Research & Academic Projects
Masters in Automotive engineering (Autonomous Vehicles)
Autonomous Planning and Control of F1/tenth cars
Implemented Adaptive Model Predictive Control for obstacle avoidance, lane keeping and adaptive cruise control of ROS based scaled F1/10th vehicles
Used MATLAB's Robotics System Toolbox to establish communication between ROS and MATLAB/SIMULINK
Developed algorithms for obstacle detection using 2D LIDAR and lane detection using RGB camera and successfully implemented trajectory tracking for obstacle avoidance
Co-author for a paper published in SAE journal (Bulsara, Ardashir, et al. Obstacle Avoidance Using Model Predictive Control: An Implementation and Validation Study Using Scaled Vehicles. No. 2020-01-0109. SAE Technical Paper, 2020 )
Control Implementation - Software
Control Architecture and Results for sinusoidal and pulse reference trajectories
Hardware In Loop Simulation
Hardware in Loop demo between F1/tenth vehicle and SIMULINK using ROS toolbox
Obstacle position detection
Counting the number of obstacles and locating them using 2D LIDAR
Anticipatory collision avoidance for mobile robots
Used Non holonomic time to collision (NH-TTC) approach to generate collision free path for robots with various motion models
Worked on control and simulation of these robots
Gazebo Simulation Scenario 1
Three agents approaching each other
Gazebo Simulation Scenario 2
Agent approaching two agents from the opposite side
Autonomous Driving of Differential Robots
Worked on ROS platform to develop algorithms for perception, motion planning and control of Turtlebot3 Waffle pi for line following, blob tracking, human tracking and obstacle avoidance.
Simulated on gazebo 7 and then successfully tested on the actual robot.
Human Tracking
Obtained leg pose estimate from the 2D LIDAR data of the turtlebot3 and developed algorithms to track human legs
Obstacle Avoidance
Performed 2D LIDAR SLAM using gmapping and developed algorithms for local obstacle avoidance
Line Following
Used the raspberry camera and OpenCV libraries to develop algorithms for line following and blob tracking of the robot
Autonomous Drifting of robot car in simulation
Developed algorithms for Autonomous Drifting of a robot car using Probabilistic Inference for Learning Control (PILCO)
Initiated model learning using Motor Babbling and the model was trained where steering was controlled to initiate the drift at constant
speedSuccessfully simulated the trained model in gazebo 7
Discrete motion planning of a simple car using A*
Implemented A* for grid based motion planning of a simple car
Implemented algorithm with varying costs on each action (straight, turn right and turn left)
Successfully found the optimal path
With unit cost on all three actions or with higher cost on right turn
For cost value of straight, right and left are [1,1,1] or [1,10,1] respectively
With higher cost on turning left
Cost value of straight, right and left are [1,1,10] respectively
Crowd collision Avoidance
Developed crowd collision avoidance algorithm using force based repulsion and time to goal methods
Successfully tested multiple agent scenarios
3 agent scenario
8 agents scenario
Crossing agents scenario
Deep Orange 10
Perception Engineer
Autonomous Mobility Concept for Ford motor company.
Developed an algorithm for road lane detection using OpenCV libraries, thresholding using sobel absolutes, magnitudes and gradients and HSV color
Successfully tested it on the video samples and then tested it on the actual vehicle
Vehicle Detection using 3D LIDAR data
Used VoxelNet architecture for detecting the vehicles using 3D LIDAR data
Trained the CNN architecture using Kitti Dataset and successfully tested the architecture using Velodyne VLP 16 3D LIDAR
Autonomous Lawn Mower
Worked on developing a robust and a low cost autonomous lawn mower for Samsung
​Worked on Semantic Segmentation of the terrain using deep learning techniques