Taipei Friends | Polar Quest
The Challenge | Polar Quest
One hundred ways to survive in Antarctic
A man find all his ways to survive after an Shipwreck in Antarctic.
Background:
Antarctica is a land that is relatively unfamiliar to people, so we create a game that can help the players to understand more about it. We set the character in our game to try to survive in Antarctica. In the process, we present a lot of information and knowledge to the player and wish to make the complicated datum to be more acceptable to the public.
Approach:
We utilize deep learning to train a model based on NASA's datum to construct the whole environment in the game. For example, we use the data of weather and predict the trend of the variation of the weather conditions in Antarctica. Moreover, we can also relate the time, latitude, and longitude to the environmental parameters (temperature, pressure, albedo,etc.). There is also a setting of a time machine, which can allowed the players to see the datum years ago. All the datum in different periods can be shown and produces by this means.
Training method:
DNN hidden layer : 2
Loss : L2 (mean square error)
Optimizer: Adam
TOOL:
Pygame 1.9.3
Python 3.6.5
Tensorflow 1.5.0
Content:
The players can know some information in Antarctica when the events in the game happen. There are some example that players may learn in the following:
- the species of penguins
- some symptoms or diseases people may encounter in Antarctic and the way to deal with them
Github:
SpaceApps is a NASA incubator innovation program.