Writing an autopilot for a 2d game with newtonian physics

Question

The subject says it all. I am making a 2d space game with newtonian physics and I need pointers on how to write an autopilot for it. The requirements are best explained by an example.

There is a target object which has speed and position vectors and there is a spaceship that is controlled by the autopilot. This spaceship also has speed, position and maximum acceleration. The autopilot needs to control the ship so that it either collides with the target or intercepts the target with matching speed and position.

Could someone give me some pointers how to achieve this behavior or perhaps even a ready implementation? I am sure someone has written something like this before and there is no point in reinventing the wheel.

Answer 1

You will probably find what you're looking for here:

Steering Behaviors For Autonomous Characters

Maybe in this article:

Steering behaviors (full article)

Answer 2

Many of the currently available steer libraries are based on either OpenSteer or Programming AI by Example and have one fatal flaw: they rely on forward-facing physics which are out-of-the-box incompatible with physics libraries such as Box2d.

To illustrate the problem: in almost every single one of the steering examples, the direction of the agent, is the same as its force. This, however, doesn't hold true with physics libraries, where the force of the object can be different to the direction the object is actually facing.

Solving this can be fairly non-trivial. However,there is a video that seems to be doing something similar related to the book "The Nature Of Code". In the book, the implementation has been left as an excercise (6.21), unfortunately.

Trivial way to handle this is to cheat, and adjust your agents graphics heading independently of physics, but this seem like a sub-optimal solution to me. What about situations where you want to preserve non-forward facing physics of the physics library, but yet be able to steer the craft using autonomous agents?

Answer 3

In case you decided to write your own autopilot, I would always start by asking myself the question: What would I do in that situation? How would I plan my course, what maneuvers would I take? Surely you have mastered the task multiple times yourself. What steps did you take to arrive at your destination?

I know it still will be a pretty sophisticated autopilot. I would break it into smaller problems, i.e. a simple autopilot for making a course correction, one to kill rotation, one to come to a complete halt etc... Then later I would write more sophisticated ones that rely on the more simple ones - to automatically go from orbit of planet A to orbit of planet B, or one to intercept another ship etc...

I have thought about writing my own game for years now, and I think eventually I will have to face these problems myself...