badri-r-s Goto Github PK

2d_particle_filter_for_localization

This project involves a vehicle that is lost in a map. It uses an initial estimate of it's position from GPS and distances from landmarks measured by LIDAR and utilizes a particle filter to localize itself in the map

ackermann_kinetic_model

Designed and implemented a controller for an Ackermann kinematic model that could take inputs such as goal location and goal heading and apply the appropriate control input to the motors of the vehicle. The inputs were target heading angle and velocity. The outputs were the steering angles and angular velocities for each of the front wheels.

advanced_lane_detection

anomaly-detection-robot

apprenticeship-learning-via-irl-f1tenth-gym

Implmented Apprenticeship learning via IRL for the f1tenth gym that uses ROS2 Foxy.

astar-diffdrive-robot

behavioral_cloning

Implementation of Behavioral Cloning algorithm to train a car to autonomously drive in a simulator.

custom_global_planner

Implementation of A*,Dijkstra,RRT and BFS in C++ for Turtlebot3 and Gazebo

delbot

The aim of this project is to build an omni directional mobile manipulator that can be used for delivery of groceries and food.

dwa_dijkstra_turtlebot3

Simulated and analyzed different path planning algorithms such as DWA and Dijkstra in a custom world modeled in Autodesk Fusion 360. The modeled world was imported to Gazebo, and the paths planned by turtlebot3 were visualized in Rviz.

ekf_sensor_fusion

Implementation of extended kalman filter in C++ that fuses data from lidar and radar to track a bicycle moving around a car

enpm662_project1

Designed a custom mobile robot with independently steerable front wheels in SOLIDWORKS and exported it into the Gazebo simulator and teleoperated with the ROS teleoperation node. Additionally, a publisher and a subscriber node were written to control the robot.

enpm667-project-2

Designed an LQG controller using MATLAB for a system consisting of a mobile crane and two pendulums suspended from it. The non-linear system was linearized, and an LQR controller was designed for both linearized and non-linearized systems. A Luenberger observer was designed for a set of observable states and an LQG controller was designed.

enpm808x-week5

ENPM 808X week 5 project

extended-kalman-filter

Implemented the extended Kalman filter for the state (x, y, yaw) estimation of a differential drive robot moving in a circle. The robot had been simulated using a Gazebo physics simulator.

implementation-of-hector-slam-and-autonomous-navigation

Designed a mobile robot controlled using Raspberry Pi. Used hector mapping and AMCL to map the environment and localize the robot. Dijkstra Algorithm was used to autonomously navigate the robot.

introspective-vision-slam

•Implemented and innovated a SLAM (Simultaneous Localization and Mapping) solution based on Introspective Vision.Specifically focusing on scenarios where scene features vary across different angles, developing methods to intelligently merge disparate features, thereby enhancing overall feature extraction efficiency and localization of robots

lqr_ilqr_ddp_feedback-linearization

Designed LQR and iLQR controllers for multiple non-linear systems describing scenarios such as cart-pole balancing, helicopter hovering, hopper stabilization, and trajectory following for helicopter flight.

pid_controller_tdd

This is the implementation of a basic PID controller with Test Driven Development.

pitch_yaw_estimation

Predict the direction of travel in camera frame from given dashcam videos

ros2_roomba_obstacle_avoidance

Simulation of obstacle avoiding robot in ROS2 using turtlebot3

rrt_connect_diffdrive

Implemented RRT_connect on turtlebot3

ukf_sensor_fusion

Implemented a Unscented Kalman Filter and Sensor Fusion to track the position, velocity, yaw and yaw rate of a bicycle maoving closely around a car.

value-iteration-and-maxentropy-iteration

Implemented value iteration to solve a Markov decision process describing a two-dimensional grid with an agent trying to reach from one corner to another. The algorithm was tested with a deterministic policy that maximizes the reward at each step, and a maximum entropy policy that maximizes entropy at each step.

visualodom

Estimated the motion of a robot using classical computer vision techniques such as the point-n-perspective, local outlier factor, LMeds, and stereo vision with the KITTI dataset.