Jagadeswara Pavan Kumar Varma Pothuri

Aspiring Robotics Engineer with a passion for autonomous systems, SLAM, path planning, and computer vision. I’ve worked on real-world projects involving UAV systems and cutting-edge technology like online visual tracking and real-time sensor fusion. With experience in Python, C++, and ROS, I enjoy tackling complex challenges in robotics, from multi-robot coordination to advanced tracking solutions. I’m excited to bring my skills, creativity, and fresh perspective to help your team innovate and succeed.

Scalable Coverage Path Planning (SCoPP) for Multi-UAV Coordination

Developed the Scalable Coverage Path Planning (SCoPP) framework for coordinating multiple UAVs in large-scale disaster response and environmental monitoring tasks. I integrated Multi-Robot Task Allocation (MRTA) algorithms into custom-built drones and Parrot Anafi AI and Anafi Thermal UAVs, using Olympe and Pymavlink for control. The system was tested in simulation and real-world environments, including the SOAR facility at the University at Buffalo, where it successfully managed coverage of non-convex areas for disaster relief.

Autonomous UAV Tracking and Pursuit using Reinforcement Learning

Designed a robust UAV tracking system utilizing YOLOv8 for high-precision object detection combined with reinforcement learning for optimal navigation and control. The system integrates state space representation to forecast the target’s future states, improving the tracking of dynamic aerial objects. The reward-based control system was trained and tested in AirSim simulation environments, achieving enhanced UAV tracking accuracy, making it suitable for applications like aerial surveillance.

Concurrent Design of Unmanned Aerial Vehicles for Real-Time Drone Tracking

This project focused on optimizing UAV tracking systems by integrating YOLOv2 for object detection with MOSSE correlation filtering to enhance tracking accuracy and responsiveness. Using depth cameras, the system was able to maintain precise positioning relative to its target. I employed MATLAB and simulation tools for system modeling, applying concurrent design principles for both software and hardware optimization. This research has potential real-world applications in surveillance, search and rescue, and geographic mapping.

Imitation Learning for Autonomous UAV Navigation and Last Mile Delivery

Developed an advanced navigation framework for UAVs using inverse reinforcement learning (IRL) to optimize routes for last-mile delivery. The system improves real-time decision-making and dynamic adaptability in urban environments, enhancing delivery accuracy and safety. By employing machine learning techniques, the UAV was able to autonomously adjust its behavior based on its environment, making it highly efficient for autonomous deliveries.

Monocular Vision-Based Object Detection and Depth Estimation for Autonomous Vehicles

Developed a system using YOLOv5 for object detection and depth estimation in monocular vision settings. This project focused on improving the perception abilities of autonomous vehicles in dynamic environments. The system enhanced the detection accuracy and depth estimation, enabling the vehicle to navigate more effectively in real-world scenarios.

Stereo Visual SLAM for Autonomous Robots

Implemented a Stereo Visual SLAM system using ORB-SLAM2 for enhanced localization and mapping in unknown environments. This system was deployed on TurtleBot platforms and custom drones, significantly improving autonomous navigation through robust mapping and environmental understanding.

Advanced Path Planning Algorithms: A*, Dijkstra, RRT, RRT* 

Implemented a variety of path planning algorithms, including A*, Dijkstra, RRT, and RRT* for efficient autonomous navigation. These algorithms were integrated into custom-built UAVs and ground robots to optimize their route planning and ensure collision-free navigation in complex and dynamic environments. The algorithms were tested across multiple scenarios, proving effective for both indoor and outdoor robotic applications.

• An Open-Source Hardware/Software Architecture and Supporting Simulation Environment to Perform Human FPV Flight Demonstrations for Unmanned Aerial Vehicle Autonomy Published in: AIAA AVIATION FORUM AND ASCEND 2024 DOI: 10.2514/6.2024-4458
• A Physical–Digital Twins Environment for Real Outdoor Testing of Multi-Unmanned Aerial Vehicle Coordination and Applications Submitted to: ICRA 2025
• Scalable and Load-Balanced Coverage Path Planning for Multi-Robot Teams Surveying Non-Convex Areas Yet to be Submitted
• Autonomous UAV Tracking and Pursuit using Reinforcement Learning Yet to be Submitted
  

• Convolutional Neural Networks
  Offered by DeepLearning.AI on Coursera, this course teaches how to build and train CNNs for image recognition and other tasks.
  Completed: August 21, 2020
  
• Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
  Offered by DeepLearning.AI on Coursera, this course focuses on optimizing deep learning models for better performance.
  Completed: August 10, 2020
  
• Neural Networks and Deep Learning
  Offered by DeepLearning.AI on Coursera, this course introduces fundamental concepts of neural networks and how they work.
  Completed: August 6, 2020
  
• Structuring Machine Learning Projects
  Offered by DeepLearning.AI on Coursera, this course provides strategies for managing and structuring machine learning projects effectively.
  Completed: August 12, 2020
  
• Introduction to Deep Learning
  Offered by HSE University on Coursera, this course covers key concepts of deep learning, including various algorithms and models.
  Completed: July 8, 2020
  
• Natural Language Processing with Classification and Vector Spaces
  Offered by DeepLearning.AI on Coursera, this course teaches NLP techniques using classifiers and word embeddings.
  Completed: July 20, 2020
  
• Natural Language Processing with Probabilistic Models
  Offered by DeepLearning.AI on Coursera, this course explores probabilistic models like Hidden Markov Models for NLP tasks.
  Completed: August 1, 2020
  
• Natural Language Processing with Sequence Models
  Offered by DeepLearning.AI on Coursera, this course focuses on using RNNs and LSTMs for tasks like sentiment analysis and text generation.
  Completed: August 4, 2020