In my research on video games path finding I'm using ant colony optimization, not only to find the shortest path, but also to add some unpredictability and adaptiveness to bots path finding. It works the way as players move in the map, they add some pheromone to the map, so it adds up a probability that bots choose path like players. I have sent the paper, but judges said you need a benchmark and distinguish to previous works. Can you tell me how can I benchmark this work?
The biggest issue here may be similarity to prior work. As for the benchmarks, benchmarks are a common means for comparing algorithms. What it means here would be to compare the end-result (your chosen goal) for each of the algorithms compared in a similar scenario, or a generated test-scenario. This will mean that you will have to utilize all chosen algorithms in your test-game-map software. To make it more original, shake it up a little with some creative/disruptive element which should lower the performance of the algorithm. Find some way to quantify how close the algorithms actually come to meeting the stated goal (or not meeting goal, like how far off they were) with and without the disruptive element.