PROJECTS

Turtlebot 3 Burger

Turtlebot 3 - Burger

TurtleBot Following AprilTag

Camera Calibration with RapiCam. AprilTags were detected using ROS packages. Proportional Controller was used to run the bot.

Behavior Planning of TurtleBot

Turtlebot follows a wall. Avoids the obstacles in its path. Follows the yellow line and avoids the human in its path. A statemachine was used for Behavior Planning.

Line Following

Path Detection with open CV library in python. P controller for following the line.

Wall following on real robot

Wall following on the basis of distance from the walls as the error for P controller.

Obstacle Avoidance In Gazebo

Obstacle Avoidance On Turtlebot3 burger

Path following in Gazebo

Wall Following in Gazebo

Moving in a square shape path

Hardcoded motion in the turtlebot for following a square trajectory.

Obstacle Detection And Collision Avoidance

Code

Motion Planning Algorithms

RRT* Algorithm With Dubins Car

RRT algorithm with Dubins Car

Probabilistic Road Map (PRM)

Force Based Local Navigation - Crowd Crossing

Force Based Navigation - 8 Agents

A* Algorithm

Breadth FIrst Search

Depth First Search Algorithm

Code

F1 Tenth Scaled Vehicle

Autonomous Cruise Control and Lane Keeping Assist

Calibrated 4 ultrasonic sensors, developed calibration functions.
Implemented Kalman Filter for sensor noise reduction, and sensor fusion. Longitudinal and lateral PID controller.

Follow The Gap

Model Predictive Control on F1Tenth Simulator

Stanley Controller in F1Tenth Simulator

Wall Following

Code

Deep Reinforcement Learning

Reinforcement Learning

Crossing Intersection with PPO

Evolutionary Staregies on Lunar Lander

Bi-Directional Traffic Crossing with DDQN

Unidirectional Traffic Crossing by DQN

Crawler Learns to Walk with SARSA

Double Q-Learning

Value Iteration

Solving the grid.

Q Learning

Solving grid with Q Learning.

SARSA

Applying SARSA on the grid.

Code

IROS 2020

IROS

Disparity Extender

Implemented for IROS 2020,
8th F1tenth Las Vegas Grad Prix (Scaled Autonomous vehicle Racing)

Map

Moto-GP track map used in the competition.

Code

CARLA Simulator

Pure Pursuit Controller

Implemented Pure Pursuit Controller in Carla Simulator.

DQN on F1Tenth

Deep Q Network on F1 Tenth Simulator with ROS

DQN with 3 action resolution.

DQN on Track with Obstacles

DQN on Berlin F1Tenth Map

Used DQN with 3 steering actions resolution

DQN with 9 Steering Actions

9 steering angle action resolution.

DQN with 15 actions resolution

15 Steering Angle Resolution

DQN on Donkey Car

Code

Traffic Sign Detection and Auto Lane Keeping

Simultaneous Lane Keeping and Traffic SIgn Detecton

Simultaneous Lane Following and Traffic Sign Detection

Transfer Learning on RCNN with CIFAR-10 data set for Traffic-Sign Detection on Machine 2. UDP/IP protocol to communicate between two machines. Utilized Hough Transform to identify the lane lines on a pre-recorded camera feed. Lane Keeping Algorithm (Stanley Controller) on Machine 1 to find the steering angle.

Software Architecture Layout

Code

PROJECTS

Turtlebot 3 Burger​

​​

Motion Planning Algorithms

​

F1 Tenth Scaled Vehicle

​​

Deep Reinforcement Learning

​

IROS 2020​

CARLA Simulator

Deep Q Network on F1 Tenth Simulator​ with ROS

​

Traffic Sign Detection and Auto Lane Keeping

Turtlebot 3 Burger

IROS 2020

Deep Q Network on F1 Tenth Simulator with ROS