85 79 69 bytes
$b to be an array of
@(1, 1), then enters a loop from
2 .. $n. Each iteration we tack onto
$b the latest calculation in the sequence with a simple
+= and the definition of the sequence. We then output the appropriate number from
$b (with a
-1 because arrays in PowerShell are zero-indexed). This works if
2 because both of those values are pre-populated into the lower indices of
$b from the start, so even if the loop tacks on junk, it's ignored anyway.
78 76 bytes
First time I've used the equivalent of a lambda as the answer, as usually an iterative solution is shorter (as you can see from all the nested parens).
But, in this case, the nested parens are almost duplicated in the iterative solution with the nested array calls, so the recursive solution is shorter. Nope, the iterative solution is indeed shorter (see above).
Call it via the execution-operator, like
&$a 20. Just a straight-up recursive call.