I need to simulate a conditional expectation of the form:
E (Xi^2 /(||X||^2 +h) | R^2 = ||X-T||^2 + ||U||^2 , Zi^2= (Xi - Ti)^2)
Where X~N(T,I_p), U ~N(0,I_n) and ||.|| being the euclidean norm, i found this result using the nadaraya-watson approximation (https://arxiv.org/abs/1306.1182), but i couldn't adapt it to the uni-variate case. i also found a function on R program called condexp {RGeos}, but i couldn't identify all its arguments

In the same frame work i need to simulate these expectations:
E (1/(||X||^2 + h)) , E (1/(||X||^2 + h)^2) (moment or order 2 of the previous) and E(X'T/(||X||^2 + h) ?
Any clue please ???