Commit dc65f7f2 authored by Matteo De Carlo's avatar Matteo De Carlo

fix horizzontal line text overflow

parent 6af003c2
Pipeline #124 passed with stage
in 1 minute and 25 seconds
......@@ -274,12 +274,12 @@ It seems to be the case that, given more evaluations, the general population fit
Finally, in Figure \ref{fig:Results:phototaxis:single_chart5}, we notice that some runs could not deliver any results.
We recorded a video of one of the best results, with a fitness of \num{0.022745}, here: \href{https://www.youtube.com/watch?v=OsFUwZ5tfXc}{https://www.youtube.com/watch?v=OsFUwZ5tfXc}.
We recorded a video\footnote{Video can be found here: \href{https://www.youtube.com/watch?v=OsFUwZ5tfXc}{https://www.youtube.com/watch?v=OsFUwZ5tfXc}} of one of the best results, with a fitness of \num{0.022745}.
We meshed three replays together: CENTER target, MORELEFT target and MORERIGHT target.
Skipping to 1 minute 30 seconds of the replay, it's evident that one of the best solutions was not even able to learn to walk to three of the five directions; in fact, the run configured to go right has gone all the way left, going to the exact opposite direction.
The high instability of the results could be fixed with a larger population and an increased number of generations.
This approach was not taken into consideration because the runs already took around 2 hours of simulated time or half a week of real time.
This approach was not taken into consideration because the runs already took \linebreak around two hours of simulated time or half a week of real time.
Increasing the number of evaluations would not fit into the creation of the \tol model.
Instead, we looked into improving the algorithm and investigated what slowed down the learning so much.
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment