Ceng 491 - Project Kickoff Document Template Page 3
ADVERTISEMENT
1
2
3WP5 - (Defining game environments to agent which will use for reinforcement learning
algorithms.)
In this work package, the following functionalities / features / work items will be implemented
1. The tools and techniques will be found to get frames from the game and process them to
introduce the environment to the reinforcement learning algorithms that will be
implemented.
2. The final game will be chosen and examined in detail ( with source code etc.)
WP6 - (Improving and integrating proper learning algorithms, environment and APIs for the final
product.)
In this work package, the following functionalities / features / work items will be implemented
1. In this stage, different tools and algorithms used and selected will be integrated to the final
game according to the needs and if necessary, datasets will be generated.
2. A sample GUI or a set of plots will be generated to evaluate the algorithm performance
visually.
WP7 - (Final game agent development and testing.)
In this work package, the following functionalities / features / work items will be implemented
1. The final product will be examined and tested with different agents, bots or by hand.
2. The source code will be examined and refactored if necessary.
3. The final product will be tried, new visual properties will be added and the product will be
prepared to be ready for final demo.
Risk Assessment
Risk #
Description
Possible Solution(s)
1
Some work to be done in limited time can be
With declaration in backlog
delayed.
reports, these works can be shifted
to following sprints.
2
The selected tools can become insufficient after a
The current tool or library could
time.
be changed with a proper one.
3
Some selected games can be much complex to get
In this situation, the game could
data or to train an agent that was implemented a
be changed.
selected kind of algorithm.
4
There is a possibility that the algorithms mentioned
Some modifications and changes
in the work-packages can’t match with the task.
in the algorithms are possible.
5
Final product type is flexible, it can be an agent that
The decision domain is restricted
may adjust to a few games or an agent that learns to
so the type of agent can be
play a specific comprehensive game
decided or changed according to
the improvements and
circumstances.
3
ADVERTISEMENT
0 votes
Related Articles
Related forms
Related Categories
Parent category: Business