We began to explore different ideas of how reinforcement learning (RL) could apply on the game. We start ed out with a broad ideation brainstorming. Everyone talked about their favorite games or the games that they are interested in. We had come up with 3-5 broad ideas of how RL could influence the gameplay. However, after talking to Carl and Melanie about our ideas, we figured out it is hard to convey our ideas just through purely elaborating them. Carl and Melanie have suggested us to categorize the existing games into different aspects and have a spreadsheet so that we could present better. On the spreadsheet, we have seperated into core game mechanics/genre/Theming/Style/Fedlity/Emotional payoff/Takaways as below.
On Wednesday, we headed down to the beautiful Google campus to meet with Erin. Due to some circumstances, we are unable to have enough time to present some of our ideas, but we meet with Marc in the Stadia team working closely with Erin. He has some useful opinions for training agents throughout the gameplay. The reason why it is rarely seen in the exisiting game is not because of the computing cost but because of the undelightful expereince the player feels since it will not be optimal throughout the game. He also mentions that this is a challenging topic since as a game designer you will want to control the game, however, applying RL means that you are letting AI to decides the rules, and there is a interesting relationship between this. It is a fun talk with Marc and he has also given his opinions for some games we have proposed.
Afterwards, we eliminated the ideas of training agents through the game, and started to go deep into the application of trained agents in game. On Friday, we had another brainstorming session with all the teammates. We started again with a mind map of what element is related to our project.
Afterwards, we talked about each of our research from differnet game genres. We defines 1) what I like/dislike of the mechanics? and 2) Which part could RL specifically play the role? We want to be more detailed in what exactly RL plays in the part of the game. In the end of the brainstorming, we have categorized around 8-10 ideas and with its solution of using RL. We planned to organize the ideas and showed it to the faculty next Monday. We will be elaborating more next week.
On Thursaday, we had a lovely team dinner at the Liuyishou Hotpot at San Mateo. Every team memeber has a blast eating bunch of food!