Lidar Obstacle Detection with ROS
Autonomous Maze Solver Robot
- π€ Developed an autonomous maze-solving robot using TurtleBot3 and ROS
- π Integrated RPLIDAR sensor for real-time obstacle detection and avoidance
- π Implemented intelligent wall-following and search algorithms
- π§ͺ Leveraged TurtleBot3 for rapid prototyping and navigation strategy testing
Skills Demonstrated:
- π οΈ Mobile Robotics
- π Navigation Algorithms
- π‘οΈ Sensor Integration
- π» ROS Development
(Video 1) (Video 2) (Video 3)
Real-Time Mask Detection using Convolutional Neural Networks (CNN)
- Developed a real-time facial mask detection system using a custom Convolutional Neural Network (CNN) model built with TensorFlow
- Designed and trained CNN architecture to accurately identify the presence of masks on faces in video streams
- Optimized model for real-time performance, achieving <100ms inference time per frame on webcam feeds
- Achieved 96% accuracy in detecting masked individuals in real-world conditions, ensuring reliable monitoring
- Demonstrated expertise in computer vision, deep learning, real-time analytics, and deploying AI solutions
(Video) (Code)
π οΈπ Mentored Projects ππ
Virtual Reality Benchmarking for Egocentric Fast Target Prediction Algorithms
Augβ22 - Decβ22
- The project was done as part of the βRobot Perceptionβ Course at NYU
- Designed and implemented an immersive VR game on Oculus for benchmarking hand-target prediction algorithms
- Egocentric Hand-Target Reaching Datasets: A compiled dataset of 21 right-hand trajectories with varying object distances
- Benchmarking and Standardized Metrics: Developed standardized performance metrics for objective algorithm comparisons
- Provided background research, raw tracking data, and documented procedures to inform system design
- Contributed to creating an engaging VR environment for reliable and practical algorithm testing
- Enabled quantitative benchmarking of prediction algorithms through an interactive game format
- Demonstrated skills in VR development, data collection, benchmark design, and documentation
(Paper)
Modelling and Control of a Quadrotor System
Final Year Project, IIEST Shibpur | August 2020 β May 2021
Supervisors: Dr. Ashoke Sutradhar, Professor, IIEST Shibpur
- Developed nonlinear dynamic model and control system for a quadrotor using Newton-Euler formalism in MATLAB/Simulink
- Derived mathematical equations representing rotational dynamics, translational dynamics, and orientation transformations
- Designed PID controllers for altitude, attitude stabilization and trajectory tracking
- Tuned gains and simulated quadrotor motion in 3D environment
- Achieved altitude settling time of 10 sec. and attitude tracking error of 10% using tuned PID controller
- Project demonstrated strong proficiency in mathematical modeling, control system design, and simulation of unmanned quadrotor systems
(Report)
Stabilization of Single Link Manipulator
IIEST Shibpur | December 2019 β January 2020
Supervisor: Dr. Aparajita Sengupta, Professor, IIEST Shibpur
- Developed control strategy to stabilize a single link robotic manipulator using Lagrangian formulation
- Derived mathematical model representing dynamics of the system.
- Linearized model and designed PID controller to achieve stabilization
- Simulated controller performance in MATLAB and analyzed results
- Obtained the average steady-state error to be around 0.5 degrees for controlling arm position using PID controller
- Project demonstrated expertise in modeling, control theory, and simulation of robotic systems
(Report)