Project goal
Design and demonstrate an autonomous robot that searches an arena, detects coloured objects, retrieves them, returns to the start area, and places each object into a matching coloured bin.
Project portfolio
Goals, contributions, evidence.
A selection of robotics, simulation, software, and engineering projects demonstrating my technical development across autonomous systems, mechanical design, numerical modelling, and applied programming. Each project includes the problem context, project goal, my specific contribution, tools used, outcomes achieved, and supporting evidence such as reports, repositories, images, videos, or demonstrations.
Robotics systems design
VISOR: Visual Intelligent System for Object detection and Retrieval
Team project using a Leo Rover 1.8 platform, low-mounted LiDAR, RGB-D camera, myCobot 280 manipulator, NUC, ROS 2 Jazzy, Gazebo, Nav2, and object-detection logic.
My contribution
- Led SLAM, navigation, and exploration development for the Leo Rover.
- Integrated Nav2 behaviour with map, TF, goal-pose logic, and camera-informed navigation.
- Developed and debugged exploration scripts using occupancy grids and Nav2 action clients.
- Identified system-integration issue caused by conflicting manipulator and rover topic names.
- Prepared demonstration evidence and explained the limits of full mission integration.
Evidence links
Gazebo simulation
Online RViz map generation
Real time object detection
Mechanical Design
Leo Rover Payload Sled Design
A modular mechanical payload architecture designed to integrate sensing, computation, and manipulation hardware onto the Leo Rover platform.
Full CAD assembly showing the payload sled mounted on the Leo Rover.
Project goal
The goal was to design a structurally sound, modular, and reversible payload sled for the Leo Rover. The sled needed to support the LiDAR, RGB-D camera, manipulator arm, Intel NUC, and cable routing while preserving sensor fields of view, access to the rover, and overall system stability.
My contribution
I completed the full mechanical design and analysis of the payload sled. This included the four-plate architecture, component placement strategy, LiDAR adapter integration, NUC holder, camera/manipulator mounting arrangement, connector design, plate thickness selection, manufacturability justification, and structural validation using finite element analysis.
Design approach
- Developed a four-plate architecture: front bottom, rear bottom, front top, and rear top plates.
- Placed the LiDAR low at the front to improve obstacle and bin detection.
- Mounted the manipulator and RGB–D camera on the front top plate for direct front-side pickup.
- Moved the Intel NUC to the rear top plate to improve packaging, balance, and LiDAR visibility.
- Designed custom I-beam connectors, adapters, and mounting interfaces for 3D printing.
Analysis and validation
- Selected a 5 mm plate thickness based on fastener compatibility, print robustness, and stiffness requirements.
- Used conservative loading assumptions for the manipulator and NUC to provide design margin.
- Performed static structural analysis to check stress distribution, displacement, stiffness, and safety factor.
- Verified that the design supported the mounted payload without excessive deformation.
Outcome
The final payload sled provided a modular and manufacturable mounting architecture for the complete autonomous mobile manipulation system. It separated the front sensing/manipulation module from the rear compute module, improved access and maintainability, reduced sensor obstruction, and provided a structurally validated platform for final integration.
Engineering simulation
Simulation-led engineering evidence
These projects show transferable competencies in modelling, numerical methods, experimental reasoning, and engineering communication.
Monte Carlo model for grain boundary movement
Goal: Predict grain boundary movement in steel during annealing using a Monte-Carlo Potts model.
My contribution: Developed a C++ simulation model using crystallographic texture and Kernel Average Misorientation as inputs, supporting a 25% reduction in annealing time and an estimated impact of over 12 crores.
CFD & MHD electromagnetic stirring model
Goal: Model electromagnetic stirring behaviour to disrupt dendrite formation during steel casting.
My contribution: Built a multiphysics simulation workflow combining CFD and electromagnetic analysis to support process understanding and new product development decisions.
Parallel Canny Edge Detector
Goal: Implement and accelerate image-processing computation using parallel programming.
My contribution: Developed a C++/OpenMP implementation and achieved significant speedup, demonstrating algorithmic optimisation and parallel computing ability.
Moldboard retention screw system
Goal: Replace a nut-bolt assembly with a cam-based retention mechanism to reduce service and assembly time.
My contribution: Modelled the cam profile, achieved 176 kN preload, incorporated non-linear Belleville spring behaviour, and supported a more serviceable product concept.