I just realized... how silly. It's too TALL. If I want a stabilized motion I should make it wider and more squat (real physics remember? I forgot). I was looking over all the small Newton projects.
I didn't realize but I can also use keys to change directions of a Newton object. That doesn't work for newton chains like the rag doll thing I'm working on unfortunately but I can use it for this. Once I have the "random" motion the way I like it I just key frame the "hover craft" or whatever it is to change direction. I have to look at the sample project to see how that works. It's the one where the object is sort of rolling through an obstacle course.
A newton object must have at lest one patch to have any substance. However you can use "invisible" groups.
Yes it can be hit by other objects. So you could "throw" something at it and cause it to "crash". I could have one of the engines get hit by a missle and cut out... then it gets unstable and crashes.
In order to make the force engines align properly using only a single translate to constraint I had to create an action object model. This model is 3 section cylinder with the top and bottom cross splines deleted so there are no patches. I then use an action to add a force to this model so I can offset the pivot. The problem with using constraints on forces for this thing is that a force has the pivot at the bottom or origin of the force. So I have trouble using a translate to constraint to point UP at the model (forces have to be upside down. Not like a rocket). The translate offset causes the movement of the force to not stay in alignment. For instance if the force is directly below the rocket and the rocket goes left and down the force goes right and up in relation to it due to the offset.
I used this "force holder" model in the action so I can "see" the force. A force as an action object isn't visible in the chor and you have to dig down in the PWS to access it. I use a bone for the force holder model to scale the points to match the size of the force in the action. I move the center "ring" spline of the model to show where the fall off is located. I think I could do this with expressions! I've started fiddling with them a while back and realize now they are pretty freaking awesome. I'm also wondering now if expressions can't be used for my rag doll rig.
So the model is centered but the force is below the center of the model. I adjust the position of the force in the action to get the proper lift. Now when I constrain that model with this action to the rocket it isn't "out of whack". The translate offset is correct.
Now... there is one thing I could have done to make this "easier". I have to play around with it some more because I got some odd results. There is an option for a force with Newton dynamics called "plane". This makes a cylinder or box force push straight up in the direction of the pivot angle. So... there would be no side to side motion at all. The way it is now the cylinder force pushes from the center of the model mass to the center of the force. Explained better if you watch the sample video on the newton page. So if an object is on the outer edge of a force it will move at an angle from the center of the object to the center of the force. With "use as plane" turned on it moves straight up from the center of the model. I asume that having "plane" turned OFF is more "real".
p.s. This is one reason I wish HA:MR was still being developed. Imagine being able to create "3D web games" using raw AM files with Newton and HA:MR! Newton physics is intended for game physics after all. If expressions were "extended" you could almost create "game AI"... rudimentary of course.