PilotAware has been supporting a pure Bearingless target mode of detection for a number of years. This has been somewhat controversial with supporters and opponents having equally considered views on its usefulness. Given this background we have been working to improve this mode of operation with enhanced capabilities. Hence the idea was developed for Mode-S/3D, to augment the existing system with MLAT data, provided by 360Radar, to improve upon the data available from Mode-S bearingless targets. The Mode-S/3D Trial is due to conclude its first phase, therefore we would like to share our findings on the usage over the last 3-6 months, of this joint undertaking between 360Radar and PilotAware.
There have been a number of questions raised regarding the accuracy of the data supplied by 360Radar. The following shows how the position reports are being gathered and analysed.
Reports are measured from 2 specific viewpoints. In-aircraft and from the ground-station
Participants of the trial have been submitting their in-flight data recorder tracks back to PilotAware for analysis.
Trial aircraft that are Mode-S equipped provide an ideal platform, because this allows the round-trip delay (latency + computation time) and the accuracy of the MLAT data to be measured. 1090MHz signals derived from the on-board Mode-S transponder are captured by 360Radar, sent to the central server, relayed to the OGN-R, and uplinked to PilotAware units. The MLAT reported position is then compared with the GPS reported position. This gives an absolute error value for a reported MLAT sample, combining network latency, computation, transmission and reception errors.
10,000's of data points have been received to produce individual and cumulative data for statistical analysis.
As the Trial progressed all aircraft equipped with PilotAware and a Mode-S transponder (not just the 45 trialists) were included in the analysis using data simultaneously captured from the OGN.
This allowed the time and position of the MLAT and PilotAware (GPS based) traces to be compared.
Combining the two methods has produce significantly more data than would have been produced by just 45 airframes alone. It has also demonstrated the UK reach of 360Radar
How good is the accuracy of reported MLAT positions?
Accuracy of Reported Positions
The graph above shows the difference in distance between the reported positions alternating between either PilotAware->MLAT or MLAT->PilotAware. These have been placed into distance buckets to fit with the GDL90/ADS-B metric of NACp (Navigation Accuracy Parameter). These quantized fields are distances in nautical miles of 0.05nm, 0.1nm, 0.3nm, 0.5nm, 1.0nm, 2.0nm, 4.0nm, 10.0nm respectively.
This shows that 98.29% of reported MLAT positions were within 0.3nm of the reported GPS positions. This takes into account all computation time, network latency, and uplink processing. Importantly over 77% of the reported positions were within 0.1nm. This is far and beyond the initial project expectations, for enhancing and augmenting Bearingless Traffic data.
A large proportion of ADS-B transponders using a non-certified GPS, emit their positions with NACp figures equivalent to NACp=3 which equates to 2nm.
Angular Bearing at Distance
An analysis has also been done from the perspective of a given ATC traffic service where a distance, relative altitude and bearing is provided. For instance,
“Traffic, one o'clock, 3 miles, northbound, same level”
How accurate is a traffic service, and how useful is it?
Firstly, the bearing can only ever be accurate to within 30 degrees due to the rounding error of using the clock face as a reference (360/12). How does this compare to the angular reports using the same data as above (Accuracy of Reported Positions), but from the perspective of the ground-station viewing 2 position reports (MLAT and PilotAware).
Please Note, the Bar chart and Line graph use different Axis, this is because the line indicating the cumulative values would snap to the top of the chart, so the secondary axis on the right is used for the line (cumulative), and primary axis on the left for the bar (individual) traces.
The data shows that for 98.67% of the reported MLAT positions, there was an angular bearing of less than 5 Degrees, from the true GPS reported positions. This data is encompassing traffic at all distances. The outliers will be derived from the traffic which is very close (as expected), whereby the positions arepotentially more in error either due the MLAT calculation error, or the refresh of the position report.
Refresh rates are defined as the time between the updates of the position reports of an Aircraft. Consider some of the current technologies.
PilotAware ~ 1.5 Seconds
FLARM ~ 0.5 Seconds (using frequency hopping on 2 frequencies)
RADAR Head, this is variable and difficult to get a definitive source, anecdotally I have found references to approximately 10RPM, which would equate to ~ 6 Seconds
ADS-B ~ 0.5 Seconds, but a full position report is fragmented over a number of frames, so 0.5 seconds is not actually a position update (TODO work out the definitive number)
Given the above it is important to note that the refresh rates of the MLAT position is based upon two important factors
Factor 1: MLAT is a purely passive system, it relies upon the interrogation of the airborne transponder, either from the ground SSR or from a TCAS system. A ground based SSR only wants to receive a response to its own interrogation, as it is using a directional beam interrogator. Any other response received is effectively unwanted noise (see FRUIT and GARBLE). In the case of MLAT, it is beneficial to see all responses, it is not using a directional beam, instead it uses an omnidirectional, antenna, utilising multiple receivers.
Factor 2: MLAT is computationally intensive, it takes time to perform many calculations in order to resolve a position given the data.
More analysis on the existing refresh rates of 5-6 seconds is required (this is a TODO). The very good news, is that we already know how to improve this figure. At the moment, the 360Radar MLAT server is being enhanced to use parallel computation to improve the throughput of the MLAT refresh rates.
The Feature of providing MLAT data from PilotAware is disabled by default, and only enabled by the user actively selecting the option to do so. Consequently it is important that users understand the operating principles of Mode-S/3D.
Electronic Flight Bag depiction is important for conveying the importance and urgency of the MLAT uplinked data. In PilotAware RADAR we display a region of ambiguity, this is calculated from the timestamps associated with the position report, and the age of that position report.
If the position report has become too old or too inaccurate, that it can no longer be effectively utilised, then the aircraft reverts to being considered Bearingless and thus if judged as requiring notification (using existing algorithms), will be depicted as a classic Bearingless target - accurate altitude, unknown bearing.
Our aim is that the user has been provided with enough advanced warning of these MLAT positions, that visual acquisition is obtained prior to any close encounter, this is a much better proposition, than simply being presented with a bearingless warning, with no advance notification. In all situations, Electronic Conspicuity, of any description, does not replace the responsibility of the Pilot in Command to maintain a continuous and thorough visual scan as the Primary Situational Awareness input.
As mentioned earlier in this paper, the improvement of refresh rates is the low hanging fruit we can and will improve. Accuracy seems to be effective when compared to existing technologies, and will be further improved by increased OGN-R ground coverage and improved computational server speed and capacity.
Additionally, we need a method of describing to the EFB what we calculate as the accuracy of the data, this can then be used to display that information in the most informative manner, or not at all.
The addition of MLAT uplinked data, is not a panacea to perfect situational awareness It is another tool in an existing toolbox of capabilities. The technology is new, and improving rapidly, along with the network capabilities we have with the 360Radar and OGN. We look forward to feedback from users to help guide and improve this technology.