Runway Project

Air combat in Kerbal Space Program

Doing a biology PhD is rewarding in many ways, but as I’ve written about before, it can often be stressful. Just today, I had to rush to the lab in the early morning to deal with a failing –80°C freezer.

To avoid a flameout, I like to do more relaxing activities. Over the last two years, I’ve been participating in a rather unique Kerbal Space Program¹ competition called Runway Project. Season 5 of this competition recently finished, and I came in first place! I’d like to take the opportunity to explain what Runway Project is all about, as well as the strategies I used that led to my victory.

So put on your helmets and fasten your seatbelts, we’re temporarily leaving biology for a lesson on aerodynamics.

What is Runway Project?

Runway Project is a combat tournament and art show based in the video game Kerbal Space Program. The goal of Runway Project is to design a plane that blows up all the other planes, and also looks cool while doing so.² Basically, participants submit craft files (which include parameters for an AI to pilot the craft) and art. Then, the craft all fight each other. Points are awarded both for combat performance and for artistic prowess (judged by the Youtuber Scott Manley, who hosts the competition).

Each season of Runway Project comprises ten rounds, each with a different theme. My first season competing was Season 4, and I came in 18^th place. In the latest season, I figured out some new craft design strategies, as well as greatly improving in art. And thanks to this, I got the highest overall score.³

Season 5 Craft Gallery

Before I get into the details of how I designed my aircraft, I’d first like to show them so you know what I’m talking about. As I mentioned above, each round had a unique theme, so the crafts are quite different. You’ll also notice that my art quality improves a lot over this season. This is partially from more experience. However, a scoring change was introduced after round 4 so that art was more heavily weighted, leading me to put more effort into it.

Round 1: “Top Gun” (craft must be replica of historic jet fighter)

Round 2: “Stretch Limos” (craft must be >3 times longer than wide)

Round 3: “Steel Rain” (craft must survive orbital reentry, then fight)

(Not pictured: heatshield and fairing)

Round 4: “Blood Gulch” (craft must be powered only by solid rocket boosters)

Solid rocket boosters, once lit, are always on at full thrust.

Bonus points to anyone who gets the joke.

Round 5: “Boonta Eve Classic” (a pod race, like from Star Wars)

Round 6: “Krimson Skies” (craft must be powered by piston engines)

Round 7: “Gunships” (mass must be >20 tons, bonus for >50 tons)

I took “gunships” rather literally.

Also, this round was a giant breakthrough for me in terms of art. I know I can’t compete with talented animators such as Concodroid and EclipticIndustries when it comes to visual renderings. But in round 7, I realized that a great soundtrack makes a great video, even if the visuals aren’t cinema-quality. So, for rounds 7, 9, and 10, I collaborated with my girlfriend (ULAhoop09) to perform silly songs to go with my craft. She played the ukulele and sang; I added a few notes on my clarinet, and put everything together using Audacity and iMovie.

Round 8: “Rocketplane Rumble” (liquid fuel rocket engines only)

Round 9: “Dreadnaughts 2 - BIGGER EDITION” (must carry 50 tons of fuel, turrets allowed, asteroid obstacles)

For this round, I took the same basic idea as Floaty McPlaneFace but made it much bigger and heavily armored. The result was something . . . unsinkable.⁴

Round 10: “Finale in the Void” (combat in space!)

Annoyingly, Youtube thinks this is a close enough musical match to David Bowie’s “Star Man” that they restrict the video embed, even though I only used originally performed audio.

What makes an aircraft good?

Art is fun, but the real reason I like Runway Project is because it poses a series of interesting aircraft design challenges. To perform well, your aircraft needs to have a few key attributes: firepower, thrust, maneuverability, resilience, and last but not least, artistic potential. However, these conflict with each other! Most notably, the heavier your craft is, the less maneuverable it will be — and guns, engines, fuel, and armor are all heavy.

So, how do you balance all these factors? Here are some tips based on what I’ve learned over the past season:

note: this will probably only be interesting for you if you’re into KSP and/or airplanes

General tips:

• Carefully consider the rules for each round and think about how they will affect the optimal design. For example, if the round allows turrets, maneuverability is less important (since turrets can swivel to shoot opponents) but tankiness is more important (since you can’t dodge as easily). This is why I added 110 tons of armor to the Titanic.

• Design your craft with at least some potential for art. It doesn’t have to be gorgeous, but the craft design should match with an artistic theme. If you don’t have an idea in mind before starting, start anyway, and as you build your craft, think, “What does this look like?” Because art points were weighted more heavily in the second half of this season, art was quite important for overall score.

  • For example, Floaty and Titanic were designed as boats from the beginning, but BOWMARC and Star Fish started out as attempts to build combat-capable craft.

• Don’t commit to a design too early. Build a few unique designs and do some preliminary testing to see which has the most potential.

• Extensively test your designs! Something that seems good in the hangar may not actually perform well. Try to maximize the information gained from each round of testing (test multiple variants against each other, not just one). The only Runway Project participant to score better than me in combat this season (Grizraz) is notorious for taking testing to an extreme. I think he has a server farm running 200 copies of Kerbal Space Program, or something like that.

Core components (guns, cockpit, engines):

• Your gun loadout is quite important, since the best way to score points is to kill opponents. For evaluating guns, it is helpful to equip a test plane with a variety of different guns, make all the variants fight each other, and look in the tournament logs to see what did the most damage. The legal loadouts vary quite a bit from round to round, so the best choice will vary. I performed well in round 10 because I used guns (in this case, lasers) that most other players overlooked.⁵

• Make sure to include enough ammunition for your guns. This will vary by round, but in general, enough for 30 seconds of continuous firing is usually good. Ammo explosions are quite bad, so make sure to place your ammo boxes somewhere well-protected. Somewhere close to the center of mass is best, so the center of mass doesn’t shift too much as your ammo gets depleted.

• Your cockpit choice is also important, since one of the easiest ways to die is having your cockpit destroyed. Fortunately, there are many to choose from. Pick one with a small size (minimizing mass and chances of it being hit), but high hitpoints for its size. Consider changing the hull material to steel if your mass budget allows. Placing your cockpit between other parts can reduce the number of hits it takes, but don’t violate the cockpit visibility rules.

• Once you’ve decided on guns and a cockpit, the next step is choosing engines. The best engines are the ones with the highest ratio of thrust to fueled weight (i.e., weight of engines plus fuel for combat, which accounts for different engine efficiencies). I usually have enough fuel for 3 minutes on full afterburner.⁶ If Saturn or Typhoon jet engines are allowed in the round, those are nearly always the best. Unlike with guns, it *is* possible to have too many engines. A final craft thrust-to-weight ratio of over 2 is usually overkill. For jet engines, make sure you have enough air intakes too – usually one more than strictly necessary, in case they are hit.

Maximizing maneuverability:

Maneuverability is usually the top factor in whether a plane is good. It is important both for aiming at other planes to shoot them, and also for evading enemies. In Runway Project, one-on-one fights between planes are nearly always won by the plane that can out-turn the other. To improve maneuverability:

• Have as much wing area as allowed (i.e. minimize wing loading). Some of my crafts in the first half of this season had a rather ludicrous amount of wing area,⁷ but the admins changed the rules and put a stop to this.

• Have your center of lift just barely aft of your center of mass. You want your craft to be aerodynamically stable, but only slightly. Use the CorrectCoL mod to check this. Make sure that depleting fuel and ammunition does not change your center of mass too much.

  

  This is the stability analysis for my Combat Limousine. For pitch stability you want the slopes to be negative, but very shallow near the origin.

• Put heavy parts (engines, cockpit, ammo boxes, fuel) close to the center of mass, to minimize the moment of inertia.

• Align the thrust vector from your engines with your center of mass, so your plane doesn’t have to counteract torque from off-center engines. You can use the Kerbal Engineer Redux mod to check this.

• Put control surfaces far from the center of mass and choose control surfaces with a fast response rate (>40°/sec). In general, control surfaces should be about 20% of your total wing area, but this can vary considerably based on the overall design.

• To minimize craft wobbling, don’t mix and match control surfaces with different response rates (this also applies to engine thrust vectoring, which it’s usually best to disable).

Resilience:

Ideally, your craft will be so maneuverable it won’t get hit.⁸ But because bad luck is always possible, it’s good to be able to take a few bullets and still fight.⁹ To make your craft more resilient:

• Minimize the harm of having parts destroyed:

  • Design your craft to have few “critical parts” where losing the part means losing the craft.

  • Strengthen any critical parts that you can’t eliminate (adding armor, or changing the hull material to titanium or steel)

  • Since parts that are attached to a destroyed part will fall off the craft, you shouldn’t attach important parts onto weak parts.

• Use fewer big parts rather than more smaller parts. Let’s say you have a 500 HP big wing, or 5 x 100 HP small wing segments. If you take three 100 HP hits, the big wing will still provide all of its lift, but you’ll lose most of the lift from the small wing segments.

  • The exception is control surfaces, since damage can cause them to become inoperable before they are destroyed.

• Keep in mind that although resilience is good, you don’t want to add mass and reduce your maneuverability.

Tuning the AI:

A good design is just the beginning — you will also need to get the AI to fly it well. The Runway Project plane AI is basically a PID controller with some extra features. The autotune function is quite useful here. To start, fix the “P” term at the default, set the “I” term to zero, and tune the “D” terms (using the 3-axis damping option). Then, decrease the learning rate and fine-tune all the parameters.

Besides the PID control, there are a few other things you should pay attention to. It usually helps to enable “don’t evade my target” and disable the “air to air extend” toggle; these changes will make your craft spend more time shooting and less time running away. The best way to not take damage is to quickly kill your target. Finally, speed, altitude, and terrain avoidance settings should be set to values appropriate for the round.

Final thoughts

Runway Project is, at least for me, super interesting and fun. If you’ve ever played Kerbal Space Program, I definitely recommend it, and I hope to see you participate next season! And although aerodynamics knowledge probably isn’t going to be practically useful for me, I think this xkcd definitely applies:

1. For those who don’t know, Kerbal Space Program is an aerospace simulator game, where players can build their own planes and spaceships out of individual parts like engines, wings, and fuel tanks. The base game lacks weapons, but of course there’s a mod to add them.

2. For those interested, the rules are here.

3. See here. I got 2^nd in combat points and 7^th in art points, but first overall.

4. This also went well with the theme of asteroid obstacles.

5. In fact, lasers were so overpowered relative to guns that the admins nerfed them the day before the competition, when they realized this. My craft still came in second in Round 10 (and the winner also used lasers).

6. The competitions are only 5 minutes long so this is usually enough, since the plane won’t use its afterburner continuously.

7. Notably, the Red Baron in round 3. But, fun fact: this is the same reason the real Red Baron used a triplane!

8. This was the idea behind the Japanese “Zero” fighter, which was very maneuverable but had no armor.

9. Of course, this doesn’t apply in non-combat rounds such as podraces.