Project

General

Profile

Statistics
| Revision:

root / trunk / code / projects / mapping / drive / multi_robot_behavior.txt @ 1390

History | View | Annotate | Download (1.82 KB)

1 916 dsschult
/**
2
3
Optimal Behavior for Mapping
4
5
Desired Properties
6
- Overlapping maps
7
- Shared origin
8
 * robots must know where they are in relation to each other
9
10
11
12
Actual Behavior
13
- make BOM contact with another robot
14
 * estimated distance reading would help
15
- try to follow another robot
16
 * could parallel path but moved over slightly
17
   to give a wider view while also corroborating data
18
 * then treat as leader/follower in actions
19
 * get groups to link up with each other and then separate to
20
   establish meeting point between maps
21
22
23
24
25
26
Thoughts on improving overall efficiency
27
----------------------------------------
28
Robot:
29
- use a shorter polling interval for rangefinders,
30
  then average readings between wireless polling intervals
31
- if the rangefinder value = MAX, then mark it as -1
32
  so the program doesn't draw that point
33
- try bursts of data every second or farther apart
34
  containing multiple data points
35
 * this reduces stress on wireless timing and allows for
36
   overhead created by token ring
37
- one solution to the 'master origin' problem for multiple robots
38
  is to start each robot at the same point with the same orientation
39
 * this isn't very useful and takes effort, but it works
40
 * could also establish an initial offset in robots,
41
   but accurate initial placement would still be an issue
42
43
Server:
44
- if the rangefinder value = MAX or -1, don't use that point
45
  or set it to infinity
46
- if robots have linked as leader/follower pair, use
47
  one point of origin for both robots
48
- when two or more robots meet, set one robot as the
49
  master coordinates, and modify all other robots to use
50
  those master coordinates
51
 * easiest is to start a new map with one robot at the origin and
52
   all others starting based off that robot
53
 * backdating the coordinate change would be better
54
 * using both approaches is likely most accurate
55
56
57
58
*/