Movement Prediction Algorithms for High Latency Games : A Testing Framework for 2D Racing Games

Sammanfattning: Context. In multiplayer games, player information takes time to reach other players because of network latency. This can cause inconsistencies because the actions from the other players are delayed. To increase consistency, movement prediction can be used to display other players closer to their actual position. Objectives. The goal was to compare different prediction methods and see how well they do in a 2D racing game. Methods. A testing framework was made to easily implement new methods and to get test results. Experiments were conducted to gather racing data from participants and was then used to analyze the performance of the methods offline. The distance error between the predicted position and the real position was used to measure the performance. Results. Out of the implemented algorithms, Input Prediction had the lowest average distance error at all latency. All methods tested did better than Dead Reckoning when above 600ms. Stored data algorithms did not do worse when predicting on a curvy part of the track unlike the other algorithms tested. Conclusions. Different methods are supported by different games and applications. Movement prediction should be tailored to its environment for best accuracy. Due to Input Predictions simple nature and its results here, it is a worthy contender as the go-to algorithm for games.