Weekly work

Author: dduro2020

docs/_posts/2025-02-24-Week_17.md

@@ -33,7 +33,7 @@ To calculate the target dynamically what is done is to save the position of the
 
 Before we could do all this we had to fix the way of using the coordinates in SMARTS, as we were interested in a reference system relative to the vehicle regardless of the orientation of the road, and as SMARTS uses this reference system, it had to be rotated to make it easy to interpret:
 
-![](SR_Smarts.png)
+![](SR_Smarts.jpg)
 
 ## Starting DQN
 
@@ -152,4 +152,4 @@ In the last tests it seemed to ‘park’ without perfect orientation and speed,
 
 That the loss does not stabilise after so many tests and parameter changes may mean that the rewards are too variable. It was something I hadn't given much thought to until now, one way to fix it and add very little variability is to normalise the rewards, but it doesn't seem to have solved the problem:
 
-![](plots_dqn/p_norm_.png)
+![](plots_dqn/p_norm.png)