Probabilistic pointing target prediction via inverse optimal control