On the Impact of Delay on Real-Time Multiplayer Games
Authors: Lothar Pantel, Lars C. Wolf
Presented by: Bryan Wong
Outline
Introduction Multiplayer Games Experimental Evaluation Related Work Conclusions
Introduction
Major problem of networked multiplayer games is caused by transmission delay
A primary goal is to provide a global consistent state of the gameOne approach is to introduce a local
presentation delay
Introduction
Simulator-like games such as flight simulators, racing and sports games are most critical because they require smooth movements of the objects
Multiplayer Games
Architectures:Server based with a dynamically
based or central serverDistributed scheme
Influence of Basic System Parameters
Designer must decide which messages are transmitted reliably/unreliably
Delay
The time between the generation of an event and the resulting update of the game state
Important issue of how the system handles old state information and how large the difference between the real and expected positions of the players in a game may be due to the delay of messages
Experiments with Commercial Games
Studied the delay-handling approaches used by two commercial games in two player modeRe-VoltNeed-for-Speed
Examine how commercial game developers handle the delay problem
Start Release
Direct play assigns the role of coordinator to one of the two players
This contains the release of the start Use the start countdown and start
sound to detect correct synchronization
Start Process
As soon as the vehicles begin to move, the first deviations in the positions of the cars can occur
Identical vehicles, motors, etc. are selected to avoid acceleration differences
Expected that both cars remain head to head
Simulated Start
Cars are stopped in the course of the race and are aligned at the beginning of a straight path
Both cars are accelerated simultaneously by one person using both keyboards
Cars should remain side by side
Driving with Constant Speed
Vehicles drive with same constant speed side by side
An identical screen situation should be presented
Occurs only if position consistency is maintained during the game
Collision Treatment
Collision of game objects is detected through a polygon-collision
Only a position difference of 0 of opposing vehicles will cause an identical, local evaluation of the polygon collision
If position consistency is not sufficient, contradictory situations can occur
Observed Results
Games provided very limited handling of delay
Both provided a synchronized start release Start process was not handled correctly –
the local car was always in leadership Both games failed tests 3,4 Neither game provided satisfying treatment
of collisions
Experimental Evaluation of the Impact of Delay Inconsistent state brings several problems
and can be confusing to the players One approach to enhance consistency is to
delay the presentation of the actual game state
How much delay is acceptable for users in real time games?
Delay includes all communication, processing and buffering costs
Principle Setup
Performed steps:
Implementation
Implemented in an RC-Car simulation Driver does not use a cockpit view but
a higher view allowing the user to view the whole area
Virtual RC Racing implements the static delay scheme
The feeling how the car reacts is influenced by the delay between steering and reaction
Experiments
Purpose was to determine acceptable presentation delay
12 candidates performed experiments For each of the following graphs, three
players were selected who were representatives for the following categories: Beginner, Average and Excellent driver
Static presentation delay was increased in 50ms steps up to 500ms
Experiments
Avg time per round in dependence on delayIt becomes more difficult to control the
car, requiring the driver to slow down Best time per round in dependence on
delayReflects the best possible value for
the driver, under the given conditions Frequency of Leaving the Course
Results
Results
Results
Results
Results
View Dependance
Delay value that can be tolerated by the participant depends on the camera perspective
A cockpit view allows at least an additional 50ms of delay to be tolerated
Related Work
MiMaze – distributed 3D Pacman game
Consistency of the is improved by delaying the presentation of the events using buckets
The whole game is 100ms in the past Players with >100ms delay cannot
participate
Conclusions
Delay of 50ms is uncritical for a racing game Delay of >100ms should be avoided Consider the racing game as a worst-case system For other games (first person shooter,…) a delay of
100ms may be acceptable For comparison, lip synchronization requirements for
audio and video streams are [-80ms,80ms] Static delay schemes are not well suited for real-time
multiplayer games