Evaluation of the Motion-Aware Adaptive Dead Reckoning Technique under Different Network Latencies Applied in Multiplayer Games

Abstract

Multiplayer games hold virtual worlds which connect dozens of players in the same session, in which these human players experience network latency in networked interactions negatively. A number of predictive techniques were developed to deliver the best experience for the players. With the growing numbers of players and the complexity of the worlds, the limitations of such techniques, like network delay, consistency, responsivity and bandwidth cost, become evident. The main technique used nowadays is called Dead Reckoning (DR) and was firstly presented decades ago. Based on it a variety of authors proposed improvements to the prediction method, culminating on Kharitonov’s proposal of the Motion-Aware Adaptive Dead Reckoning (MAADR) technique. The author’s evaluation did not consider latency, an important factor that affects the consistency (i.e. veracity) of the information and the Quality of Experience for the players. The proposal of this paper is to evaluate both the MAADR and DR prediction techniques in 4 situations ranging from simple to complex movement patterns, with 0 to 300ms of network delay. The results show that the MAADR performance is superior when compared to the classic algorithm for each situation and each different latency evaluated. It shows that the classic DR presents a large decay in precision within medium and high latencies and it is not ideal in situations with great intolerances to network delay. Because of the obtained results, it is recommended the usage of the MAADR technique when medium and high latencies are expected or when there is an intolerance for the loss of the Quality of Experience.