02 January 2014
When evaluating the suitability of various location intelligent solutions, it is useful to have some technical understanding. In the following blog post we will discuss how location accuracy and tag range play an important part in choosing a location intelligent system.
LOCATION ACCURACY AND LOCATION CONSISTENCY
Location based applications are most often compared in terms of the accuracy provided by the sensor system. Location accuracy is certainly a critical parameter for quantifying system performance, but when quoted out of context it can prove misleading. A location sensor system is only fully defined when location accuracy is quoted in conjunction with a consistency metric.
Consistency in this context means a measure of how often the location system delivers the stated accuracy. The game of darts provides a simple example: even the most amateur player can score a bulls-eye if enough darts are thrown, but the measure of a good player is: how often can he or she hit the bulls-eye?
Figure 4-1 (above) shows an example of the output of a typical location system (or, the places where the dart hit the board in the above analogy).
This plot is generated by counting the number of measurements that lie within the error circles shown in Figure 4-1. In this example, 20% of the measurements lie within the 3’ error circle, 70% are within the 15’ circle (including those in the 3’ circle) and 97% are inside the 30’ circle. Figure 4-2 shows that data plotted in blue. The orange line shows an example location system with a much tighter constellation of measurements around the true location, with 95% of measurements inside the 3’ error circle.
The blue and orange lines in Figure 4-2 represent a true comparison between location technologies. By simply comparing location accuracy, misleading results can be interpreted if the accuracy figures are quoted at different consistency levels. In the darts analogy, the orange line represents a good player, and the blue line a poor one.
Figure 4-3 shows how consistency levels become important in location enabled applications. In this figure seven zones are shown, with the item being tracked located in the center zone. In the case of a 3’ error measurement the item is correctly associated with the center zone. When the error is 15’ the item could be in one of three zones, and with 30’ error in one of five.
Reading from Figure 4-2 the different location systems can now be fully compared as shown in Table 4-1.
The importance of Table 4-1 is that it compares systems based on not just how well they perform, but how often they perform to that level. It’s clear that both systems can claim to identify the location of a tag to within a single zone, but that the high accuracy system does so with significantly more consistency than the low accuracy system. In a real production environment, location system performance must exceed 99.9% for many mission critical applications.
OVERVIEW OF LOCATION TECHNIQUES
All location intelligence technologies share the common theme of some kind of tag attached to the object being tracked, and an infrastructure of sensors that determine the ID and location of the tag. Such tags may emit optical, ultrasonic or radio frequency energy, but radio frequency (RF) tags are the only ones suited to industrial environments (since optical and ultrasonic signals are too easily jammed or blocked in typical factories).
In all cases the sensing infrastructure attempts to measure either the range or bearing (or both) to a tag from one or more sensors and from that information the tag’s location is calculated. The technologies are predominantly differentiated buy how well they are able to measure range and/or bearing and, as a result, how accurately they are able to determine tag location.
MEASURING SIGNAL STRENGTH
Many location systems make use of the fact that the signal strength from a tag becomes weaker as the tag moves farther from a sensor. These systems measure the signal strength measured by multiple sensors, use that information to estimate the range of the tag from each sensor and then calculate the tag location. Figure 4-4 shows how this works.
In Figure 4-4 it is clear that three sensors are required in order to locate a tag in two dimensions. (By extension, four sensors are required for a 3D location determination.)
Tag location errors are cause by many sources, the dominant of which being variations in signal strength caused by effects other than range. If some other effect acts to reduce the signal strength then the sensor interprets this weak signal as longer range than reality. Figure 4-5 shows how range errors from each sensor contribute to overall tag location error.
Range errors are the limiting factor of signal strength systems since many factors affect signal strength beyond just the spacing of the tag from the sensor. Examples include attenuation due to signal blockage and variations due to tag orientation. The dominant factor, however, is an effect called “multipath” which is a factor in all location system types, and so is worth a short aside to aid in understanding.
When a tag emits a signal, that signal is, by design, emitted in all directions so as to reach as many sensors as possible. This scattergun signal will bounce off any reflective surface, such as metal objects, and a single sensor may hear the signal from one tag echoing off multiple surfaces. The effect is exactly the same as voice echoes heard when shouting in a large cavern or in the mountains.
In a manufacturing environment, which is filled with metallic objects, the chief concern of location systems is to cope with the damaging effects of multipath. For signal strength systems multipath can be very damaging indeed due to an effect known as signal cancellation.
A radio wave is just like a wave on the ocean: an undulating surface which moves along at a certain speed (the speed of light in the case of radio waves). In the ocean, waves can come from multiple directions and when two peaks meet a very large wave can build up locally as the peaks add together. Similarly a peak from one direction can coincide with a trough from another direction and these completely cancel each other leaving no wave at all. The phenomenon is familiar to surfers who understand that waves come in sets: they wait through the periods when the peaks and troughs are cancelling each other leaving very small waves, instead hoping to catch a period of peaks combining with peaks for the biggest surf of the set.
The exact same signal cancellation phenomenon affects radio waves as they bounce off multiple reflectors and arrive at a sensor from many directions. The adding and cancelling of peaks and troughs look to the sensor like increases and decreases in signal strength, which that sensor interprets as decreases and increases in tag range even though the tag may not be moving at all. The result is that the error bands in Figure 4-5 can grow very wide due to multipath, causing large errors in tag location.