COMP 790-058 course project: Strategies for maintaining visibility of a moving target.

Strategies for maintaining visibility of a moving target.

COMP 790-058 Course Project.Georgi Tsankov

Date: 9/19/2007

TOPIC

Given a moving target in environment with obstacles, try to find a strategy for the observer, that maintains visibility of the target for as long as possible. The target trajectory may predictable or partially predictable. There may be other goals, for example: to minimize the total distance traveled or to keep the observer invisible to the target (stealth tracking).

MOTIVATION

There are various applications that require monitoring of a moving target:

surveillance cameras.
monitoring a robot, that is involved in automated process.
moving a camera around a surgery site.
following any object of interest in virtual environments.

There are different constraints - usually the obstacles obstruct both movement and visibility. The distance of sight is normally restricted, as well as the speed of the observer. Other kinds of limitations can be imposed as well.

PRIOR WORK

Several papers are related to this project:

Steven M. LaValle, He'ctor H. Gonza'lez-Ban~os, Craig Becker, Jean-Claude Latombe: Motion Strategies for Maintaining Visibility of a Moving Target.
Steven M. LaValle, David Lin, Leonidas J. Guibas, Jean-Claude Latombe, Rajeev Motwani: Finding an Unpredictable Target in a Workspace with Obstacles.
Tirthankar Bandyopadhyay, Yuanping Li, Marcelo H. Ang Jr., and David Hsu: Stealth Tracking of an Unpredictable Target among Obstacles.
Tsai-Yen Li, Tzong-Hann Yu: Planning Tracking Motions for an Intelligent Virtual Camera.

GOALS and SUBTASKS

The goal is to implement and extend the algorithm in [1] for following a target with unpredictable motion in a known 2D environment. There are different possible extensions to it:

Taking dynamics into account and trying to minimize the energy (not the same as minimizing the distance).
Analyzing the prediction schemes and proposing a better one.
Adding moving obstacles.
Other goals, constraints.
Multiple targets and/or multiple observers (might be too complex).

The specific goals are not clearly defined yet. I will try to do the algorithm for unpredictable target + (1), (2) and will look into other papers for more possible improvements.

Update on Oct 31

I have started working on the framework and it should be ready soon. It will show a moving robot (without motion planning yet) and its visibility region, also some user interface to specify the motion for the target robot and possibly UI to create and edit a maze (this can be skipped probably, since it is a tedious work and not of primary interest). In the beginning the maze will be read from a file.
Once this is done, I'll concentrate on the motion planing algorithm itself.