Thursday, November 27, 2008

End course project - Finding project

End course project - Finding project


1) Monte Carlo Localization
The robot is supposed to localize itself in a known world, by having an internal geometrical representation of the surrounding.
2) Balancing robot
Deal with the problem of a two wheeled balancing robot. Extend the idea of using a PID controller to full state control. And apply a remote control for tilting the robot so its driving forward or backwards.
3) Evolution
Exchange genes for solve a task better.
4) Navigation
Make four robots work together to move from a specified position to a specified target.

Three projects in detail

We have excluded no. 3, because we didn't have a good idea for a project including the evolution theory.

1) Monte Carlo Localization
How should a map look like for easy localization?
Detailed measurement of the light sensors.
How to design a robots which is able to apply accurate light sensor measurements.
Finding out, how the communication between PC and NXT works.
PC program with visual interface showing the possible position of the NXT. At the same time the PC is doing the heavy calculation.
NXT and PC algorithm programming.

- Ordinary localization: from random samples to the robot position.
- Lost/hijacked robot problem: robot was found and set to another position. Will the samples find the new position?
- Tracking: samples already know the robots position and try to follow the robot.

needed parts:
- 1 ordinary NXT package
- 3 Light Sensors
- 1 Compass
- Landscapes

2) Balancing robot
Detailed mathematical analysing of the robot physics for the fullstate feedback control description.
Building the robot, maybe different configurations.
Try to apply the bluetooth communication between PC and NXT for the remote control.

- Balancing time of the robot while starting in an vertical position.
- Testing the remote control.

needed parts:
- 1 ordinary NXT package.

4) Navigation
Find an appropriate path model.
Determine the distances between the robots while using local sensors.
Dealing with the communication between the robots.
Try using a Kalman filter.

- The point to point navigation (while obstacle avoiding).
- with and without Kalman.

needed parts:
- 4 ordinary NXT packages.

Choosing Project:
We have chosen the project no. 1 (Monte Carlo Localization). Because the project includes many of those subjects we have been through in the course. Example navigation, behavior layers and more may come. Project 4 was excluded because we couldn't see how to measure the distance between the robots.

Schedule for next time:
- How to make a map, using plotter etc.
- Build robot.
- Color test for light sensor
- Overall schedule for project.

No comments: