research areaseach links to the matching projects
Decentralised cooperative transport — planning and control in the payload frame, with no shared map and no central controller.
Visual SLAM written from scratch; 3D LiDAR SLAM flown on a UAV. Invariant EKF on SE(3), AMCL, pose-graph optimization.
Nav2, global and local planners, pure pursuit; whole-body QP control across a mobile base and arm.
Depth cameras, ArUco pose, homography, bird's-eye-view lane pipelines.
ACT, diffusion policy, SmolVLA, π0, X-VLA — action chunking and flow matching for manipulation.
Chassis and mechanisms, power distribution, ESP32 low-level control, Jetson, LiDAR and depth-camera integration.
publications
Soft autonomous mobile manipulators in agricultural automation – a review
Industrial Robot, Vol. 53 No. 2, pp. 376–398
Soft grippers, actuation and control, localization and motion planning for agricultural manipulators — and why integrating them, rather than any one of them, is the open problem.
research experience
Decentralised Cooperative Mobile Manipulation
Researcher · IIT Bombay
Advised by Prof. Shivaram Kalyanakrishnan, Prof. Harshad Khadilkar, Prof. Nived Chebrolu
- Cooperative transport of a rigid payload by two nonholonomic mobile manipulators, using onboard vision and decentralised compute only — no motion capture, no shared map and no force/torque sensing at the grasp.
- Planning and control are posed in the payload frame, so transport depends only on the relative pose between the robots; global localisation reduces to coarse landmark-based estimation.
- Relative pose is recovered through the kinematic chain of the shared grasp rather than odometry, making it immune to wheel slip and map drift; a right-invariant EKF on SE(3) estimates each grasp transform with χ² innovation gating at 50 Hz from 30 Hz wrist-camera detections.
- A 14-DoF whole-body QP carries the closed-chain constraint as a soft cost weighted by the live estimator covariance, so coupling relaxes as the estimate degrades and the solver stays feasible mid-transport. Full stack running on hardware.
e-Yantra Robotics Competition — Theme Development
Senior Project Technical Assistant · Embedded Real-Time Systems (ERTS) Lab, e-Yantra, IIT Bombay
Advised by Prof. Kavi Arya, Prof. Shivaram Kalyanakrishnan
- Research and development of Niti Vahan, the eYRC 2026-27 theme — an autonomous car operating under city road rules, selected by 1500 teams nationally. Contributed the electronics and the perception and control algorithms for a scratch-built miniature Ackermann vehicle.
- Monocular pipeline on a Raspberry Pi with a single USB camera: lane detection via bird's-eye-view warping, colour segmentation and polynomial fitting under PID steering control, with obstacle detection, lane switching and ArUco-based docking for parking. Teams build against a matched CoppeliaSim digital twin in Stage 1 before moving to physical hardware in Stage 2.
- Developed the Krishi CoBot theme (eYRC 2025-26) for 649 teams and 2436 participants, delivering 475 hours of remote access to a UR5 arm and a mobile robot on a mock greenhouse — part of ongoing research on scalable, low-cost remote laboratories for robotics education, with a paper in preparation.
- Mentored three eYSIP summer research projects and five interns in 2026, and delivered faculty and outreach workshops reaching 120+ participants.
Visual SLAM for UAVs
Research Intern · eYSIP 2025 · e-Yantra, ERTS Lab, IIT Bombay
- Built a complete visual SLAM front end and back end from scratch using only Open3D and OpenCV, after judging ORB-SLAM3 and RTAB-Map impractical to adapt and finding no end-to-end 3D VSLAM plus navigation stack comparable to Nav2.
- Implemented ICP registration, keyframe selection and incremental map building, producing consistent maps of indoor environments; added local and global loop closure with pose-graph optimisation, and relocalisation using a bag-of-words place recognition front end refined by ICP.
- Validated handheld on a RealSense D455 with an Orin NX; designed an extension to tightly coupled VIO by adding IMU preintegration factors to the pose graph.
Soft Autonomous Mobile Manipulators in Agriculture: A Review
Undergraduate Researcher · NITK Surathkal
Advised by Prof. Mervin Joe Thomas
- Systematic review of soft mobile manipulators for agricultural automation, published in Industrial Robot (Vol. 53 No. 2, 2026).
- Compared soft gripping and actuation technologies across actuator families, assessing compliance, payload and crop damage trade-offs for harvesting; reviewed classical and state-of-the-art localisation and path planning for mobile bases in unstructured fields.
- Analysed the mobile manipulator design space, identifying where gripper compliance relaxes precision requirements on the base and arm.
Autonomous Ackermann Platform for Field Exploration
Research Intern · eYSIP 2024 · Autonomous Robots and Multi-Systems (ARMS) Lab, IIT Bombay
Advised by Prof. Arpita Sinha
- Built a ROS 2-controlled 4WD Ackermann vehicle for field exploration, from low-level motor and steering control through to autonomous navigation on the physical platform.
- Fused IMU with Vicon velocity estimates through a Kalman filter for odometry, ran LiDAR SLAM with AMCL localisation in Nav2, and implemented a pure-pursuit planner for waypoint navigation.