Trajectory opimization: Fast preview algorithm

[Profiler]

Hello everyone,

I have been interested in trajectory optimization for a while now and I have read a few papers on that topic and bought the book "Spacecraft trajectory optimization" from Cambridge University Press and want to start programming with the goal to optimize a trajectory in a simplified solar system, not in real time but as close as possible, without the need of an exact solution but a preview of a good one.

The books and papers propose different approaches like

• Analytical solutions with the primer vector
• Direct transcription + nonlinear programming
• particle swarm algorithms
• evolutionary/genetic algorithms

From what I have understood so far genetic algorithms (GA's) are pretty good for finding global minima of the cost function even with a bad first guess but are pretty slow. I am looking for a solution that is as fast as possible but where finding the global minima is not required - a 'good' solution is fine. So I am still unsure of which approach to take; as mentioned before it should be consider a preview.
Particle swarm algorithms have been suggested to me as well and seem to be faster than GA's but converge slower towards the optimum, which is fine to me but I could not find any papers on the topic of speed comparisons regrading these optimal control problems.

If this is the wrong forum for questions like mine please move this post; I was unsure because this question is related to computer science an engineering as well.

Thanks for your help!

 

[BastianQuinn]

I am by no means an expert, but I figured a reply from someone similarly interested would be better than no reply at all.
Heuristics in algorithms is entirely about defining the metrics of "good enough," how often those metrics are calculated, and the "time" wasted making those calculations. While an effective heuristic algorithm will come up with an equitable answer faster than a precise algorithm, they take longer to code. In a sense, you're adding three new dimensions of complexity to test. The reason to turn to heuristics is when and only when a precise algorithm takes too long to complete, and you have plenty of time to experiment.

These types of problems are addressed in real time game AI programming, specifically in path-finding algorithms, so sources from that field may also be helpful.