Testing: Four, five, six

We commented on Testing: One, two, three, that we would like to develop a software tool which would allow anyone to fine-tune his/her installation to a position as accurately as possible. This tool is ready and available in mide repository.

It is a Java command-line tool developed using Ubica code. To use it, it is only necessary to consider that:

• The name of each dataset indicates in which coordinates it has been created (x1000y2000.txt = {“X”:1000,”Y”:2000}).
• The configuration is defined in the config.properties file.

When we wrote the first Testing post we captured five datasets placing a beacon in different coordinates (the datasets are also available in the repository):

{"X":1800,"Y":1000}
{"X":3400,"Y":1000}
{"X":3400,"Y":1900}
{"X":5300,"Y":1900}
{"X":6300,"Y":1900}

We have tested the tool with those datasets and the results have been better than we expected. To put you in context, in the first Testing post we indicated the following:

At the beginning of the challenge, we carried out several tests until we achieved a maximum error of two meters when positioning. This is enough for us because it allows knowing the rough position of each beacon. In a nursing home, we could know the room in which each resident is located. If it is a medium-size room, we could also know the area in which the resident is located.

After trying different configurations we have obtained the best results with the following values:

dataset=dataset
distance.algorithm=Accuracy
least.squares.algorithm=NonLinear
retention.time=60
publication.rate=1000
delay=5
scanning.window=10000
attenuation=0.5
cutoff.rate=0.9

With this configuration the results have been as follows:

{"X":1800,"Y":1000}: 1,137 m.
{"X":3400,"Y":1000}: 0,736 m.
{"X":3400,"Y":1900}: 0,570 m.
{"X":5300,"Y":1900}: 1,084 m.
{"X":6300,"Y":1900}: 1,277 m.
Mean error: 0,961 m.

An error of less than one meter. Far beyond our expectations 🙂