Samiran Gode

Hello! I am an ELLIS PhD Student working with Professor Wolfram Burgard. My research interests span a lot of fields but broadly I work on building robust models for robot perception across multiple modalities - vision, language, sonar, audio etc.

I am a graduate of Carnegie Mellon University where I was fortunate to be a part of the Robot Perception Lab. At CMU, I worked on Underwater Perception and Object SLAM. I was also a Robotics Software Engineer at a startup where I worked on the localization stack for a mobile robot system.

Before that I interned with Jupiter's Data Science team, focusing on NLP for FAQ search and user behavior analytics. Additionally, I contributed to the EqWATER project at IISc, specializing in leak detection for smart water distribution systems. I've also gained valuable experience as an Area Manager Intern at Amazon.

Email / CV / Google Scholar / Github / Linkedin

News

[Nov'24] Presenting our work FlowNav in workshops at CoRL 2024.
[Jun'24] Started my PhD as an ELLIS PhD student.
[Jan'24] SONIC accepted at ICRA 2024.
[Oct'23] Participated in Closing the Loop on Localization workshop at IROS 2023.
[Oct'23] Presented our work SONIC at the Advanced Marine Robotics Workshop at IROS 2023.
[Sep'23] Submitted one paper to ICRA'24
[Aug'23] Our paper was published at the AI Magazine!
[Feb'23] Presented our work on Understanding Political Polarization using Language Models at AAAI'23 AI4CEW
[Dec'22] Graduated from CMU!

Research

	FlowNav: Learning Efficient Navigation Policies via Conditional Flow Matching Samiran Gode, Abhijeet Nayak, Wolfram Burgard, CoRL Workshop on Learning Effective Abstractions for Planning (LEAP), 2024 CoRL Workshop on Differentiable Optimization Everywhere: Simulation, Estimation, Learning, and Control, 2024 arXiv - Used Conditional Flow Matching(CFM) for efficient goal-conditioned robot navigation as opposed to diffusion policies - Achieved the same level accuracy as the SOTA Diffusion policy for a 8x speedup.
	SONIC: Sonar Image Correspondence using Pose Supervised Learning for Imaging Sonars Samiran Gode, Akshay Hinduja, Michael Kaess, ICRA, 2024 2nd Advanced Marine Robotics TC Workshop IROS, 2023 arXiv / Code - Solved data association for underwater SLAM through a novel method for sonar image correspondence using Learned Features. - Introduced a pose-supervised network that generates feature descriptors robust to changes in viewpoints, enabling more reliable feature matches in sonar based localization and mapping.
	Understanding Political Polarization using Language Models: A dataset and method Samiran Gode, Supreeth Bare, Bhiksha Raj, Hyungon(Clay) Yoo AI Magazine The AAAI 2023 Second Workshop on AI for Credible Elections Journal / arXiv / Code - Finetuned Longformer on a scraped Wikipedia dataset to find most important tokens based on attention score. - Used other conventional techniques such as Word2Vec, Doc2Vec and BERT based models understand words which lead to polarisation. - Paper accepcted at AAAI 2023 Workshop on AI for credible elections and selected for publication at the AI Magazine Fall 2023.
	Quadric SLAM - Implemented on Object based semantic SLAM, created low-memory metric semantic maps for multi-robot communication. - Formulated a graph based SLAM. Used quadric factors with the factor graph with underlying visual inertial odometry. - Designed feature descriptors for SONAR using unsupervised learning for underwater SLAM. - Used an encoder decoder structure with CNNs with custom loss functions to learn without labels.
	Detecting and Localizing Leaks in Intermittent Water Distribution Networks Samiran Gode, Sheetal Kumar K R, Sindhu H J, P G Prasad, M S Mohan Kumar, Rajesh Sundaresan, - Developed an algorithm for detecting and localizing multiple leaks in Water Distribution Systems with Intermittent water supply for Bengaluru a city of 8.5mil.(As part of EqWATER funded by Ministry of Human Resources Development, Govern- ment of India). - Automated intermittent water supply in an experimental system(60m long*100mm dia network, 4L/s) using LabVIEW to generate scaled-down comparable data with identical disturbances and leaks analogous to field data.

Projects

	3D Dense SLAM system using ICP - Camera Localisation on the ICL-NUIM dataset using Iterative Closest Point Algorithm. - Used point-based fusion to create a point cloud map
	NERF (Volume Rendering and Neural Radiance Fields) Code - Implemented a Differentiable Renderer for emission-absorption volumes. - Implemented a ray sampler for optimising volume parameters. - Used a MLP to map 3D positions to Volume Density and colour
	Particle Filter - Monte Carlo Localization(MCL) based robot localization for an indoor robot using laser rangefinder and odometry. - Implemented the raycasting based sensor and motion model along with resampling.
	Single View to 3D Learning 3D representations using single views.

Thank you to Jon Barron for the website template!