Middle East Technical University - Department of Computer Engineering
Home
Schedule
Assignments
Readings
Projects
Software
Resources
 
saranli@cs
Uluç Saranlı
  

CENG786 - Robot Motion Control and Planning
Homework #2 - Potential Fields

  • Out: Nov 20, 2015
  • Due: Dec 1, 2015

For this homework, you are required to compare repulsive+attractive potential fields with navigation functions for a planar point robot in a 2D sphere world where both the obstacles and the boundary are circles.

As described in the assignments section of this web site, your submission should be in the form of a web page which nicely describes your work. Please do not collaborate with your classmates except exchanging ideas and other inspirational materials.

In summary, you should do the following for this homework

  1. Implement a simple potential field planner with attractive and repulsive fields for the 2D sphere world with circular obstacles and boundary. The number, locations and sizes of obstacles should be configurable. Note that unlike the first homework, your planner has access to a global map for this homework.
  2. Implement a navigation function based planner for the same domain, once again with configurable obstacles.
  3. Construct a number of example settings with different number of obstacles and apply both algorithms to these domains. Compare the two implementation with respect to their performance, speed of convergence (i.e. how "long" does it take to get to the goal), difficulty of implementation and/or other issues you think are relevant.
Bonus problem: Implement both algorithms above in an environment with polygonal star-obstacles by constructing a diffeomorphism to a topologically equivalent sphere world and following the gradient. Illustrate your implementation with a few examples.

Note that using Matlab for visualization will probably a good idea. Another important issue is how you compute the path that results from following the negative gradient. I recommend that you use a continuous model and the ode45 function of Matlab to obtain your result accurately through numerical integration. However, you are allowed to use a discrete approximation with a properly chosen step size as well. Make sure to discuss related issues and potential problems in your report.

Good luck!

[ Home | Schedule | Assignments | Readings | Projects | Software | Resources ]

saranli@cs, Uluç Saranlı
The overall design of this web page was inspired by similar pages by Frank Pfenning.