The part you're really paying for is their collimator/optics. A constant current LED driver capable of the maximum 24W supplied by their power supply is easily under $10. The LEDs, heatsink, etc. would all be well under $50. 

The instructable assumes a paper type of fuse ("quickmatch") that a firecracker/bottle rocket would have. If you're trying to light the thicker, woven type  of fuse ("visco") you're going to need more oomph. For that you can coat it in a wet slurry of fine black powder mixed with ~10% dextrin. Dextrin would be corn starch cooked in an oven for an hour or so until it turns a dark yellow color.

There are quicker, easier and cheaper ways to break the law if that is what you want to do.

In the US it's legal to build and use a firearm. I'm assuming that regulations vary by state but where I am it's up to .6 inch caliber. Similarly it's legal to own and use explosives and the only license required is $100. There are plenty of rules regarding safely storing or transporting explosives, but nothing to really stop the average person to getting their hands on them. Aside from distributors of those explosives being very wary of who they're selling to.

A quick test would be to play with the number of averaged readings. If you're not getting less flicker with ~50 vs ~200 then it's not an issue with your averaging.

LED drivers are typically pretty flickery at very low output values. That would be my guess at your problem.

You'd need a triple pole switch (maybe just double?) rated ~30A to do that. Brushless motors have three leads, don't forget. It would also be impossible to put the retracts up while flying.

Sorry I'm not being very helpful here; While I'm familiar with these KK boards I can't think of any simple ways to implement what you're asking. Is it worth mentioning that these boards already have an arming sequence (requires the throttle at zero and rudder full right to arm)?

Use a PWM signal to the L298 instead of just a full on or off.

I'd suggest finding a small DC motor to use with the L298 and practice controlling the speed of the motor.

Triangulating position based on sound waves might work outdoors but not indoors. Too many echoes. Even outdoors you're going to get lost if the copter starts to yaw.

If you poke around on Youtube for examples of high precision positioning of quadcopters you'll see it's typically done with cameras mounted around the room. A computer then locates the copter in the camera picture and sends commands to the copter to adjust its position. As far as computer vision goes it's pretty simple to locate a copter if it's the only thing in the room with a bright (or specific color) light on it. The 3D locating using multiple cameras would be a bit of math though.

How many? Should be no problem with just 15. A Mega has plenty of input pins to handle just about any type of ultrasonic sensor. If you actually hit a limit you can always extend your number of input pins with something like a CD4051 multiplexer.

The 15 degree angle is the XY angle. Think of it like a cone (with a 15 degree tip) pointing at the sensor.

There's a difference between powering down a distance sensor and having a sensor that automatically wakes up when it senses a new distance. The former can be done by driving the sensor's power line with an IO pin as previously suggested; the latter... I don't know of any distance sensors with a wakeup function.

What you're describing isn't possible with a servo alone. While you can command a servo to move to a certain position you never know for sure if it is in that position -- there is no feedback. Well, you can make a guess based on how much power the servo is consuming (a servo consumes more power when it's trying to move rather than when it's stationary) but that's not very easy to implement.

 You need some other way of determining if your retracts are up/down; probably a small lever microswitch would work best.

You never have to drive it full current. You'll just get a little less torque, and even then you can get that torque back by just reducing the rotational speed of the stepper (the faster a stepper moves the less torque it has).

The number one rule is that if the motor gets too hot to hold your hand on then you're using too much current. Datasheets be damned; just start with a low current and keep slowly increasing it until it gets hot.

As far as drivers go you'll be much happier with a driver based on the A4988 (~$10 at pololu.com) rather than the L293/L298.

There are no interrupts being used in his code posted on GitHub.

You're going to need to be more clever in how you set up the scheduling of the various functions of your gimbal. It looks like in your code you considered implementing a timer interrupt and that is probably the best way to run things. Decide how quickly you want your gimbal motors to be able to turn and set up an interrupt so that those motors are always updated with that frequency. All of the other tasks of reading from the gyros/accelerometer or calculating the kalman filter, etc. should be considered secondary; you don't need to be doing everything on every loop().

IR receivers are kinda specific to their protocol. You have to match the modulation frequency (36KHz, 40KHz, 56KHz etc.) and length of the data packet.

Sony is 40KHz and RC-5/6 are 36KHz. That's probably going to be a show stopper right there.

Going by the second picture, your big round thingy with all the coils appears to be a combined ignition coil (the top coil) and two alternator stators (the bottom left and bottom right coils). The "condenser" (fancy name for high voltage capacitor), which works with the ignition coil, is sitting at the bottom between the two alternator stators. And then it looks like your points are jammed in there as well. A lot of stuff going on in there.

You'll want to splice into either line 4 or 5 for AC current; it shouldn't matter which because the AC waveform must be identical with either. If you're not using much amperage then dc42's suggestion of using four diodes would work, but personally I'd use a full rectifier in a GBPC case which would be easier to work with.

You should download Wireshark and learn how to use it. You need to be able to figure out what kind of proxy authentication is required; there are so many ways it can be done.

Microsoft has a good example using NTLM that shows the style of the messages going back and forth. While I'm not saying your proxy server is using NTLM (which is popular in the Microsoft world) it's assuredly using some type of encrypted authorization and that's going to prove a big stumbling block for you.