FYI, in 2.42.9t, the rotational cost is acessed per radian of turn independent of mass and radius. Thus if the cost was set to 3, a bot performing 628 .aimsx store would be charged 3 * pi units of nrg.
At some point, we will want to move to a force based rotation paradym where mass and radius can be factored in.
Let me see if I can relate this.
For linear movement, f = m*a. For a given linear movement cost c (specifed by the user) the bot is charged N nrg units where N = c*f = c*m*a.
Thus, if a bot of mass 20 applies an aceleration of 10 (vector sum of all 4 movement sysvars) it gets charged N = c * 20 * 10. Note that bots only get charged for the acceleration actually applied I.e. a bot already at max velocity that trys to accelerate in the same direction it is alrady moving will not be charged anything. Note also there is a max charge of 100 nrg per cycle for linear acceleration.
It would be similar for rotaional cost except that the current convention doesn't use forces, it uses absolute turn requests, which is totally broken. I don;t think there is a sane way to include mass and radius using the durrent paradym, but I'm open to suggestions.
Once we do use forces, then we woud use the moment of interia I as Nums has pointed out in another thread. f = I * a / r where I = 2/5 * m * r^2 (the 2/5 assumes bots are uniform denisty spheres). For a given rotaional cost c', the bot is charged N' nrg units where N' = c' *f = c' * 2 / 5 * m * r * a. Note that in this model, bots would keep their rotaional inertia in the same way they keep linear inertia, subject to friction, etc.