Robotics & Simulation Engineer

Hi, I'm Bharath Desikan

Robotics researcher specializing in information theory and spatial field reconstruction. I build robust simulation infrastructure using Linux, Docker, and the PX4 stack for autonomous systems research.

Expert in ROS2, Gazebo, containerized deployments, and Gaussian Process methods.

About Me

I'm a robotics researcher and simulation engineer focused on building autonomous systems that can intelligently sample and reconstruct spatial fields. My work bridges theory and practice, from Gaussian Process methods to production-ready containerized deployments.

Information Theory

Applying information-theoretic principles to optimize sampling strategies and sensor placement for environmental monitoring.

Gaussian Processes

Working with stationary, non-stationary, and uncertain-input GP methods for robust spatial field reconstruction.

Simulation & PX4

Building high-fidelity robotics simulations with PX4 SITL, Gazebo, and ROS2 for autonomous vehicle development.

Linux & Containers

Developing reproducible research infrastructure with Linux systems, Docker containerization, and automated deployments.

Technical Skills

Robotics

ROS2 Gazebo PX4

ML / GP

GPyTorch PyTorch NumPy Scikit-learn

Systems

Linux Docker Git Bash CI/CD

Languages

Python Bash YAML

Publications

ICRA 2025 Robots in the Wild Workshop

Low Cost ASV for High-Resolution Spatio-Temporal Aquatic Field Reconstruction via Dynamic Kernels

Rodney Staggers Jr*, Bharath Vedantha Desikan*, Jnaneshwar Das (*contributed equally)

We present a low-cost ($2,500) Autonomous Surface Vehicle (ASV), R/V Karin Valentine, for high-resolution aquatic field reconstruction. Our approach uses Gaussian Process regression with non-stationary Matérn kernels that adapt spatially to capture varying environmental dynamics across water bodies.

Code

Projects

Research tools and simulation infrastructure for autonomous spatial sampling.

Field Sampler

Autonomous Spatial Sampling

ROS2 Jazzy GPyTorch Gazebo Docker

Aquatic Mapping System

Complete simulation infrastructure for autonomous spatial field reconstruction using USVs. Features PX4 SITL integration, GP-based reconstruction with multiple methods (Standard GP, McHutchon NIGP, Girard), and a web-based control panel for running batch simulations.

  • Docker containerized simulation trials
  • Real-time GPU reconstruction with CUDA
  • Web dashboard for monitoring and control
  • Comparison of GP methods (RMSE/NRMSE)

Get in Touch

Interested in collaboration or have questions about my research? Feel free to reach out.

ICRA 2025 - Low Cost ASV for Aquatic Field Reconstruction