

Matlab implementation of a K Nearest Neighbor and Weighted KNN Oct 2016 – Nov 2016

• Built a k-nearest neighbor (KNN) that used the leave-one-out cross validation approach, and determined the best value of k, which is the number of neighbor nodes considered.

• Training and testing data and labels were provided based on which the prediction and errors were calculated.

• Coded a weighted KNN classified in Matlab, where the weights of the k neighbors were also considered unlike KNN.

• Calculated the best λ parameter in the weighting function for WKNN.

Implementation of Kernel Logistic Regression algorithm for different kernels Nov 2016 – Dec 2016

• Coded up a kernel logistic regression in Matlab for Gaussian, polynomial, and linear kernels on heat statistics data.

• Calculated the testing error, validation error, and training error for different values of regularization parameter λ (lambda).

• Obtained an accuracy of around 80% for the dataset. INDUSTRIAL EXPERIENCE

Sistema Shyam Teleservices Limited (MTS) June 2014 – July 2014 R&D Department-RF Kolkata, West Bengal.

Worked with BSC and used Narda Radiation meter to measure RF and microwave EMF. On field-experience with conical dipole antenna and measured the antenna signal directivity.

Oil and Natural Gas Corporation Ltd. June 2013 – July 2013 Infocom Department Nazira, India.

Trained on the networking that exists in ONGC and the working of the internal communication within ONGC. LEADERSHIP AND INVOLVEMENT

Co-founder and General Secretary of "Assamese Association of UPES”. July 2011 – May 2015 Cultural Organizer and Singer in “BIHU” at ONGC, Dehradun. Jan 2012 – Apr 2015 Program Chair of "SAE INDIA UPES Collegiate Club", (Elected member). Mar 2013 – Nov 2014 Student Organizer of "SAE NIS Student Convention 2013". Oct 2013 SKILLS

• Machine/Deep Learning • FPGA, GPU • VHDL • Linear Algebra • CAN,ECU

• Embedded Sensor Network • Control Systems • Linux • C,C++,Python • MATLAB

• Hybrid Electric Vehicles • German language • NodeRed • Microsoft Office • Stateflow

• Model Based Control Design • LabView • AWS • Matlab,Simulink • MultiSim

Contact this candidate