oyster_catcher

Sorry. I've been off KSP for a long time. I suspect you've found out yourself. I don't know precisely. I think it might work ok. I'm guessing the switch forward/back in time will be like loading/saving the game in which case the correct settings for BG and correct phase of flight should be set to the correct values from that game state. So I hope it should work. Let me know what you find/found.

Sorry everyone. Its been a long time for me. So many comments I haven't responded to. This inspires me to get back to it and sort out any compatibility problems with the latest KSP. So the mod should work whether FAR is installed or not, it willl just use FARs atmospheric prediction if available - though this was quite hairy to get to work. Let me know if you see any bugs

If you can arrange to save the game the BG settings will also be saved with that game as they get saved with the command module that BG is attached too. That doesn't cover all situations but may be it'll help you.

Thanks so much for the kind words and beautiful video. What was that Kerbal thinking? I think 10-100m error is about the best I can achieve right now. The biggest issue is when aero-forces and rocket thrust just about equal each other, one pushing one direction and other pushing the other. At this point you have no control authority, and the mod is quite conservative and applies a big safe zone around this when it will not steer as steering the wrong way is quite a possibility as sometimes the forces aren't predicted accurately enough. It does use the built in physics get these forces but still it doesn't always work out quite right. I've struggled with this sometimes and have tried to optimise as much as I can but retain accuracy. Usually the drop in FPS should be manageable, sounds like something is going very badly wrong. Switch on Debug in the advanced tab as I think it may show the XXms to do the simulation. it should be 4-60ms and it does this out 10 times/second I think so its quite significant and worse case takes 50% CPU time. Please let me know what that says. The mod does a lot of work and predicts the flight path all the way to the ground. It will be slower the more time it'll take to land, though it takes much bigger steps when far out. It possible your craft would never land for some reason, say if minimum thrust was too high, or too far away, in which case the simulation would run until it hit a maximum time of 600 secs, which would be quite slow.

It designed to work with FAR if its present or else use stock physics, and I thought I'd tested this well enough. So the error shouldn't mean that BoosterGuidance failed all together. So you can ignore the error. FAR is not required. Does BG still appear to be working or not? Thats odd. A video would help, but anyway the thing to work out is, is BG trying to do the right thing but the craft is aerodynamically unstable? Select Debug in the advanced tab and you will see green line showing what direction BG is trying to steer the craft too. If this isn't achieved them probably BG is trying to do the right thing but the craft is not stable enough to achieve that.

Hmm. The mod shouldn't be causing the booster to flip. If you toggle "Debug" in the "Advanced Tab" you will then see a green line which is the desired orientation of the rocket pointing in the direction of the front of the rocket, so it should be pointing upwards. It'll be very useful to know if this is still pointing up, if so then somehow the aerodynamics of the booster has flipped it, not BoosterGuidance, but possibly BoosterGuidance has made its steering unstable. So I suggest, enable "Debug", try setting up the steering gains to far left so it won't try to steer at all, and seeing if it still flips. Also, at which stage does it flip, in re-entry burn? (I have seen a rocket start to turn with thrust vectoring from the main engine, the main engine cuts and then it doesn't have enough RCS or reaction control wheels to stop the turn, this could be happening - more RCS thrusters or reaction control wheels, and less steering gain in re-entry burn would help this)

Fantastic! Glad it worked for you and thanks for the kind words. I've tried to make the mod as easy to use as I can but it still takes quite a bit of experience to get it working really well. Thanks for persevering to get to that point. It'd be interesting to know if your super-heavy had enough thrust to slow down with the hard dive without using HyperEdit, perhaps it did, the mod will aim to use 90% of the maximum thrust to slow down so I can be very late with the sufficient thrust and a light booster. If you can land a super heavy with 10m error than I think you are ready to build a tower that can catch it out of the air the way Elon is proposing. That would be a fantastic feat to pull off! I'm not quite sure how to build a launch tower that could do that, but perhaps you can or definitely someone in the community can and probably will be doing that. Part 2 of BoosterGuidance tutorial uploaded. Covers controlling multiple boosters/vessels (for Falcon Heavy style dual landings), setting landing burn engines and logging/plotting. See The logging can unfortunately sometimes fail to enable due to some kind of C# file opening error (files not getting closed or some file opened twice or something). If it fails (BoosterGuidance won't get enabled with logging switched on) delete all the ~/${KSP_INSTALL}/*.dat files and try again. Sometimes it might require a KSP restart. I've been working on fixing this for a while but its tricky to debug, so sorry if logging sometimes fails.

I agree. Thats very painful to have to set those values every time. The settings get saved when you use "Save Game". I think the problem is that the craft is saved with the default settings from the VAB. So when you reset to launch, which is like loading the saved game at launch the settings are reset. So I'll at least try and work out either how your craft can get saved with your settings, or someway they don't get reset with a revert to launch. Thats the first step. Thanks

Excellent, glad it works for you. Ok, that's an interesting thought, would do you tend to change for the different types landings? I tend the settings depend a bit more on the vessel rather than the landing type so they are saved with the vessel. I'm a bit wary of adding the complexity for something I'm not very sure had I'd use it so I'd be interested in your reasons

A video would really help please, but, when did it go wrong? In the re-entry burn, aero-dynamic descent or landing burn? The most likely cause is that the booster is being steering the wrong way, this can happen when the engine is ignited and travelling fast. This could be re-entry or landing burn. The aero-dynamic descent is usually well behaved. If this happens then move the gain slider for that phase far left and the error should at least not grow any more. I'm struggling to work out what the specification would be. In reality you would boostback after you separate the 2nd-stage. I guess you want something like the furthest point on the predicted trajectory where you would still have enough fuel for boostback, re-entry and landing? For this I'd need to predict fuel consumption and fully understand the change in mass from staging, and understand which components you want to send back after staging. This is very difficult if possible at all as the KSP API doesn't really tell you what you would get in the future after staging. I think it might be possible the work something out without trying to predict staging, but I'd have to work out fuel consumption and thats quite complex too. I'll bear it in mind as feature but its some way off, sorry.

I suggest you check out this video which gives a rough idea of an easily manageable boostback, but you can actually be much more extreme than this but it gets a bit touch and go! It took a while before I got a real for it. This video is great! Generally you want to still be climbing after you have finished boostback which means you have good height and will fall not too far off vertical. So, I tried to reproduce your trajectory with a near horizontal re-entry burn. I tried it out in realism overhaul in v1.1.1 (just released). I have now fixed a number of bugs and amazingly I managed to 'land' (almost!) within a few hundred metres of the landing target. I accidentally took off again as the throttle switched to maximum when BoosterGuidance got disabled on landing! But I did get some extreme jitter of the landing prediction due to the highly non-vertical trajectory before landing. See my attempt here:

v1.1.1 released Basically quite a few accuracy improvements and bug fixes. v1.1.0 had quite a major bug, the simulation wasn't computed correctly which could lead to very large errors in the landing prediction. The simulation ran fast though so didn't slow down the game much. Now it works again but is slower, but I've tuned it a lot to run as fast as possible which minimizing 'jitter' on the landing predictions. This is still higher than I'd like to some cases and it runs slower than I'd like. In my machine I've seen to take up to 50ms (x10 times a second, so taking 50% of the CPU!) when furthest from landing but mostly its not worse than 20ms. The jitter is generally low but I've seen it get high when coming to land at a low angle in Realism Overhaul. - Upped throttle gains which avoids very heavy vessels hitting ground (depends on dt in simulate) - Target crosses now move when selecting vessel or navigation target - Landing burn works correctly when clicking button even if aero-descent not enabled first - Fixed bug in simulate that led to landing burn not being simulated below 500m (leading to jitter in predictions) - Tuning of simulation to make it as fast and accurate as I currently can - Now runs faster for planets with no atmosphere - Reduced maximum gains to try and reduce the near to tweak gains to reduce max angle-of-attack in RO (gains still needed to be low for RO) - Reduced jitter to landing prediction by reducing timestep of simulation when near ground (can still get bad jitter is some scenarios) Get it here https://spacedock.info/mod/2587/BoosterGuidance and here https://github.com/oyster-catcher/BoosterGuidance/releases/tag/v1.1.1 Released under GPL v3. I'd definitely recommend users upgrade. A full tutorial video is *still* in the works.

Thanks for answering. Correct. It depends on the reduction in velocity you need for re-entry burn plus landing burn. If you don't need the re-entry burn then you just need enough dV is bring your booster from terminal velocity need the ground to a stop (if on Kerbin), these depend on the settings and fast your booster is going at the reentry burn so its complicated. However the mod would be able to do a rough calculation of dV, but calculating the how much fuel is used looks complex and I have up on that. For now its not possible, you just need to experiment and get a feel for it. Do you think 500m is high enough? If not raise the touchdown margin to give me time. If its just the tumbling thats the problem I would pull the steering gain for the landing burn all the way to the left so it just steers retrograde, this should stop the tumbling. I've be pleased to see if this work. If you are able to make a video (make sure you include the BoosterGuidance main window) that'd help me diagnose it. Also, I've found a few bugs in v1.1.0, v1.1.1 is coming soon which will be more accurate.

Thanks, thats very useful. Its great to see people using the mod. I have a few suggestions. (1) You kept all engines enabled for boostback. This makes for the mod to reduce thrust enough to delicately reduce the minimum target error when it gets small since checks are only done 10 times/sec. Use fewer engines for boostback, re-entry and landing burn as the booster is much lighter and is super-sensitive to thrust (I set up an action group to toggle nines). Your video didn't show what the final target error was at the end of boostback. Take a took at this, it'd be good to see how low it gets. (2) You were quite ambitious with a boostback burn so far away from the launchpad. The re-entry burn is nearly horizontal. I would expect the re-entry burn to be 20 degrees off vertical max for good accuracy. Because the trajectory is very horizontal prediction of aerodynamic drag makes the landing spot very prone to error (you can see that if the aero-drag calculation is wrong with a vertical trajectory then the prediction error on the ground won't change much). The landing was 126km off target was is embarrassingly bad though I couldn't tell if this was an overshoot or undershoot. It is possible to get the landing error to a few metres. Staying a bit closer when doing boostback will make it easier for the mod to be accurate. (3) Again because of the large number of engines with high thrust and landing burn thrust is applied very late (even though landing burn is enabled many seconds earlier thrust is not applied). With fewer engines the landing burn will span over more time and will work more accurately. (4) Raising the "touchdown margin" in the advanced tab by perhaps by 30-40m will help, even if you do nothing else. It looked like the booster hit the ground at 80-90 m/s and decelerating at perhaps 40m/s/s so it needed just 2 extra seconds of deceleration. On (4) I'm working on making the landing burn more accurate for high thrust rockets. I've learnt some more things - Upping the rate of thrust calculation estimate from 10 per sec, to the maximum physics rate - The vessel altitude given from the probe corer rather the origin of the booster of the CoG. This meant some of my calculations were off. - Upping the gain of the throttle. So the throttle will re-active more rapidly of the booster is travelling too fast. The gain can't be too high otherwise the throttle will oscillate around the required amount. The first change was in v1.1.0. The next changes are coming soon along with a full tutorial. I hope you get it working a bit better.

I'm not sure. Could you make a short video? That would help see the problem. Even the earlier version shouldn't have a bug that appears this big

v.1.1.0 released - Finally fixed steer gain calculation (perhaps) as several bugs discovered. Further improvement that mean craft should steer in the correct direction during landing burn (or not steer at all) - Added action group to toggle guidance - Added language localization tags for all text. Included en-us.cfg - Reports messages to screen from controller such as phase changes - Landing burn gain reduced so steering less aggressive - Fixed bug where target position wasn't copied to all cores, so target was lost on staging - Stopped slowness of 40-50ms of computation time for some vessels by stopping very frequent calls in simulate Released under GPL v3. Download from https://spacedock.info/mod/2587/BoosterGuidance or https://github.com/oyster-catcher/BoosterGuidance/releases/tag/v1.1.0 If it is slowing down but hits the ground a little too fast, SN8 like(!) then increase "touchdown margin" in the advanced tab by 10m or so, which means at will slow down earlier and then slowly descent at "touchdown speed" from there on. So it gives a safety margin. If you see "too heavy" then the vessel has insufficient thrust for its weight to slow down enough. You might also see an improvement with v1.1.0 I just uploaded. Previously it would re-compute the required thrust only 10 times/sec. This might be insufficient if you have really high thrust engines where all the deceleration happens in the last second. v1.1.0 will compute the thrust every frame, like 20-60 times/sec when near landing. This might help. It'd be great if you tell me if any of this works for you

Just ModuleManager. If you don't install it you will see the message "No BoosterGuidanceCore" in the main window

@TheKurgan I've added Enable and Disable BoosterGuidance action groups in v1.0.4a. See https://github.com/oyster-catcher/BoosterGuidance/releases. I missed out Toggle BoosterGuidance but that will be in the next version with other improvements

Thanks. Currently it would steer the vehicle retrograde all the way down like a Falcon 9 so it wouldn't do the belly flop. Its definitely possible to extend the mod to do the belly flop though. It would require a different 'plug-in' for the belly flop and different re-entry. Its quite a big change and the keeping the UI simple would be a challenge but I'd be up for trying it!

On v1.0.4 - The button to set the number of engines to use for the landing burn is not yet working. Fixes for this coming soon.

As well as fixing some of the problems with RO I've refactored the code quite a bit to make it cleaner and more reliable when saving BoosterGuidance settings with craft and in saved games. The major breakthrough is the use of FerramAerospaceResearch as I finally managed to link with it, even though I could not get the way the Trajectories mod does this to work for me so I tried an alternative method. I changed how the steering gains work which means much better landing accuracy in Realism Overhaul. A droneship landing with RO is definitely possible, though I've had trouble getting one out to sea in RO. (Any tips on this greatly appreciably since HyperEdit won't work ) I think RO users will enjoy the improvements. Its recommended to lower the aero-dynamic steer max angle to 1-5 degrees to avoid over-steering. In Stock you can have this at 30 degrees with no problems. I'm calling this a pre-release, though I think its good. I'd just like others to try it and report any bugs, as the code has changed quite a lot internally. Enable guidance with multiple vessels works well in Stock KSP but I've encountered some strange issues with this in RO which I don't yet understand. Download from here https://github.com/oyster-catcher/BoosterGuidance/releases Pre-release v1.0.4 - Will now always use FerramAerospaceResearch for aerodynamic calculations if its available. FAR seems to predict higher drag so landing burn alt often way lower - Better calculation of lift which means vessels should steer in the correct direction in landing burn (mostly!) - Considers aerodynamic lift in re-entry burn so if this lasts to the thick atmosphere steering will reverse - Only sign of gain (steer direction), rather than value is calculated by aerodynamic lift vs thrust. This means that gains are more stable between different vessel, RO, FAR, etc... and leads to less over-steering. Steer gain is still best set to max angle of 1-5 degrees with FAR though (for a typical Falcon-9 with gridfins) otherwise over-steers - Will cutout engine in final landing burn if too much minimum thrust. Only occurs in RO. This is hit and miss due to the time to throttle down engine - Major re-factoring to use BoosterGuidanceCore (the vessel part) as a central class. Code is cleaner and more reliable and fewer bugs for load/save/switch vessel - Target error is reported in km when >100km - Bug fixes to show/hide prediction/target - Size of prediction/target on map reduced - Click through prevention for pick target when under window - Many minor bug fixes - Stores engine ignite delay with BoosterGuidanceCore

@rhZhaoThanks for your videos and craft file, it really helps and inspires me to make the mod better. I've discovered that the use of FerramAerospaceResearch is the reason for the early landing burn. It seems it does make a quite significant difference to the terminal velocity and can drastically change the height of the landing burn. I've also improved the steering when under engine thrust and aerodynamic forces. Rather than approximating lift I can now calculate it directly. This should mean the rocket steers in the correct direction. In this video you can see the temporary button "use FAR" which can switch between stock and FAR aerodynamics, I do this a few times and you can see the landing burn altitude change. Also the rocket now steers correctly during the landing burn, whereas previously I had most success setting the steering to 0 degrees. However, the gains look very high so the rocket over-steers a lot but it does land within 9m of the target. Also note that the green line shows the orientation the rocket is trying to steer towards, which helps me to work out whether the rocket is trying to steer correctly but cannot. I hope to release this in a day or two. I will probably remove the "use FAR" toggle, and hopefully stop the over-steering in RO. I was able to download your craft and engine part, but it seems I'm missing quite a few required mods. I'll need to spend some more time testing your craft. But anyway, I think the improvements I'm making will make it work better for your craft too even if I can't test it. To answer your questions... I think the landing burn height becomes TOO HEAVY may be the result of undershooting, unless the vessel pitches over 45 degree then it will reach the target "theoretically". It is true that if the vessel is pitched over it would more thrust to slowdown since the vertical component is lower (it will compensate for this). But actually I don't think "too heavy" could result from that as the calculation made for the landing burn is simpler. Its almost certainly that the stock aero-dynamics give less drag hence too much thrust is estimated to be required. So the kOS calculate landing burn starts at 2914m, so it is far accurate. But I did not even know the different between retrograde and surfaceRetrograde in the kOS, so I was struggled to control the horizontal velocity. Thanks for showing me that. It looks like you are really close. For the landing burn I basically compute a steering vector = -vel_air + 20*up. Where -vel_air means surfaceRetrograde. up is a normalised vector pointing straight up. This calculation means any vertical and horizontal velocity is cancelled until the velocity becomes low and then just staying vertical begins to dominate. I also add a small correction so thrust can be used to steer towards the target point. But almost all steering is done aerodynamically rather than with the engines since this dominates when travelling > ~100 m/s. If you out grow kOS you might like to try kRPC which allows you to write much more complex code in Python and other languages and interface it to KSP.

Thanks so much for the video. Its very helpful. 1. I can see that the landing burn came in quite early as the thrust was as at minimum for the whole landing burn. Thats disappointing. The mod does try to give the latest suicide burn taking aero-braking into account. It does this by finding a maximum velocity for each altitude assuming no air resistance. It also calculates a falling trajectory with no engine thrust which includes air resistance. When these lines cross at the lowest possible altitude determines when the landing burn is started. This has mostly worked well in RO too, but I've tested very few vessels, and there is a safety margin added (perhaps this is too much). The landing burn altitude is extremely sensitive to the weight. If you use HyperEdit and tweak the fuel quantity you will see the landing burn altitude change radically. But anyway it looks quite badly wrong for your craft. Could you provide me a craft file please so I can test? Also the vessel was steered in the wrong direction. This is because the lift provided by the vessel is balanced against the thrust. In v1.0.3a I make an estimate of the lift as a proportion of the drag, and this was clearly wrong. BoosterGuidance assumed the engine thrust would dominate hence it steered to thrust towards the target, but the aero forces pulled it away from the target so the target error grew from 200m to 800m. This lift estimate will be gone in the next version, and the lift is calculated correctly (though only with Stock aerodynamics, not FAR so far). So the vessel should be steered towards the target even when some engine thrust is applied. 2. In principle I need the altitude of the suicide burn for the boostback since they depend on each other. In practice this is a rough estimate at boostback anyway due to fuel consumption being ignored and reducing the weight. So calculating the landing burn altitude later is probably fine and your suggestion will work quite well I think. It will avoid continuously calculating it which isn't ideal other but I didn't want to add extra parameters, it could probably calculate it once when starting the aero descent. I'm not too keen on the mod having to be told its in RO. I think the mod should be able to do without that or work it out. I'll consider this more. 3. Yes. I could probably drop that. Its used to find the height above ground. Its really a hangover from the HopperGuidance mod where targets can be at height rather than on the ground. It think it would be a good idea to drop it entirely as it is redundant. Also the target altitude can be wrong if you land somewhere nearby if there is a slope, it would be better to consider height about the terrain where you are. So the player doesn't need to consider the vessel height, the target is for the bottom of the vessel. Thanks. I'll definitely incorporate some of your suggestions.

Sorry. It looks like I forgot to upload the packaged zip file to GitHub for v1.0.3a, only the sources. Fixed up now, or use spacedock.info. You are looking for BoosterGuidance-v1.0.3a.zip

@TheKurganYes I should think I could add triggering by an action group. Perhaps then I could be triggered immediately on staging. I've seen a bit of this part of the API but I'm not familiar without so I'll need to work it out. The best bet is probably for me to look at the MechJeb code so thanks for that tip. Hmm. Sorry about the loss of your vessel. The mod does only simulate the Stock aerodynamics, as I couldn't get FAR to link, but that probably doesn't make a major difference. A few people have found the thrust comes in too late. There are quite a few tweak factors here since getting the height of the bottom of the vessel above the ground is rather tricky as the vessel origin moves around a bit but you need to know where the bottom of the vessel is relative to that, and this will change is landing gear is lowered. So please increase "touchdown margin" in the Advanced tab by 10m or so for now, the vessel will treat the ground is being higher which provides a safety margin.