user

About me

I'm pursuing Masters in Robotics and Autonomus Systems at Arizona State University.My current coursework is interdiscplinary. I completed my bachelors in Mechanical Engineering from National Institute of Technology Jalandhar, India.I have 2 years full-time experience as Research and Development Engineer at Suzuki India Ltd.I have internship experience as Automation Engineer at Robic Rufarm. Currently I'm looking for Internship opportunities and Full time opportunities.

Skills

Experience

Education

  • Robotics
    ROS, gazebo, SLAM
  • Coursework
    Machine Learning, Deep Learning, CNN
    Computer Vision, Motion Planning, Vehicle Dynamics
    IC Engines, Design of Machines, Controls and Modelling
  • Languages
    Python, C++
  • Framework
    Tensorflow, Pytorch, Keras
  • Mechanical
    CATIA V5, UGNX, Solidworks, MATLAB
    INCA, PUMA, AVL Concerto
  • Research Volunteer - Apr 2023 to Current
    Arizona State University
  • • Implemented Probability edge detection algorithm from scratch resulting better than canny and Sobel detectors.
    • Detected corners using Harris corners and stitched images and created panorama of image using homography and perspective transform.
    • Used Neural Radiance fields (NeRF) to synthesize novel views by optimizing a continuous volumetric scene function using a sparse set.
    • Reconstructed 3D scene using Sfm(Structure from motion) with different viewpoints using state of art classical approach.
    • Developed pipeline of ResNet and VGGNet for image processing and trained on large image data.
  • Suzuki India Ltd - July 2020 to July 2022
    Research and Development Engineer (Assistant Manager)
  • • • Prepared Design Validation, Product Validation and DFMEA reports for troubleshooting meetings and reverse engineering
    • • Performed accelerated aging testing on hardware components of engine and evaluated life cycle of products.
    • Performed Measured Data Analysis for debugging and automating new test algorithms in ECU and test frame via CAN and ETK.
    • Designed Jigs, fixtures, and spare parts in UGNX for the testing purpose and optimized equipment for better insulated testing.
    • Data Acquisition and root cause analysis by investigating test data in AVL PUMA, Concerto.
    • Validation of Components through a set of endurance, function, friction test cycles execution by identifying hardware critical parts.
    • Identified anomaly case in testing by implementing machine learning techniques after collection of imbalanced data of 5% minority
    • Developed real time test data graph generation of Measured data and curve fitting using recursive least squares.
  • Robic Rufarm PVT Ltd- June 2019 to July 2019
    Design & Automation Intern
  • • Automated a length measuring machine by integration of pulleys, motors, VFD and rotary encoders.
    • Measured length by rotary encoders and calibrated to Arduino code to deliver tested products with budget cut by 40% for customer needs by benchmarking market specifications. Experience of working in small team(startup).
  • Masters in Robotics at ASU : Aug 2022 - Current
    CGPA 4.0/4.0
  • Bachelors in Mechanical Engineering at NIT Jalandhar :
    2016 - 2020
    CGPA 7.52/10

My Work

Face Mask Detection

Point cloud NERF

Swarm Robot Navigation

Turtlebot4

Mapping

Parrot Mambo Drone Path follower

Image Stitching

Planar Drone

Multi Resolution Face Blending

My Projects

Face Mask Detection

• Real time Face Mask Detection using OAK-D camera feed of turtlebot4 by deploying Neural Networks.
• Made our own Neural Network for the feature generation by taking training data of images from OAK-D camera
• Deployed model to ROS node which does face mask recognition in real time by subscribing to real time data current_frame.

OVERALL ROS GRAPH
user
FLOW
user user

Point Cloud NERF

• Implemented a pipeline outperforming NeRF using point cloud based NeRF by including depth loss function.
• Implementation of ICP point cloud registration, pruning techniques and developing neural point clouds.
• Improved density of point cloud using Recurrent Neural Networks.

OVERALL PIPELINE
user
Point Cloud and Individual views of clouds
user user Complete Point cloud
user user

Swarm Robot Navigation

• Developed swarm robot navigation technique for surveillance using utility maps and correlation density mapping.
• Developed A-star and dijkstra algorithms for path shortest path generation from maze start to end and weighted branches.
• Developed trade-off between cost and time, negotiated dynamic obstacles by shape navigation and Artificial potential fields.

OVERALL PIPELINE
user
Vector fields around the obstacle
user user
Shape Navigation function
user user

Statistical Machine Learning

• Built Machine learning algorithms such as Linear regression, Logistic regression, K Means, Naive Bayes, Gaussian Distribution, Perceptron, SVM, decision trees, Adaboost, Gradient Boost, XG Boost, random forest, and Neural Networks from scratch.
• Applied Lasso, sparse, Coordinate descent, Gradient descent, and Stochastic gradient descent from with different algorithm from base.
• Induced Boosting technique for imbalanced dataset of Wi-Fi dataset, defected glass datasets for better F1-score of 0.85.
• Compared F1 score, Accuracy, G means of Adaboost, Gradient Boosting and logistic regression performance on the datasets.
• Implemented all ML algorithms on MNIST and CIFAR 10 dataset using the statistical approach without using Canny algorithms.

BIAS and VARIANCE
Bias and Variance Classification and the Training and Testing error w.r.t iterations.
user user

GRADIENT DESCENT
Gradient Descent with cost function and error.
user user

STOCHIOSTIC GRADIENT DESCENT
Stochoistic Gradient Descent with cost function and error with batch size of 1.
user user
Stochoistic Gradient Descent with cost function and error with batch size of 100.
user user

Regualrization and K Means
Regualrization of weights and K means for the number Classification.
user user

LASSO
LASSO implementation and count of number of zeroes for sparse configuration.
user user

RBF and Poly functions
RBF and Poly Kernel which are combined functions of the radial individual functions.
rbf user

Bootstraping and K means for RBF and Poly functions
RRBF and Poly kernel implementation with K fold and Bootstraping to increase number of samples.
user user user user

Turtlebot4 Applications

• Experimented with computer vision techniques by creating dataset using stereo camera of turtlebot.
• Collected IMU data and implemented Kalman filter and FIR filter for the velocity and acceleration prediction.
• Implemented line follower using the computer vision thresholding, mapped surrounding environment using bot.

Student Prediction Kaggle project:(Dataset of 60000 samples with 31 features)

• Performed feature engineering and generated heat maps to find the best features to develop algorithm.
• Implemented XG Boost, SVM, Neural Networks and Logistic regression to find the best F1 score of 0.71 on testing dataset.
• Compared performance time and F1 score for the dataset and implemented hyper parameter tuning along with K means.

FEATURE ENGINEERING
user
CHECKING F1 SCORE AMONG METHODS
user

Edge detection and panorama

• Implemented Probability edge detection algorithm from scratch resulting better than canny and Sobel detectors.
• Implemented Deep learning techniques along with Adam Algorithm for the better accuracy over the training dataset.
• Detected corners using Harris corners and stitched images and created panorama of image using the concept of disparity.
• Performed the tasks using the classical state of art and deep learning as well.

Robot Navigation

• Developed code for Robotic Arm navigation avoiding static obstacles using Inverse kinematics solutions and iterative loops.
• Reviewed navigation of soft robotics by exploiting the contact with the environment by lumped mechanism.
• Developed Algorithm for the optimization of waypoints for the better path exploiting obstacles than nominal path.

Automation of Length Measuring Machine

• Automated a length measuring machine by integration of pulleys, motors, gear train, VFD and rotary encoders.
• Designed product in CATIA and fabricated prototype for testing with actual edge band tape product.
• Measured length by rotary encoders and calibrated to Arduino to deliver tested prototype with budget cut by 40% for customer needs by benchmarking market product specifications for small scale industries.

BAJA SAE INDIA

• Managed to develop timer by LDR and Laser setup for calculation of vehicle acceleration with the Arduino RS 232.
• Simulated CVT in MATLAB for optimal acceleration by reducing lag of 0.5 sec and validated specifications of CVT for different terrains.
• Implemented MATLAB code and technically validated flyweights and variables of CVT for different terrain by physical testing.
• Designed and fabricated a gearbox, CVT, driveline of ATV using FEA analysis (Ansys) and decreased rotational weight by 20%.
• Fabrication, assembly, and disassembly of powertrain components of ATV for 3 years and cut down collaborative service time by 30%.

Design elements
user
user
user
user
user
user
user
user

PUBLICATION

• Performance evaluation of PV module integrated with PCM filled container with external fins for extremely hot climates.
• Worked as Research Assistant and completed project which got published in Elsevier, ‘Journal of Energy Storage’.
• Attached PCM along with fins on rear to a PV panel to mitigate the drop in power output by 6.59%.
• Automated wiper system on PV panel for growth of efficacy by 10% using Arduino and case study of dust accumulation.

Contact Me

praveenpaidi4@gmail.com

Download CV