RESEARCH & PROJECTS
Current and selected research
Current investigations and developed systems in robotic perception, trustworthy AI, and secure research infrastructure.
Risk-aware perception and navigation
Research study
Uncertainty-Aware Visual SLAM
A visual-SLAM project using TurtleBot 4, OAK-D, RPLidar, ROS 2 Jazzy, and SLAM Toolbox. The study examines illumination, image blur, speed, surface disturbance, jerk, feature quality, inlier ratio, and tracking risk.
The labeled dataset contains
28,133 samples
, with
ROC-AUC 0.8813
and
PR-AUC 0.9038
. The study evaluates adaptive speed, exposure, and gain; closed-loop control effectiveness remains to be established.
Privacy-preserving robot mapping
Active research benchmark
PrivateMap-Bench
A privacy–utility benchmark for sharing indoor SLAM maps. It examines floor-plan, room-graph, trajectory, and routine leakage from robot maps and trajectories while measuring navigation utility.
The benchmark is under development, with a focus on quantifying privacy risks alongside navigation performance.
Explainable surgical-video intelligence
Ongoing Ph.D. research direction
Surgical Skill Assessment
Exploring explainable AI for arthroscopic surgical-video skill assessment. The proposed “Agentic Evidence Tribunal” uses AI agents to locate, challenge, and verify video evidence before assigning a rubric-based skill score.
This ongoing research investigates how verifiable video evidence can support interpretable surgical skill assessment.
Autonomous indoor guidance
Developed system
SMART-TourBot
An autonomous indoor guide robot integrating ROS, SLAM, semantic navigation, A* path planning, sensor fusion, reactive obstacle avoidance, finite-state execution, and human–robot interaction.
The project brings localization, planning, and interaction together for indoor guide-robot operation.
Applied AI and secure research infrastructure
Earlier research · 2019–2023
AI assistants & cryptographic validation at IISc
Research at the Indian Institute of Science included a virtual physics laboratory and AI assistant using BERT, AllenNLP, and IBM Watson, alongside the Automatic Cryptographic Algorithm Validation System (ACAVS).
Contributions included DSA, ECDSA, and RSA validation and research infrastructure built with Java, MySQL, Linux, and Git/Bitbucket. Demonstrations and resources are available in the earlier projects below.
