We know that EA Sports College Football will use the Madden engines and technology. We also know that EA Sports developed a neural network locomotion engine as recently as 2020. Will we see an entirely new locomotion engine in Madden and EA Sports College Football on next-generation consoles? Perhaps. But it’s instructive to look at how EA has managed its portfolio of games in the past and what it can do with a new prototype engine.
Evidence shows EA’s track record. It develops engines and implements them across its entire portfolio. The Frostbite Engine was purchased and implemented in many EA titles, some of which were not a great fit. This decision was a setback. Instead of advancing technology for its sports games, EA had to develop foundational components for Frostbite to deliver the same sports games on the Frostbite engine. Accordingly, gamers have found EA behind schedule on revamping its animations, locomotion, and A.I. logic.
Fortunately, we have a potential preview of at least one of EA Sports’ development projects. It is a locomotion engine driven by neural-network, A.I. technology. According to a 2020 paper published on locomotion by EA’s Scientist Sebastian Starke, EA Redwood City is potentially constructing a “motion prediction network” engine for its sports titles.
The future of EA Sports
In collaboration with EA, Starke was developing this prototype engine specific for sports gaming, as demonstrated by Starke’s published research. The prototype is focused on basketball moves; it may represent a preview of NBA Live, and the project demonstrates an engine that can apply to many sports and genres. In light of EA’s track record, we can reason that this technology, or its premise, is in development and a candidate to be used on next-generation sports titles.
Per Starke, the engine uses Unity3d, TensorFlow, and Pytorch environments. Whether or not EA will integrate the prototype engine with Frostbite remains to be seen, and it is possible for the locomotion engine to live within the Frostbite environment. The two are not necessarily mutually exclusive.
New Animation Control
Organic controls and animation technology have lagged as the gaming landscape becomes increasingly performance-based. It follows that game engine limitations are often realized shortly after each new sports title release. As a result, sports gamers become mass beta testers for titles such as 2K, Madden, and others at title launch. Advances in player control are needed to keep up with gamers’ speed and skill-level.
A key component to animation-control is “a character’s relation to its environment,” notes Starke. This dynamic “largely defines the fidelity of its motion.” Such spatial awareness is critical for Madden and the highly anticipated EA Sports College football game in development. Defensive A.I. has been extremely problematic on the Playstation 5 and Xbox Series X versions of Madden 21. The defensive players stand to benefit from improved spatial awareness.
The defensive A.I. must improve its play recognition, and improved intelligence will certainly help. The line-play also stands to gain the from smarter animations. Line-play has seen the least amount of advances over the last console generation – aging animation logic has played a role in the run-game’s struggles. The line interactions that may look organic are likely taking too many frames to play out, hindering the line-play’s overall fluidity.
According to Starke, his prototype engine factors in several variables: the character’s locomotion state, environmental conditions, opponent information, and the contact points for a character (two feet, two hands). His research project segments motions into smaller, independent subsets (limbs) that work in coordination.
This “subset” methodology, in theory, should optimize frame-rate, and it could help resolve unresponsive control that we experience where sharp movements look and feel unnatural. The prototype engine blends multiple algorithms from each “subset” or limb, producing full player motion control.
The End of Scripting?
Animation technology that segments animations into subsets for each limb, conducting micro-animations at each contact point, sounds quite complicated. Continuing, the engine described above also sounds primed to deliver smoother running motions.
Foot-planting and cutting is everything in football games, and EA Sports has generally struggled with sharp cuts and momentum. Based on the demo, agile cuts are a point of emphasis for the EA prototype. Per the research article, animation branching may be computed by the engine much faster than in the past.
The neural network technology that EA Sports cites in its study shows promise in A.I.-driven animation technology. EA’s prototype projects to increase the number of movement combinations by assimilating current and new animation libraries. Such an engine can learn from the datasets of motion-captured animation libraries through “deep learning” programming.
On paper, the technology looks exciting, and it may lead to more natural player motion. Even if EA does not deploy this engine for Madden and College Football, the principal idea is there: EA is developing animation engines that can execute organic motion with less “scripted” outcomes.