Ending study

This is a game played recently I annotated. the egregious blunders aside, it was a very interesting ending/ Queenless Middlegame (R+R+B+N+5 vs R+R+B+N+5). 

I follwed some principles incorrectly before looking for alternative moves. My 17 for black is an example of seeing a natural move to protect and control the d-file. However I had a strong double attack with 17... Qe4. Secondly The choice on move 24 between b6/ g5 was difficult. I thought stopping the knight's activity more imperative, than restricting the scope of the bishop. 

After Blacks 30...Nb6 The board looks sterile, clogged center unable to pass for the kings without a run for opposite wings.

Move 32, From a Natural looking move 32. Nc6 the long mostly forced sub variation i gave went right to a draw, unable to convert.

The final critical Junture came after 36. Kxd2, here I considered two central ideas in endings, mobilization of the king and piece activity vs inactivity. My personal preference being activity I chose that path without fully considering the king initially. I wanted to lock his knight out and gain the advantage of 2 active pieces vs 1. We agreed on an earlier draw but a justified one.