Cycling Game (Mini Neural Net Demo) (www.doc.ic.ac.uk)

A new mini neural net demo for a cycling game has been unveiled, showcasing how different neural networks can control virtual riders in a race. Users can select a rider, observe their neural controller, and trigger evolutionary changes that influence the riders' performance metrics based on various percepts, such as speed, power, and battery levels. Each rider, despite having identical physical traits, possesses unique brains that evolve over multiple races, allowing for strategic variations in their approach, such as optimizing power output or utilizing drafting techniques. This demo is significant for the AI/ML community as it illustrates practical applications of evolutionary algorithms and neural networks in simulating real-world scenarios, like cycling dynamics. Users can experiment with a variety of parameters, including rider behaviors that emerge from genetic mutations of neural network weights to refine speed and strategy over generations. The simulation’s detailed physics, incorporating slope, air resistance, and physiological limits, offers a rich environment for studying adaptive behaviors in competitive settings. This engaging visualization not only serves as an educational tool but also highlights the potential of neural networks in optimizing performance in complex systems.