Virtual duellists could soon get a more realistic Game of Thrones battle experience after scientists programmed machine learning to improve enemies’ sword-fighting skills.
Researchers at the University of Bath with games studio Ninja Theory set out to make contests more realistic after discovering video game opponents were mostly coded to follow pre-recorded strikes, counter strikes and blocking actions.
They found these set moves were fine when playing using a console gamepad, but felt clunky and formulaic when the player was immersed in virtual reality, wearing a headset and using a handheld movement controller.
Computer scientists used actors wearing motion capture equipment to record an array of new sword-fighting moves that could be learned by a computer to upskill an avatar opponent, in a system called Touche.
It means games designers can also adjust the “skill and aggression” level of the avatar when fighting a human enemy.
Twelve volunteers told researchers that the data-driven opponents updated with intelligent moves resulted in more responsive gameplay and felt less “clumsy”.
Dr Christof Lutteroth, senior lecturer in computer science at the University of Bath, said: “With screen-based video games, a player presses ‘attack’ and their character displays a sequence of animations.
“But in a VR game, the player input is much harder to process.”
“Touche increases the realism of a sword fight by generating responsive animations against attacks and eliminating non-reactive behaviour from characters.
“Using our model, a game character can anticipate all possible fight situations and react to them, resulting in a more enjoyable and immersive game experience.”
