This paper presents a control strategy based on a extension of the generalized predictive controller (GPC), called bilateral GPC, for scaled teleoperation systems in the presence of communication delays, packet losses and slave force feedback. The basic idea behind this approach is to take into account the case where the reference trajectory is not a priori known in a generalized predictive controller (GPC). This difficulty is due to the slave force feedback which alters the master reference trajectory. Experimental results show the overall system stability with the proposed BGPC approach using a frequency-domain technique. Several configurations of scaling factors have been used; it is shown that the stability conditions strongly depends of the external environment. Another experimental result verify the robust performances of the approach in terms of tracking behavior with both strongly variations of time-delays and packet losses in the communication network. For the experimentations, the communication network is Internet using UDP protocol while the slave robot is a 6-DOF anthropomorphic robot with a force sensor. The presented performance results have been obtained between two remote sites of a distance ap 1000 km.