Undergrad Research Project - Navigating the Ballbot in extremely crowded environment

Fall 2015

Student
Sicong Wang
Advisor
George Kantor
Project description

Introduction

Ballbot is a tall and slim self-balancing robot, designed to be able to easily navigate in crowded environment. Previous researches have demonstrated the ability to efficiently generate kinematically and dynamically feasible trajectories in milliseconds. However, this only works if there exists an obstacle-free path to the destination, which is often not the case in extremely crowded environment full of people. Therefore, it is desired that the robot can gently “push” its way through the crowd to reach the destination.

Objectives

The objective of this project is to enable the Ballbot to navigate through environment so crowded that no obstacle-free path is available, by interacting with the obstacle to “push” a way out. I will be assisting PhD candidate Michael Shomin who is currently pursuing a research in this direction, and in the meantime will also conduct independent research on specific aspects of the problem being investigated.

Methods

  1. Complete background knowledge training. Should get familiar with: (1) the ROS navigation stack; (2) the Ballbot software stack; (3) the Ballbot dynamics; (4) trajectory planning.
  2. Experiment with different methods to update the cost map based on interactions with the obstacle.
  3. Come up with a method to analyze the performance and characteristics of each method used for updating the cost map.
  4. Help Michael with his experiments.

Anticipated results

The expected outcome of this project is an optimal method to update the layered cost map in the ROS navigation stack such that the robot behaves in a desired way in a packed environment with no obstacle-free path. The definitions of optimality and desired behavior are also questions to be investigated.

Return to project list