Worked on the navigation stack of differential drive robots used for warehouse automation

Technologies used: Used ROS1, ROS2, Python, C++, Docker, Git

Crowd Simulation:

• Developed Gazebo plugin to dynamically control simulated humans through velocity/path commands sent via ROS topics

• Created a Gazebo-based test-bed for AMR navigation amidst a realistic human crowd simulated using the Social Force Model

Global Planning:

• Generated motion primitives to create kinematically feasible paths for free/cart-joint bots for graph search methods

• Evaluated and benchmarked lattice-based planners, namely SMAC & SBPL, to generate geometrically predictable paths

  • Optimized cost multipliers for different motion primitives to create a similar path for a given start and goal points

  • Saved an average computational time by 70% by advocating the usage of SMAC over SBPL in the existing navigation stack

Local Planning:

• Built ROS/ROS2 test-beds in Gazebo to evaluate different local planners for AMRs across various speeds and confined spaces

• Simulated precise localization with Gazebo’s data using TF broadcaster; ended global plan replanning with behavior trees

SLAM:

• Integrated Webots simulator and ROS2 to test the SLAM-toolbox on AMRs; reported a flaw in LiDAR's measurement update

Worked on pendulum actuated Spherical Bot mounted with a fisheye camera

Guides: Prof. Leena Vachhani, Prof. Abhishek Gupta

System Modeling: 

• Modeled wobbly behavior of a ball-shaped bot to obtain Lagrangian considering forward & steering motion coupling

• Obtained bot’s nonlinear control-affine system via Lagrange D’Alembert equations  adhering to non-holonomic constraints

• Simulated system response using MATLAB solver ODE15s and generated 3D visualization of bot’s motion on Simulink

• Formulated expressions for wobbling frequency and amplitude by system linearization during circular motion

Controller design: 

• Designed a non-linear controller based on input output feedback linearization to control the underactuated bot with 2 inputs

• Eliminated lateral oscillations (wobbling) during a turning maneuver while maintaining stable pendulum motion

Product Management: 

• Filed three utility patents & two design patents to secure bot’s design, vision system and scalability method

• Evaluated product market fit for existing bot via field trials; drafted technical requirements for next prototype

• Negotiated with bidder for product development by aligning on technical specifications within project’s budget

Worked with bureaucrats towards Indian Robotics Policy making and to reduce C&B costs for an Indian-based tech MNC

Policy making: 

• Benchmarked India's robotics capabilities with 10+ nations; suggested 5-year targets and means to attain them

Cost reduction: 

• Formulated FY22 savings targets across 14 business units to achieve 5% organization level C&B% reduction target

• Conducted a pilot for managers to plan savings across 12+ levers; developed macro to compile 80+ worksheets

• Instituted saving planning portal with planner-approver model for 500+ managers; drafted BRD & user stories

Awarded a Pre-Placement Offer for exemplary performance and contribution to various projects for data governance team

 
Data life cycle management: 

• Co-authored a white paper; formulated 7 stages of data lifecycle mgmt. solutions to promote PwC’s capabilities

Security: 

• Designed a methodology for the risk assessment of 15 assets of savings banks for securing 40 data attributes

• Proposed a 5 layered security strategy encompassing 40 solutions to safeguard the assets from cyber attacks