mobileprint 0.0.1

GUI for measuring human performance on mobile construction task

Setup

• Python >= 3.6 installed on your machine

Start in interactive python shell:

$ python

import mobileprint
mobileprint.run()

Instructions

The goal of this game is to move around an environment and lay bricks in the environment such that they match a given plan.

Example of a plan.

Environments

There are 8 different environment configurations to choose from:

• 1D Static (Plan 1)
• 1D Static (Plan 2)
• 1D Static (Plan 3)
• 1D Dynamic
• 2D Static (Dense)
• 2D Static (Sparse)
• 2D Dynamic (Dense)
• 2D Dynamic (Sparse)

Static plans are fixed, pre-determined plans. You can not view the overall plan for static environments, but you can use 'Training Mode', and infer the plan layout by interacting with the environment and examining the rewards.

Dynamic plans morph with each new episode, but you can see the overall plan layout.

Each environment has a predetermined maximum number of steps and number of bricks that you may use.

Controls

LEFT : Move Left

RIGHT : Move Right

UP : Move Up

DOWN : Move Down

SPACE  : Drop a brick

Game Mode

You can either play in 'Training Mode' or 'Evaluation Mode'.

In 'Training Mode', you will be able to see the reward for each move you make, as well as your total cumulative score. This will allow you to learn what leads you to a good or bad score.

In 'Evaluation Mode', you will no longer be able to view your score, and will simply have to rely on your memory of the plan (or in Dynamic plan environements, you will be able to see the overall plan, but still have no access to your score).

Results

At the end of each episode, you will be prompted to save your results for that episode. Assuming the episode was an honest attempt to complete the task, please choose 'Yes'. Your results will be logged into a directory 'results' in a .csv file. When you have finished playing the game / played as many episodes as you wish to, please send us your .csv results files so that we can compile the results and build a good benchmark!