Here's the link to our demo of the project.
In order to play this game, you need to have Unreal Engine 4.2+ installed. Then clone this repository and start the project. Delete the contents in your plugins folder and replace with the contents found here, based on the version you have.
Then launch Steam and hit play on your engine. You can either join an instance or start one! Enjoy the game!
Our objective was to implement a 1v1 multiplayer first-person shooter that accurately analyzes inputs from two players and generates an acceptable gaming experience. Using Unreal Engine, we created a multiplayer level environment where two players can fire weapons at each other using FPS mechanics. The locomotion method used is the steer-to-center redirection algorithm with translational gain. Users may not leave their bounds when playing the game.
Our hypothesis is that a 1v1 multiplayer-shooter game that is set in a first-person POV would be more immersive than the same game played in a 3rd-person POV. Null Hypothesis: First Person and Third Person POV would have no difference. Alternate Hypothesis: Third Person will be more immersive. Furthermore, we wanted to survey players on the multiplayer experience.
- SAP ’16: Proceedings of the ACM Symposium on Applied Perception July 2016 Pages 113– https://doi.org/10.1145/2931002.2931018.
- N. C. Nilsson et al., ”15 Years of Research on Redirected Walking in Immersive Virtual Envi- ronments,” in IEEE Computer Graphics and Applications, vol. 38, no. 2, pp. 44-56, Mar./Apr. 2018.
- M. Azmandian, T. Grechkin and E. S. Rosenberg, ”An evaluation of strategies for two-user redirected walking in shared physical spaces,” 2017 IEEE Virtual Reality (VR), Los Angeles, CA, 2017, pp. 91-98.
- M. Usoh, E. Catena, S. Arman, and M. Slater, ”Using Presence Questionnaires in Reality,” PRESENCE: Virtual and Augmented Reality, vol. 9, no. 5, pp. 497-503, Oct. 2000.
We first began by constructing a steer to center algorithm. The code for this was done in a previous class assignment. The same build was then restructured: the camera was oriented to first person, character model was added, with access to weapons and a clean, minimalistic HUD. Our S2C algorithm essentially accelerated the rotation of the VR user’s space with respect to the user’s input rotation.
Once we implemented the steer-to-center lo- comotion algorithm locally, we began using RPCs for the server and client’s pawn move- ments. In our approach, we decided to go against an authoritative server and instead have the locally controlled pawn implement its own redirected walking. If the pawn is not attempting to escape its bounds, we send its updated transform to both the server and client. We were able to achieve replicated movements between the server and client with no apparent lag. For the shooting and health mechanics, we take the user’s left mouse input and notify the weapon blueprint that the user may start firing. Once the event reaches the weapon blueprint, we begin and end the user’s fire from the server with notify. Any time the user starts or stops firing locally, this notify instance alerts both the Network Authority and Remote Machine of the change in value.
Figure 1: This custom event replicates the client’s pawn movements to the server. When the
event is called with authority, the server pawn’s movements cast to the client.
We used Unreal Engine’s Advanced Session Plugin and Online Subsystem Steam to run multiplayer. The multiplayer is simply designed for one user to create the session and spawn the first person that is able to join successfully.
Figure 2:The host creates a public multiplayer Steam session and open the game level.
Figure 3:Another Steam player will join the created session.
Subjects (found through family and roommates) were assigned to either a third-person POV control group or the first-person POV in our 1v1 shooter game implemented in this study. In both groups, the players played until they defeated their opponent (reach HP of 0). Both groups then took a Slater-Usoh-Steed (SUS) Questionnaire to assess their immersion in the game. Due to Coronavirus, our study’s effectiveness was reduced. In a fully operational environment we would encourage finding more participants, implementation of actual VR devices and creating a more robust multiplayer network.
A Univariate ANOVA was run in SPSS using data to analyze whether there was a significant differ- ence between groups. Our ANOVA resulted in a p-value of 1, indicating that there was no difference between the groups. We fail to reject our Null Hypothesis. In retrospect this result does make sense to some extent. We had a limited number of participants who had happened answer identically in both groups. Furthermore, we realize that our users did not have VR headsets. They were simply playing the same game, but in different viewpoints. There wasn’t much of a difference they could particularly feel. Our users also felt that the multiplayer experience did not have much impact. We believe this was the case because it wasn’t a true VR experience.
In conclusion, there was no significant difference between 1st Person and 3rd Person POV in a multiplayer VR game. Future studies should use VR devices and have more participants. Addi- tionally, a comparison of games in similar realms and purposes may provide a more stark difference in results than with two differently styled games. Analysis of other variables, such as enjoyment, excitement, or physiological response, could be measured in the future. We believe that our analysis are important in the field of VR development. In an optimal setting our results could further solidify factors that effect immersion in a virtual world. And analysing these factors in different settings will help game-makers and creators push their creative limits. Furthermore, multiplayer in VR is fairly untested, our analysis in these areas could help transform how VR is played.
The main challenges we faced were replicating the pawns’ movements to be seen by both the server and client. By only running the S2C algorithm on the local controller, we were able to successfully replicate the transforms of both users’ pawns and VR tracking spaces. Another challenge we faced was implementing the UI and pawn components to successfully register between the server and client.
We would love to try our project with two users wearing their own VR headsets. We specifically de- signed the game to be compatible with virtual reality by using first-person mechanics. Future work would include the analysis of the user’s multiplayer experience in a virtual world using redirected walking. Some limitations we endured included Unreal Engine’s PlayerStart spawning implementa- tion. At times, we had issues with Unreal spawning the actors in the same starting location and thus allowing for one of the pawns to not spawn at all.
Special thanks to Technical Writer Michael Prinke of Unreal Engine for covering the fundamentals of the server-client model. We would also like to thank our instructor Nick Rewkowski for the steer-to-center redirection algorithm implementation. Assets from turbosquid.com.
Alex worked on the data collection and analysis. Austin implemented the RPCs and Steam Multi- player. Surya worked on S2C, game functionality, sound, and scene setup. Surya, Austin and Alex all worked together on the write-up and analyzing our results.