1. Data Set Analysis
Wall-Following Robot Navigation Data Data Set(4 US Sensor data)
Wall-Following Robot Navigation Data Data Set(24 US Sensor data)
Amit Ghosh
011132134
4. Dataset Information
(Source : UC Irvine Machine Learning Repository)
Type Classification
Characteristics Multivariate
Number of Instance 5456
Number of Attributes 5
Attribute Characteristics real
Date Donated 2010-08-04
Distinct Class 4
Missing Values? N/A
7. Naive Bayes
Accuracy
89.11 %
a b c d <-- classified as
1883 176 106 40 | a = Move-Forward
36 753 37 0 | b = Slight-Right-Turn
104 48 1927 18 | c = Sharp-Right-Turn
5 0 24 299 | d = Slight-Left-Turn
*10-fold cross-validation
Confusion Matrix
8. Decision Tree
CART and Random Forest
give same accuracy
*10-fold cross-validation
ID3
100 %
9. Applying this algorithm in 24 US Sensor model
*10-fold cross-validation
KNN NaiveBayes OneR ID3
88.1782 % 52.456 % 75.6415 % 99.6518 %
10. I select ID3
● Highest Accuracy
● Easy to implement in
Robot
● Only if else condition
● We can remove
unnecessary sensor like
US3 and US4
● Reduce cost
ID3
100 %
11. Implementation
if US1<=0.9
Sharp-Right-
Turn
else if US2<=0.494
Slight-Right-
Turn
else if US2<=0.901
Move-Forward
else if US2>0.901
Collect sensor data from Robot
Apply ID3 and Generate Rules/Condition
Put the condition in the controller program
The data were collected as the SCITOS G5 robot navigates through the room following the wall in a clockwise direction, for 4 rounds, using 24 ultrasound sensors arranged circularly around its 'waist'.
The data were collected as the SCITOS G5 robot navigates through the room following the wall in a clockwise direction, for 4 rounds, using 24 ultrasound sensors arranged circularly around its 'waist'.