Mayuri Bhadra

The aim of the project is to develop a circuit for pixel by pixel addressing of the novel display prototypes of IGM (Institute of Large Area Microelectronics), Universität Stuttgart and allowing to show content from either internal memory or from an external source. - To develop the initial addressing system using an FPGA board and driving IC 'Hitachi HD66115T'. - Technologies: RTL Coding, High level synthesis, ISE, VHDL

Autonomous Ship Lock Mechanism: - First-ever implementation of autonomous entering of ships into the lock chamber. - Data from Blickfeld Laser Scanner and RPLIDAR are obtained and converted into ROS compatible format. - Simultaneous Localization and Mapping (SLAM) algorithm combined with the Laser Scan Matching technique is then used to correctly estimate the pose of the Ship with respect to the lock chamber. - Technologies: ROS, C++ - Tools: Blickfeld Laser Scanner, RPLIDAR, Raspberry Pi.

Recursive Sorting Algorithm: - Designed a Data sorting Algorithm using tree-based recursive algorithms to improve the existing sequential known algorithms. - Hierarchical Finite State Machines (HFSM) was used as it permits recursive algorithm to be implemented on hardware. Technologies: VHDL, Data Structures, HFSM Tools: Xilinx, Vivado

Automated scripts to improve productivity using RESTful API - Worked in the Automotive Application Lifecycle Management (ALM) department. - Hands-on experience with IBM RTC tool(Rational Team concert) - Change and configuration tool and its REST API. - Implement the template converter tool that migrates templates (JASON) between servers by automated, and user-controllable mapping via REST API. Technologies: Python, REST API Tool: RTC (Change and Configuration tool)

Perception algorithms for mobile service robots using ROS: - Implementation of Food classification using Deep Learning Techniques in Python Pytorch. - Interfacing of the Food classification algorithm with the care-o-bot using ROS in C++. Technology: C++, Python, Pytorch, ROS.

A gaze gesture detection project using Machine learning and Heuristic based approach to detect the type of gesture in applications related to autonomous driving.

Master thesis: Implementing Stochastic Computing based Recurrent Neural Network on a ZedBoard-Zynq FPGA board. -The neural network correctly predicts the handwritten digits recognition dataset of MNIST. -The concept of parallel cells is used along with the multi-core design technique and FFT for complex data computation. -Creating a bare-metal application on the CPU of the Zynq board with communication to the FPGA using AXI interface. - Technologies: VHDL, Python, C - Tools: Vivado, PyCharm