• 🔭 I’m currently working on: Autonomous robots and drones for GPS-denied environments
• 🌱 I’m currently learning: Difusion Models, NeRF, Autoware AI
• 👯 I’m looking to collaborate on: Robotics projects, AI/ML, and computer vision applications
• 💬 Ask me about: Robotics, drones, sensor fusion, and SLAM
• 📫 How to reach me: ([email protected]) | LinkedIn
• ⚡ Fun fact: I’m an avid traveler and love building things from scratch!

About the Project: This project explores a collaborative system where a TurtleBot (UGV) and Crazyflie drone (UAV) navigate through a set of predefined waypoints generated using the Traveling Salesman Problem (TSP). The TurtleBot acts as a mobile platform for the UAV’s takeoff and landing, while autonomously performing obstacle avoidance, waypoint navigation, and dynamic waypoint resampling. The UAV, after take-off, visits multiple waypoints and returns to land on the TurtleBot at its final waypoint. This report details the system’s design, implementation, and performance evaluation. Watch the video

About the Project: The primary goal of this project is to design, implement, and train a reinforcement learning (RL) agent capable of autonomously navigating a vehicle through a variety of urban driving scenarios within the CARLA simulator. The agent will be required to make decisions in real-time that allow it to avoid obstacles, adhere to traffic regulations, and successfully reach predetermined destinations without human intervention. This project aims to demonstrate the feasibility and effectiveness of using RL techniques in solving complex navigation problems in simulated environments that closely mimic real-world conditions.

About the Project: This project investigates the methodologies and techniques employed in real-time traffic prediction, augmented by synthetic data generation using TimeGAN and other advanced models. The study focuses on analyzing and preprocessing real world traffic data, followed by generating synthetic data that closely replicates the original dataset. Through the implementation of machine learning models like Long Short-Term Memory (LSTM) networks, traffic patterns are predicted with high accuracy. Synthetic data is generated using Generative Adversarial Networks (GANs) and Synthetic Data Vault (SDV) methods, ensuring that the synthetic data retains the key statistical and structural characteristics of the original dataset. The robustness and generalization capabilities of the traffic prediction model are validated by comparing its performance on both real and synthetic datasets, highlighting the effectiveness of these approaches in real-time traffic forecasting.

• Programming Languages: Python, C++, MATLAB
• Robotics Platforms: ROS/ROS2, Gazebo, Autoware, CARLA
• AI/ML Tools: PyTorch, TensorFlow, YOLO
• Hardware Expertise: Raspberry Pi, Nvidia Jetson, Arduino
• Protocols: UART, I2C, MAVLink, MSP

• LinkedIn: Srinivas Peri
• GitHub: github.com/MAHASHANA