Autonomous 3D Mapping of unknown structures Senior Thesis

Guide: Prof. Jerome Le Ny (Polytechnique Montreal)

Co-Guide: Prof. Bharath Bhikkaji (IIT Madras)

Abstract: We developed a system capable of autonomously mapping the visible part of a bounded three-dimensional structure using a mobile ground robot equipped with a depth sensor. We propose motion planning strategies to determine appropriate successive viewpoints and attempt to fill holes automatically in a point cloud produced by the sensing and perception layer. The emphasis is on accurately reconstructing a 3D model of a structure of finite size rather than mapping large open environments, with applications for example in architecture, construction and inspection. The proposed algorithms do not require any initialization in the form of a mesh model or a bounding box.

FIRA RoboSoccer SimuroSot league

Team Saahas, Center for Innovation, IIT Madras

• Developed an AI agent capable of playing a simulated 5vs5 RoboSoccer game.
• Adopted a two level Strategy-Role framework for the decision making and implemented real-time path planning algorithms.
• Implemented a Case Based Retrieval system for action selection for the defender agents as a course project in Memory Based Reasoning in AI under Prof. Sutanu Chakraborti (IIT Madras)

Simualted RoboSoccer Game with 1 Attacker, 2 Defenders and 1 Goalie.

The project report for the design of the CBR system can be found here.

Controlling LEGO EV3 robots via ROS a testbed for coordinated control

Guide: Prof. Bharath Bhikkaji (IIT Madras)

• Developed a ROS package for controlling LEGO EV3 robots using the ROS Navigation stack; thereby creating a low-cost testing platform for robotics research.
• Developed an indoor localization system using multiple colored patterns and an overhead camera to correct for the drift in dead reckoning.
• The code has been open sourced and made available via GitHub here

The LEGO EV3 robots with coloured patterns used for indoor localization

The project report can be found here.

Control Lab using LEGO Mindstorms kits for sernior undergraduate students

Guide: Prof. Bharath Bhikkaji (IIT Madras)

• Designed 3 benchmark control experiments
  1. PID Controller for a DC Motor
  2. Inverted Pendulum
  3. Feedback control for a Pendulum on a cart system
  using LEGO Mindstorms kits and Simulink.
• Recorded video lectures for demonstrating and explaining the theory behind the experiments which has been made publicly available to the students.

Wrist-Flexer as part of National Service Scheme (NSS)

Guide: Prof. Manivannan (IIT Madras)

Overview

Children with cerebral palsy have limited wrist-movement capability due to dominance of flexor muscles over extensor muscles in their forearm. The objective of this project is to make an exercise equipment which would help relax the flexor muscles from the early stages in order to avoid permanent setting of wrist position.

To achieve this, we came up with a low cost design which can be made from things generally available at home. We used two bands - one worn around the palm and the other near the elbow - which are connected by means of an elastic band. The elastic band keeps the flexor muscles in the extended position by default and the user has to exert force opposing to the band to move his/her wrist. This provides a good way to exercise the forearm muscles. This device can be adjusted to different age groups but are primarily meant children below 10 yrs of age.

Prototype II: Wrist-Flexer, under National Service Scheme (NSS) 2011

The project report can be found here.

---

Smart Navigation

Summer Project under Center for Innovation, IIT Madras

Overview

The objective of the project is to develop a 'Smart Navigation' system for mobile robots which is independant of external directional-aids as opposed to Line Followers. A learning-based system is proposed here; where the mobile robot maps its locomotion (using Rotary Encoders) while being remotely controlled by a human from a hand-held device. During this process, the user can save required places as checkpoints (nodes). On sub-sequent use, the user can summon the robot to any saved checkpoint remotely from the controller. As the agent starts with a blank map, it remembers the path details using a unique, memory-efficient data structure which allows us to use our vairant of dijkstra's algorithm to find the optimal path to the target checkpoint. Furthermore, S-Nav has been primarily constructed for a differential drive system.

During testing of Prototype 1, made for CFI Summer Project 2012

A project documentation can be found here.

Understanding the Propagation of Electromagnetic Waves through General Coaxial Structures using Numerical Simulations

John F Welch Technology Center (JFWTC) GE, Bangalore

Guide: Dr. Vikram Melapudi

A patent application based on this work has been filed by GE GRC, Bangalore.

Abstract

Effective propagation of electromagnetic (EM) waves has a wide variety of applications. Here we study the practical requirements necessary for maximizing the bandwidth of frequencies in general Co-axial waveguide structures and also maintain mode purity. In this regard, well known antenna array design is verified using Numerical Simulations. Further these antenna designs were optimized for efficient coupling. Signal processing methods were also explored to enhance the SNR (signal-to-noise-ratio) and the utility of such EM waveguide systems. Finally, the effect of change in the dimensions of the coaxial waveguide on the transfer of energy through the waveguide is also analyzed using numerical simulations.

Summary of work done

• Analysed Microwave based Pipeline Inspection methods for Coaxial Waveguide systems using Numerical Simulations.
• Developed 3D models of Co-axial waveguides in COMSOL Multiphysics.
• Verified the effectiveness of Antenna Array design which was proposed by Robin E. Jones.
• Analysed the effect of radius and length of probe on the coupling and optimized them for maximum coupling (i.e) minimum impedance mis-match.

Apart from this, the internship also included development of a Simulation Framework for performing 2D Magnetostatic simulations using Ansys Parametric Design Language (APDL).

The redacted internship report can be found here.