I'm developing a game for my bachelor's thesis and originally planned to use Polyanya, a published pathfinding algorithm with academic papers and an official implementation. I understand the algorithm's overall idea, but the reference implementation is more than 1,000 lines long, heavily optimized for competitions and benchmarks, and written in a different programming language. I underestimated how difficult it would be to study and reimplement properly.
The main focus of my thesis is the game itself and how users perceive and respond to it. I'm not planning to benchmark the pathfinding system or compare it with other algorithms. Would it be academically acceptable to implement a simpler, "naive" version that follows the algorithm's general concept while leaving out the advanced optimizations? If not, I could switch to a simpler algorithm such as A* or Theta*.
3 Answers
Be careful with the terminology. If you remove essential parts of Polyanya, you may no longer be implementing Polyanya in the technical sense; you may instead be implementing a simplified or inspired variant. That is not automatically bad, but your thesis should name it accurately and cite the original paper. A small experiment or basic validation showing that it works for your game would also strengthen the project, even if you are not doing a full benchmark.
This is ultimately a question for your supervisor or thesis guidelines, since acceptability depends on your program’s requirements. In general, a simplified implementation can be fine if you clearly explain which parts of the published algorithm you implemented, which optimizations you omitted, and how those omissions affect its behavior. You should not present it as a complete or state-of-the-art implementation if it is only a conceptual version.
A 1,000-line reference implementation is usually a particular engineering implementation, not the algorithm in its shortest possible form. Removing competition-oriented optimizations may be reasonable for a game prototype, especially if pathfinding is not the research focus. However, you still need enough understanding to implement the core correctly and describe the limitations honestly. If that is not realistic within your schedule, changing to A* or another simpler method is a perfectly valid way to reduce risk.
That makes sense. I can still change algorithms if my supervisor says a simplified Polyanya implementation would not be defensible, so I’ll ask before committing to it.
I hadn’t considered that the result might need to be described as a variant rather than Polyanya itself. I’ll make the distinction clear and discuss the scope with my supervisor.