Risk sensitive
particle filtering for localization and augmented MDP for planning on Lego
robot
I am using a
IntelliBrain board with
RoboJDE on this project.
more details coming soon!!
Monte Carlo Localization using RCX and LejOS
This program was used and presented at the following workshop:
"Particle Filtering using the RCX", Thinking Outside the
(Desktop) Box, National Science Foundation Workshop, University of
Mississippi, November 19-21, 2004.
Here is simple map created with five toolboxes:
D* on a Lego Robot with handyboard
I used just a single two-dimensional integer array for storing the map, Expanded nodes and costs to fit everything in the memory.
|
15 |
6-14 |
5 |
4 |
3 |
2 |
1 |
0 |
|
Node Expanded |
Cost for
getting to this point |
Parent node |
Obstacle |
||||
 | Bits 0-2 are the obstacle bits. Each time the sonar sees a obstacle in a grid point it will increase the value of obstacle by one. |
| Bits 3-5 are parent node. The value of these bits
shows how we got to the current grid point. For example if we get to point x,y from x-1,y-1 the part value is 1 |
|
3 |
4 |
5 |
|
2 |
(x,y) |
6 |
|
1 |
0 |
7 |
| Bit 6-14 is accumulative cost of getting to point x,y. Cost of going one grid horizontally or vertically is 3 and cost of going from x,y to x±1, y±1 is 4. |
| The heuristic is a direct line from start to goal. |
| Set bit 15 when the node is expanded |
