Measures of Central Tendency: Mean, Median and Mode
Terje Midtbø - Test of an indoor navigation system
1. Are indoor positioning systems mature
for cartographic tasks?
Evaluating the performance of a commercial
indoor positioning system
Terje Midtbø
Norwegian University of Science and Technology
4. Overview
• Indoor maps
• Indoor navigation
• Testing the WiFi positioning system at NTNU
• Measurements
• Results
• Discussions and conclusions
5. Do we need indoor navigation?
• Localization of persons and equipment at
for example hospitals
• Find your way in large complex and
unknown buildings
• Find products in large shopping areas
• Most efficient shopping based on shopping
list
• Tourists at a museum
• Alternative emergency exits
• Localization of firefighter
• …….
23. WiFi positioning systems
• Time for signal one way
• Time for signal both ways
• Time for signal in combination with angle
to the access points
• Phase measurements
• Signal strength
28. Measurements
• Accurate measurement of control points by
surveying equipment and methods
• New measured coordinates for the access
points
• Coordinates measured in UTM, Zone 32N
• Transformed into local system used by the
Cisco system. Units in feet.
29. New coordinates for the access points
• Average 2D differences: 11.7 feet
• Min: 3.7 feet
• Max: 19.2 feet
32. Equipment for the measurement
• Laptop with external antenna
• Fixed height over the floor
• Easy to move to next position
33. Measurements based on 4 configurations
• Old coordinates for access points – no
fingerprinting
• Old coordinates for access points with
fingerprinting
• New coordinates for access points – no
fingerprinting
• New coordinates for access points with
fingerprinting
34. Measurements based on 4 configurations
• 10 points were measured in the 1st floor of
Lerkendalsbygget
• 5 measurements for each point in each
configuration
42. • More accurate coordinates for access points gave no
significant improvement for measured position
• Fingerprinting as it is used today shows no significant
improvement in the measured position. Lower variance
is still observed.
• Correlation between different types of measurements in
the same control points indicate that geometry and
obstacles are the dominating sources of error
• Further research: include obstacles in the model
43.
44. • More accurate coordinates for access points gave no
significant improvement for measured position
• Fingerprinting as it is used today shows no significant
improvement in the measured position. Lower variance
is still observed.
• Correlation between different types of measurements in
the same control points indicate that geometry and
obstacles are the dominating sources of error
• Further research: include obstacles in the model
• Still too poor precision and accuracy for turn-by-turn
navigation
• Acceptable for localize equipment
• Acceptable for overview of peoples position in social
applications