Implementation of landmarc method with adaptive K-NN algorithm on distance determination program in UHF RFID system

This research was conducted to find the distance prediction between reader and tag using distance determinant program that called “distance program” which applied LANDMARC method with adaptive k-NN algorithm. This method works by assigning a weighted value to k-NN algorithm between all reference tags and tested tag with k determined by key reference tag. This research is different from the research using the same method before [5] which used 2 antennas and has the position of tag in form of coordinates as the output, this study uses 1 antenna and has the distance estimation between reader's antenna and tag as the output. The use of 1 antenna is expected to increase the efficiency of the number of antennas used in one environment to search tags by distance, but still has a good accuracy, in order to not to reduce the performance of the LANDMARC method to get distance determination between reader's antenna and tag. The test was performed on 4 tracking tags, with a distance of 1.4 meters, 1.9 meters, 2.8 meters, and 3.35 meters respectively. Data retrieval is done 5 times on each tracking tag. There are 2 experiment that are applied. The first experiment is to apply 2 test scenarios, first scenario is when there is no object around the tag and second is when there are object around the tag. The second experiment is to calculate the difference of percentage error from test result from both scenarios. The first experimental result showed that the scenario 1 can produce result with the average percentage error of each tracking tag is 1.280%, 1.452%, 2.107%, and 2.470%. While scenario 2 can produce larger percentage error, with the average percentage error for each tag is 3.687%, 4.225%, 4.466%, and 7.430%. The second experimental result showed that the scenario 2 results can have larger percentage error than the scenario 1 results because of the surrounding objects near the tracking tags. The average difference of percentage error between two scenarios is 3.125%.