Lomash Relia

M.Tech student and robotics researcher specializing in autonomous systems, space robotics, computer vision, deep learning, and edge AI.

My research interests lie at the intersection of autonomous systems, robotics, computer vision, and edge AI, with a focus on bridging simulation-based development and real-world deployment.
My work emphasizes practical system-level analysis, resource-constrained inference, and deployable perception–control pipelines.

Depth-Aware Rover: Edge AI and Monocular Vision for Real-World Implementation

Springer | IEEE INGARSS 2025

This research presents a depth-aware autonomous rover system designed under real-world hardware and thermal constraints.

Summary:

  • Developed a low-cost rover platform using monocular vision and edge AI on Raspberry Pi
  • Compared stereo-based depth perception in simulation with monocular depth estimation in real-world deployment
  • Integrated object detection and depth estimation for navigation under constrained compute and power budgets

Key Contributions:

  • Demonstrated system-level trade-offs between depth accuracy, inference latency, and thermal stability
  • Highlighted challenges in simulation-to-real transfer for depth-based navigation
  • Provided insights into edge deployment feasibility for autonomous exploration platforms

Publications & Presentations:

  • Springer Book Chapter (July 2024)
  • IEEE INGARSS 2025 Symposium (Oral / Poster)

đź”— Springer: https://link.springer.com/chapter/10.1007/978-981-96-3420-0_17

Deep Learning Models for Tropical Cyclone Intensity Estimation

WGNE Blue Book | IITM Collaboration

This work focuses on applying deep learning techniques to multi-channel satellite imagery for estimating tropical cyclone intensity.

Summary:

  • Designed a 4-channel image-to-intensity regression pipeline using deep convolutional networks
  • Evaluated rotation-blended CNN architectures from existing literature
  • Benchmarked experimental results against published performance metrics

Key Contributions:

  • Systematic replication and validation of prior deep learning approaches
  • Analysis of model robustness on complex meteorological data
  • Emphasis on reproducibility and scientific benchmarking

Publication:

  • WGNE Blue Book (November 2024)

đź”— WGNE Paper:
https://wgne.net/bluebook/uploads/2024/docs/010_Relia,_Lomash_Evaluating_Deep_Learning_Models_for_Cyclone_Intensity_Estimation.pdf

Master’s Thesis Project (Ongoing)

ABV-IIITM Gwalior

As part of my M.Tech program, I am currently working on a thesis project focused on robotic manipulation and autonomous control.

Research Focus:

  • Mathematical control systems for robotic manipulators using MuJoCo-based simulation
  • Eye-in-hand visual servoing and perception-driven control strategies
  • Exploration of Vision-Language-Action (VLA) models for robotic manipulation in space and planetary environments

Objectives:

  • Develop robust control pipelines that integrate perception and action
  • Study the applicability of foundation models to embodied robotic systems
  • Improve reliability and generalization in simulated and real-world manipulation tasks

Research Interests

  • Autonomous and space robotics
  • Vision-based control and visual servoing
  • Edge AI and embedded perception systems
  • Simulation-to-real transfer
  • Embodied AI and physical AI systems

Academic & Conference Activity

  • IEEE INGARSS 2025 Symposium
  • PRL Venus Science Conference (2024)
  • ICHCSC Conference (2024)

Notes

  • Detailed experimental results, code repositories, and supplementary material are available upon request.
  • Ongoing research updates will be added as work progresses.