EA Finding Way To Create Realistic Game Worlds Using Map Data
Electronic Arts | Source: Electronic Arts
Electronic Arts is potentially working on a system to streamline the creation of realistic virtual worlds in video games by using a neural network to procedurally generate positions of map items, such as buildings and other structures, for placement on a virtual map.
The system makes it convenient for video game developers to create realistic virtual worlds in video games by iterating through different designs of virtual worlds in real time without any manual configuration of the map items.
The system also considers the various features on the virtual map, such as roads, waterways, and vegetation, to generate map items realistically.
The system is expected to streamline the video game development process by making it more flexible and accessible for video game developers, not to mention fast.
Electronic Arts is one of the foremost players in the video game industry, with a long history of creating and publishing popular games for various platforms, including PC, consoles, and mobile devices, since its inception in 1982. In addition to creating some of the most widely recognised video game franchises of all time, like FIFA, The Sims, Battlefield, and Mass Effect, the company also has a substantial presence in the e-sports industry, hosting and sponsoring events for popular video game franchises, such as FIFA and Apex Legends. With a diverse portfolio of video games and a firm reputation in the video game community, Electronic Arts continues not only to dominate but also to revolutionise the video game industry.
In August 2021, Electronic Arts announced the Accessibility Patent Pledge, which allows all developers to utilise the company’s accessibility-focused technologies. Since the pledge was made, the company has released many patents focused on enhancing the accessibility and playability of its video games. One such patent awarded to the company expects to streamline the creation of realistic virtual worlds in video games through a system for procedurally generating positions of map items, such as buildings and other structures, for placement on a virtual map using only map data provided to a neural network.
Earlier today, we came across a recently awarded patent to Electronic Arts, titled “Generating Positions of Map Items for Placement on a Virtual Map,” which was filed in September under the name of Electronic Arts Inc. The patent, which was published only a couple of days ago, describes a system for video game developers to quickly and easily create realistic virtual worlds in video games by iterating through different designs of virtual worlds in real time and conveniently controlling the configuration of map items generated on the virtual map without having to tune the many parameters manually.
“This specification describes a system for generating positions of map items such as buildings, for placement on a virtual map. The system comprises: at least one processor; and a non-transitory computer-readable medium including executable instructions that when executed by the at least one processor cause the at least one processor to perform at least the following operations: receiving an input at a generator neural network trained for generating map item positions; generating, with the generator neural network, a probability of placing a map item for each subregion of a plurality of subregions of the region of the virtual map; and generating position data of map items for placement on the virtual map using the probability for each subregion,” reads the abstract for the patent. “The input to the generator neural network comprises: map data comprising one or more channels of position information for at least a region of the virtual map, said one or more channels including at least one channel comprising road position information for the region; and a latent vector encoding a selection of a placement configuration.”
Conventionally, video game developers have used existing real-world maps to generate virtual worlds. However, these maps may be incomplete, lacking information about buildings and other structures. Electronic Arts aspire to solve this issue by allowing video game developers to conveniently control the configuration of map items generated on the virtual map without manually tuning the many parameters required by conventional methods. In addition, the system takes into account the various features on the map, such as roads, waterways, and vegetation, to generate map items realistically. As described in the patent, the system will supposedly use a neural network trained for “use in generating positions of map items for placement on a virtual map.” The training instances, which may include “(i) map data, comprising one or more channels of position information for a region, (ii) a latent vector, and (iii) at least one ground truth placement of map items,” will determine, based on probability, where a particular map item needs to be placed within a virtual world based on its virtual map.
According to the patent’s claims, the system, which the patent refers to as the “generator neural network,” is trained using a set of examples, which include a map of an area and information about where certain structures should be placed on that map (called the “ground truth placement”). The system uses this information to try to place structures on the virtual map in a way that is similar to the “ground truth placement.” Likewise, another system, which the patent refers to as the “discriminator neural network,” is used to evaluate the placements of the generator neural network to ensure the system learns to place structures correctly. The discriminator neural network evaluates the virtual map and the system’s placements and tries to determine whether the placements were made by a computer automatically or a human manually. Based on this evaluation, the system’s ability to place structures correctly is improved.
In some cases, the system described in the patent includes additional information in the training examples referred to as a “latent vector.” This latent vector is a set of data that encodes certain characteristics or properties of the structures being placed on the virtual map. During training, the latent vector of a training example is updated based on how well the system places structures on the virtual map. This helps the system learn to place structures in a way that is consistent with the latent vector. The training examples may also include “cross training examples,” which are created by combining information from two different training examples. This helps the system learn to place structures in a way that is consistent with different latent vectors. The latent vector of the second example can also be updated based on how well the system places structures on the virtual map.
In addition to using a discriminator neural network to evaluate the system’s placements, the system’s placements may also be compared to the ground truth placements to see how accurate they are — a process that the patent refers to as the “reconstruction loss.” The system may then use this reconstruction loss to adjust its ability to place structures correctly. The reconstruction loss may be calculated using the Laplacian pyramid method, which involves breaking the map and placements into smaller pieces and comparing them. In some cases, the system may also use a second discriminator neural network to evaluate the placements of the map items on the virtual map.
It’s unclear when or if Electronic Arts plans to implement this system into its video game development processes, but the patent does suggest that the company is exploring innovative ways of making video game development more flexible and accessible, not to mention fast. The system’s ability to control the configuration of map items and take into account the features of the virtual map without any manual configuration can potentially make it a worthwhile addition to the video game development process. The patent represents a significant advancement in video game development and will likely be widely adopted in the industry if it does come to fruition.
What do you think about this? Do tell us your opinions in the comments below!