Rotate to face a direction, by applying angular acceleration?

Besides the current heading, you probably want to take the current angular velocity into account. You want to reach the desired heading with an angular velocity of 0. The problem is similar to the "arrival" steering behavior. I would start by looking at some implementation of that and adapt it.

Yeah, I need to do that. So I've worked out that angular rotation is represented by the axis of rotation, and the magnitude is the amount of rotation. This means that I can use the cross-product between the two vectors to work out the torque I need to apply.

This seems to work for small rotations, but the moment my ship crashes into something the whole thing goes haywire. I'm going to debug some more but if using the cross-product is a bad idea for some reason please let me know! Are there edge cases when the difference between the two vectors is large?

No, I wouldn't say there are edge cases, but some naive implementation might leave you with a tiny torque in situations where you are facing close to the opposite of where you want. You can use the cross product to determine which axis to rotate, but then use something like the arccosine of the dot product to measure the angle.

The exact torque is

torque = -(1 / timestep^2 * angle * stiffnes + 1/timestep * angular velocity * damping) * moment of inertia

where stiffnes and damping is an arbitrary number between 0 and 1, defining how fast you want it done. If you would like the ship to align in one loop, just set both to 1.

Cheers,

Mike

@Mike, just a few questions about that equation:

• Is "timestep" the fractional 1/60.0 or the number of steps in a second (e.g. 60)?
• Is "angle", the result of acos(desired.dot(current)) ?
• Is "angular velocity" the current angular velocity of the ship?
• What's "moment of inertia"?

Sorry if I'm being dumb :)

moment of inertia is a physics term, the rest I'm not sure.

(variables with descriptive names are nice, if one was called currentAngle or desiredAngle it gets more intuitive, but I digress)

Hi Kazade,

Timestep is the delta time of your game, for instance 1/60.

I supose you could use acos(desired.dot(current)), but it is not safe, since it could in some situations force the ship to rotate the longer way around, rather than the shorter. The unambigious and 100% safe way to do it would be atan2( desired.perpdot(current), desired.dot(current) ).

Perpdot is the perpendicular dot product or "2D cross product". The "regular" dot product between two normalized vectors return cosine to the angle between them, and perpdot returns sine to the angle.

http://mathworld.wolfram.com/PerpDotProduct.html

Yes, the angular velocity is the current one.

Moment of inertia is "rotational mass". Just like mass, you can in principle give it any value you want, but if you set it too high or too low, the physics will look weird.

Check out these wikipedia articles to understand the concepts of rotation. You'll need it if you want to play with physics. :-)

http://en.wikipedia.org/wiki/Moment_of_inertia

http://en.wikipedia.org/wiki/Torque

http://en.wikipedia.org/wiki/Angular_acceleration

http://en.wikipedia.org/wiki/Angular_velocity

Cheers,

Mike