Dino is a simple implementation of Google Chrome T-Rex. The goal is to control tyrannosaurus so he won't collide with obstacles — cactuses and birds — for as long as possible, while it slowly accelerates.
You can play it just by running play_dino.py (by executing python3 [path]/play_dino.py). If you run it with no commands you can control T-Rex by yourself with keyboard arrows. Otherwise you can run the script with option -h to see all available options. Particularly you should be interested in option -a Agent which allows you to specify agent to control T-Rex.
There are two ways how to specify options for the script. You can either run script with options specified in terminal e.g.
python3 play_dino.py -a=MyAgent --seed=0 --vis_rect -d
or alternatively you can specify options as a parameter of method main at the end of play_dino.py and executing script (with all options from terminal taking priority):
main(["-a=MyAgent", "--se=1", "--vis", "--deb"])
...
python3 play_dino.py --seed=0
Note that argparse allows long options to be abbreviated if it is unambiguous.
Before you start to implement your agent, you should look into script game/dino.py, that is complete implementation of the game. As you will find, there are not many getters or methods for interacting with the game state, however your agent will get reference to the current game state so you gain complete access to all components. You should use it as read-only variable but during debugging your agent you may modify the game state as you like. Note that ReCodEx tests won't allow you to do any modifications.
The game itself is just a plane with rectangles. All obstacles are stored in game.obstacles and they are represented by single rectangle that can be accessed by obstacle.rect. T-Rex consists of two rectangles — head and body — and you can access them by game.dino.get_rects(). Your goal is to prevent collisions T-Rex and obstacles for as long as possible so you might want to check collision methods of Rect and Dino. To avoid collision you provide which movement should T-Rex make, that will be realized by Dino.tick method.
Your agent should implement interface given by Agent from game/agent.py, i.e. it should implement single static method get_move. Keep in mind that your agent won't be initialized since it is a simple-reflex agent, only its class variables verbose and debug will be set with respect to game options. Your static agent class should be implemented in script with name "class_name_lowercase.py", so you can use it to play the game.
You can implement your agent in prepared script agents/myagent.py. Example of agent implementation can be found in agents/dummy_agent.py.
For easier debugging you can run the game with option -d that allows your to visualize rectangles, lines and texts by calling methods of DebugGame class either from play_dino.py — add_initial_debug_visualization method or from your agent.
You can find example of this usage in agents/dummy_agent.py.
Take in account that speed of all obstacles is slowly increasing and that some birds are not static along x-axis.
You are expected to submit a single file containing your agent class. The system will perform a simple test running 50 simulations with random seed. You will be assigned up to 10 points based on the average score achieved by your agent. To earn full 10 points, your agent needs to achieve a score of at least 5000.