1. Robot Recognition of Complex Swarm Behaviors
Aisha Walcott-MAS622J-Dec. 11, 2006

2. Courtesy James McLurkin
Introduction
Dispersion
Orbit
A Swarm is a large collection of autonomous mobile robots
No centralized control
Group behaviors are produced from local interactions of many individual robots
Goal is to develop a suite of primitive global behaviors that combine to form more complex group programs
The goal of the Swam project is to developing techniques for programming a large swarm of autonomous mobile robots.  Each robot is autonomous and there is no central controller.  Instead, group behaviors are produced from the local interactions of many individual robots.  A programming system for the Swarm would have to reverse this process; given a desired group behavior, what combination of local behaviors are required?  Our approach is to design primitive global behaviors that can be recombined to form more complex group programs.
Courtesy James McLurkin

3. Build multi-classifiers to classify Complex Swarm Behaviors
Project Goal
Build multi-classifiers to classify Complex Swarm Behaviors
Disperse
Orbit
Cluster
Bubble Sort
Example Features
hi source
source *
source
low source

4. Approach Collect raw behavior data sets Determine Features (8D)
Pattern Recognition Algorithms
KNN
Neural Nets
Bayes Nets
Analyze results of each algorithm

5. KNN Tested a range of values for nearest neighbors random tie break
Overall Correct Classification
Average Class Classification
Cluster= 100%
Disperse = 12.5%
Clump = 50%
Orbit = 44%
Bubble Sort = 82%
k =
total_class_correct =
0.6250
0.6000
0.6500
0.5250
0.5500
in_class_correct =
tests_per_class =
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN

6. Neural Nets Single Hidden Layer Layer 1: nodes [50,70]
Max percent = 65%
Logsig
Two Hidden Layer
Layer 1: nodes [50,70]
Layer 2: nodes [25,25]
Max percent = 65%
2 layer
percentCorrect =
Columns 1 through 12
Columns 13 through 21

7. Bayes Nets
Mapping to discrete domain by applying k-means clustering to each feature
Preliminary Results
Classification of Cluster
Possible bug in code
Modify the discrete mapping
Cluster, Disperse,Clump,Orbit, Bubble Sort
To ensure that a model of an environment is current, mobile robots must use information in the model to determine which regions in the environment from where to gather data. The data is, in turn, used to update the model. We refer to the process of continuously updating large-scale dynamic models of the physical world as world crawling.
Database: 1) What is the object?
2) Where is the object relative to robot “me”
3) Where am I relative to home (or local origin)?
4) What time is it? 1 2 8

8. Discussion KNN and Neural Net performed well
Determining the mapping from real numbers to a discrete domain may affect Bayes Nets classifiers
Overall high classification of clustering behavior
-Features tuned to behavior
-Not enough variety of samples
Need more samples of varying behavior

9. Next Steps
Feature selection-which group of features work best for each classifier
Additional experiments to determine why certain classifications are much better
Future
Use the temporal information to learn hidden emergent sub-behaviors
(feature vector changing similarly over time)

10. Thank You