Martin Rhodes and Simon Coupland from De Montfort University, England, developed the “Pro Evolutionary Soccer” AI. In the competition, they presented a 15-minute demonstration of free kicks taken within a soccer simulation. Several of the examples were based on real free kicks, including one by LA Galaxy and England free kick specialist, David Beckham.
Whilst players like Beckham use years of experience to judge the spin and power to place on a ball, the AI creates many solutions and tests them to see which are most likely to give a goal.
The AI rejects obvious failures, slightly alters the best and then uses these altered solutions as input for the the next generation of calculations. This way the system “evolves” a solution. A process known as an evolutionary algorithm.
The developers see an application in video games. They say:
The current generation of football video games suffers from a predictable, repetitive style of gameplay. Pro-Evolutionary soccer uses a stochastic approach to give varied, organic style of play in the form of free-kick taking. Given a physics model and the positions of the opposing team, our system finds a near optimal free kick from all possible free kicks giving a varied and challenging gameplay.
Pro Evolutionary Soccer faced some tough competition. Other competitors included a robot that can locate and then follow a person in a complex environment and an AI avatar called Halo, who lives in Second Life.
- Have you met Halo yet?
- Did you realize you were interacting with an AI?
Image Credit: University of Leeds