Restarting the Propulsive Landing Program (Patreon)

Hi Patrons, I wanted to let you know that landings are coming back!

It's been a good while since we did any Echo landing tests - I formally halted the program around December of 2019. Generally, I didn't love the idea of continuing to build on that old codebase Echo used, and there were so many aspects of those UAV drop tests that were hard to dial in, that it didn't seem worth it.

That said, I've got a few new ideas, and I'd love to quickly talk about why I think it's worth trying this again!

For the next landing attempts, we'll be going with full launch and landing each time. The UAV drop tests were meant to simplify things - if you can just focus on landing, it's half the complexity, right? Well... sorta... but those drop tests introduced a lot of fun new ways to add problems as well. Looooonnnng flight times comparatively, string-induced roll issues, pitch yaw oscillation near separation, etc. I'm glad I did lots of drop testing, but I decided to go that route at a time when many of my launches were still failing.

With Sprint, the latest rocket I've been flying, I feel like I have my control system dialed in. The success rate on launch is much higher than it used to be, so let's just get the whole flight in all at once!

The next vehicle to try propulsive landing will be Scout E(we've had A, B, C, and D so far). It'll be a 66mm airframe just like Sprint. AVA will control the vehicle, using TVC for pitch and yaw, reaction wheel for roll. Using that fancy new Kalman filter from the last flight, I have confidence we can lower the horizontal drift rate on touchdown, which will help a lot.

I'm also nearly done with a refreshed design for the thrust vector control motor mount, which you can see above. I'm using very high quality micro servos, which will help quite a bit. Faster response time, and about 0.25 deg of resolution in the servo - the mount has roughly a 2.4:1 gear ratio, so we get roughly 0.1 deg of resolution in the mount, with a much faster response time than the old TVC mount. It's also built to automatically eject the ascent motor and slide the landing motor into place- haven't done the CAD for that yet, but I've got a whiteboard full of ideas! Staging the motor to slide into place like this will solve the Krushnic effect problems I had on Falcon Heavy, which used the same setup that I had intended to use for landing.

I've also designed a set of servo actuated fold-out fins, to help reorient the vehicle for the landing burn after apogee. The design works pretty well so far, but like everything I do, it needs just a few more revisions before it's ready. I'm not totally convinced we need this given how crazy low the COM on the vehicle will be, but glad I designed it anyway!

In order to FURTHER lower the COM, which ultimately helps with not tipping on touchdown, I had Parker(the incredible BPS.space intern!) design a micro chute ejection bay to go right above the flight computer. On a normal flight, we shouldn't need chutes, but it's a bad idea to fly without them, just in case something goes wrong. Having chutes in an isolated bay means the entire top half of the rocket can be dedicated to tucking those fins inside, instead of holding the parachutes.

Parker is also working on new landing legs - this time, NO BOUNCE. That was a big problem with Echo - landing on grass didn't help, but the way those landing legs were designed was inherently flawed. Any kinetic energy on touchdown was always put right back into the system by bouncing, rather than getting absorbed by the material. The new legs have a ways to go, but they'll be primarily made of carbon fiber spars with a thin overwrap to make them look less skeleton-y. The legs will be slightly longer than last time as well - all this combined with the lower COM should help mitigate the tipping issue.

Finally, the big ticket item, throttling! This was always a weak spot with Echo, but again because of that fancy Kalman filter, I'll have much more accuracy in my position and velocity measurements, which means the rocket can make decisions based on the expected vs measured position, velocity, thrust, etc. Here's how it works...

After apogee, the vehicle generates the ideal ignition point to attempt a landing at 5m AGL, not zero. We do this because the biggest variable I observed with Echo was the motor come-up time(how long from ignitor firing, to the start of a normal thrust curve). If we light a little bit late, no problem, we can recover because we initially targeted a high landing point! But if we ignite on time, or anything other than the upper bound on allowable ignition time...

We'll "throttle" the motor, by planning and executing a divert maneuver. I have not run this in a simulation yet, but after speaking to several folks, I have confidence I can set it up. Let's say that we light the motor roughly on time, and we're on track for a solid "landing" at 5m AGL like we planned. We can use a linear model of the vehicle with live performance of the motor with its estimated thrust curve to simulate this, on the rocket, in the air. So we do this, and come up with the percent we need to reduce thrust by. Let's say that reducing the nominal thrust by 5% will get us to 0m/0m/s/perfect landing.

So at T+1 into the burn we have enough data to start burning off that 5 percent - we spend 2.5% of our thrust horizontally by pitching over a bit downrange. Then at T+2 or 2.5, we do that maneuver in reverse to burn off the remain 2.5% and change our final altitude from 5m to 0. There are a lot of flavors of how this can work, and depending on how powerful AVA is feeling, we could run this live simulation multiple times during the burn to update our "cosine throttle". As mentioned, it's also far from ready. I know what I want to do, I have a pretty rough idea of how to do it, but simulations will really help. I'm hoping to get to that this week or next.

Anyway, hope you enjoyed this massive block of text! There's more to come, as always, and I'm very exited to be giving this another shot. Blue skies!

Joe