You should use -nestedLoad param like this:
a = load 'input' using com.twitter.elephantbird.pig.load.JsonLoader('-nestedLoad') AS (json:map[]).
And then you use the following code:
b = FOREACH a GENERATE (json#'AD') as AD:bag{t:Tuple(m:map[])};
Then your json array become a bag datatype. You can flatten it to get tuple.
c = FOREACH b GENERATE FLATTEN(AD);
d = FOREACH c GENERATE AD::m#ID AS ID, AD::m#C1 AS C1, AD::m#C2 AS C2, AD::m#ST AS ST, AD::m#PO AS PO
At this time, you will get the tuple data type which the schema is (ID:bytearray, C)