Abstract: In this paper, we propose a new attitude dead reckoning paradigm in collaborative virtual environment (CVE). We derive a general polynomial construction scheme of attitude trajectory that uses a number of previous packets in order to extrapolate the future trajectory of objects by quaternion representation. The scheme allows consecutive attitudes received from the network to propagate in a smooth manner. This cumulative trajectory construction scheme helps developing our adaptive prediction and convergence mechanism of the overall estimation paradigm. In contrast with the traditional dead reckoning, our approach can produce more pleasant shape of trajectory in many cases meanwhile eliminate the network transmission of derivative information. The proposed approach provides a robust dynamic management of shared attitude of objects and camera views while at the same time reduce the bandwidth consumption and ameliorate the network lag. Experiments are carried out to illustrate the effectiveness of our proposed algorithm.