Robotics - Perception | Autonomous Driving C++ Software Developer

Aravind Krishnan • Peerlist

* Autonomous mobile robot to relocate household objects of specific interest, inside a house.
* Stereo SLAM for localization and navigation on ROS
* 9Axis IMU Odometry to assist SLAM, processed on Arduino Mega
* NVIDIA Jetson TX1 as the main controller board
* 3DOF lightweight manipulator controlled by servos to grip the object (eg. coffee cup)
* Encoded dc motors connected to treads for robot locomotion
* Ultrasonic sensors for Deep Q-learning based Object Detection and Recognition and simple Motion Planner


Autonomous Mobile Robot for Object Relocation inside a House

Autonomous navigation of an RC car in an unstructured environment (trails in remote locations) to assist specifically in fetching/carrying water from wells to remote villages. This is a concept that we want to participate with in the Autonomous Robot Challenge held by ARM.
Perception | YOLO for Object Detection | State Estimation | Planning | Control, all under ROS framework and Microsoft Airsim with Unreal Engine. NVIDIA TX2, ZED Camera, VESC controller, Razor IMU on an RC Car platform. Also using the developments from NVIDIA Redtail project.

Autonomous RC car navigation in an unstructured environment

* Explored the benchmark algorithms for online Traffic sign Classification and Recognition
* Analysed ConvNet structure's impact on Traffic Sign Classification and Recognition of GTSRB & LISA dataset
* Empirically evaluated self-developed CNN models on various hardware configurations for real-time output

Empirical Evaluation of Convolutional Neural Networks in Cla

Camera Calibration
Camera -- LiDAR cross calibration
Visualisation of camera data in point cloud

LiDAR - Camera Cross Calibration and Sensor Fusion in ROS

* Worked on STM34Fxx and uC-OS II (RTOS) to develop MMI to control throttle & monitor Li-ion Battery stack systems through CAN communications
* Coding in C. Kiel IDE, CANalyzer, Trace32 software were used

Man Machine Interface (MMI) for an Electric Glider

* Explored path planning algorithms like A*, LPA*, D*lite for a grid robot that should navigate in an unknown environment
* Path planning for a rectangular grid robot of 3x1 size to navigate the environment with modified A*
* Python language for programming and used Pygame for simulation

Path planning algorithms for a grid robot in a Dynamic envir

* Data acquisition from multiple sensors (pressure, loadcell, GPS, motion...) controlled by TI's MSP430
* Data pushed to light server on Raspberry Pi (Server) and timely update on Robert Bosch's main server (Client)