Ads-b and sailplanes

On Monday, April 6, 2015 at 3:39:17 PM UTC-7, Andy Blackburn wrote:
The filtering of collision threat versus non-threats in a glider scenario is the entire difference in suitability between Flarm and ADS-B. Trying to do this kind of filtering without a position forecasting algorithm is challenging because you can't effectively filter threats from non-threats. Without a common position forecasting algorithm across all systems you might have one system give collision advisories based one one set of assumptions and another system issuing advisories on another set of assumptions and could lead to "you zig, I zag" kinds of asymmetric warnings and pilot reactions.

I'll explain my (over-simplified and and I'm open to accept wrong) view. Flarm and ADS-B start out with the same source data, a position and velocity vector obtained from GPS. The primary difference is that Flarm calculates a forecasted trajectory (I've never looked at the details, but I wouldn't be surprised if it's more of a probable trajectory sphere at some time t+x) from the raw data, ADS-B simply transmits the raw position and velocity vector. The great advantage of the Flarm approach is that all it need do is intersect all of the received trajectories (or spheres) with the calculated trajectory for your own glider, which is great if you need to do it on an 8 bit processor. A lot more computation is required to accomplish the same thing with received ADS-B data, as the trajectory calculations need to be performed for all received targets likely to be a threat. That is not impossible, it just takes a lot more computational power in your glider, but we live in the age of low power consumption 64 bit processors that cost a USD or two in reasonable quantities.

The fact that all Flarm devices calculate threats in the same way is helpful, but in the end, it has no control over what any of the pilots involved do to avoid a collision, whereas with TCAS-II, the pilots most certainly should do exactly what their device calls for. I don't know if Flarm transmits data more frequently than the once per second slots I believe are allocated for ADS-B, perhaps someone else knows. Beyond that, however, I see nothing preventing anyone from taking the received data, filtering out the more distant targets, and applying target type (including glider) specific trajectory calculations to more intelligently provide traffic advisories. If someone has facts that suggest otherwise, I'm interested in hearing them.

I think the important thing to keep in mind is that gliders probably won't be exempt from the ADS-B (or transponder) mandate in the US forever. Boeing, Amazon, Google, and the rest of the US drone industry will eventually figure out (if they haven't already) that the easy solution for their problems is to lobby hard for all aircraft in US airspace to be ADS-B out equipped.. We don't have that kind of clout, nor does AOPA or EAA.

Marc

On Monday, April 6, 2015 at 6:37:14 PM UTC-7, wrote:
On Monday, April 6, 2015 at 3:39:17 PM UTC-7, Andy Blackburn wrote:
The filtering of collision threat versus non-threats in a glider scenario is the entire difference in suitability between Flarm and ADS-B. Trying to do this kind of filtering without a position forecasting algorithm is challenging because you can't effectively filter threats from non-threats. Without a common position forecasting algorithm across all systems you might have one system give collision advisories based one one set of assumptions and another system issuing advisories on another set of assumptions and could lead to "you zig, I zag" kinds of asymmetric warnings and pilot reactions.

I'll explain my (over-simplified and and I'm open to accept wrong) view. Flarm and ADS-B start out with the same source data, a position and velocity vector obtained from GPS. The primary difference is that Flarm calculates a forecasted trajectory (I've never looked at the details, but I wouldn't be surprised if it's more of a probable trajectory sphere at some time t+x) from the raw data, ADS-B simply transmits the raw position and velocity vector. The great advantage of the Flarm approach is that all it need do is intersect all of the received trajectories (or spheres) with the calculated trajectory for your own glider, which is great if you need to do it on an 8 bit processor. A lot more computation is required to accomplish the same thing with received ADS-B data, as the trajectory calculations need to be performed for all received targets likely to be a threat. That is not impossible, it just takes a lot more computational power in your glider, but we live in the age of low power consumption 64 bit processors that cost a USD or two in reasonable quantities.

The fact that all Flarm devices calculate threats in the same way is helpful, but in the end, it has no control over what any of the pilots involved do to avoid a collision, whereas with TCAS-II, the pilots most certainly should do exactly what their device calls for. I don't know if Flarm transmits data more frequently than the once per second slots I believe are allocated for ADS-B, perhaps someone else knows. Beyond that, however, I see nothing preventing anyone from taking the received data, filtering out the more distant targets, and applying target type (including glider) specific trajectory calculations to more intelligently provide traffic advisories. If someone has facts that suggest otherwise, I'm interested in hearing them.

I think the important thing to keep in mind is that gliders probably won't be exempt from the ADS-B (or transponder) mandate in the US forever. Boeing, Amazon, Google, and the rest of the US drone industry will eventually figure out (if they haven't already) that the easy solution for their problems is to lobby hard for all aircraft in US airspace to be ADS-B out equipped. We don't have that kind of clout, nor does AOPA or EAA.

Marc

Yes.

There are a couple of additional consideration on prediction pre- versus post- transmission. First, which is which is more robust in the event of lost packets, which happens pretty regularly with RF transmissions. Second, while both Flarm and ADS-B transmit at once per second, the inherent lag associated with post- versus pre- processing apparently works in the favor of Flarm in highly dynamic situations like thermalling. Add to that the lag associated with whether the traffic is ADS-B direct or ADS-R which increases the uncertainty of any predicted probabilistic position bubble and you can start to see why using ADS-B for common glider scenarios could easily devolve into a cloud of false alarms which, if the lags get close to the human reaction time of ~300 milliseconds (plus whatever dynamic response time constants of the gliders are), could melt down the prediction-reaction dynamic loop entirely.

I know less about the politics of it - whatever goes on drones will need to be cheap, cheap, cheap, and possibly lower power and range to keep the aggregate bandwidth requirements down (especially if Amazon gets their way). Flarm may have a leg up on ADS-B on both counts, but the FAA would need to agree and counting on the FAA to do sensible things would.

Andy