oxbrbtx/PrimAITE
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/dev' into feature/917_Integrate_with_RLLib

Browse Source 
# Conflicts:
#	tests/test_reward.py
This commit is contained in:
Chris McCarthy
2023-07-03 11:59:16 +01:00
parent c38c13b829 4f0f542570
commit 123ec8343c
1 changed files with 18 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
tests/test_reward.py
View File

@@ -21,19 +21,27 @@ def test_rewards_are_being_penalised_at_each_step_function(
When the initial state is OFF compared to reference state which is ON. When the initial state is OFF compared to reference state which is ON.
On different steps (of the 13 in total) these are the following rewards The config 'one_node_states_on_off_lay_down_config.yaml' has 15 steps:
for config_6 which are activated: On different steps, the laydown config has Pattern of Life (PoLs) which change a state of the node's attribute.
File System State: goodShouldBeCorrupt = 5 (between Steps 1 & 3) For example, turning the nodes' file system state to CORRUPT from its original state GOOD.
Hardware State: onShouldBeOff = -2 (between Steps 4 & 6) As a result these are the following rewards are activated:
Service State: goodShouldBeCompromised = 5 (between Steps 7 & 9) File System State: corrupt_should_be_good = -10 * 2 (on Steps 1 & 2)
Software State (Software State): goodShouldBeCompromised = 5 (between Hardware State: off_should_be_on = -10 * 2 (on Steps 4 & 5)
Steps 10 & 12) Service State: compromised_should_be_good = -20 * 2 (on Steps 7 & 8)
Software State: compromised_should_be_good = -20 * 2 (on Steps 10 & 11)
Total Reward: -2 - 2 + 5 + 5 + 5 + 5 + 5 + 5 = 26 The Pattern of Life (PoLs) last for 2 steps, so the agent is penalised twice.
Step Count: 13
Note: This test run inherits from conftest.py where the PrimAITE environment is ran and the blue agent is hard-coded
to do NOTHING on every step.
We use Pattern of Lifes (PoLs) to change the nodes states and display that the agent is being penalised on all steps
where the live network node differs from the network reference node.
Total Reward: -10 + -10 + -10 + -10 + -20 + -20 + -20 + -20 = -120
Step Count: 15
For the 4 steps where this occurs the average reward is: For the 4 steps where this occurs the average reward is:
Average Reward: 2 (26 / 13) Average Reward: -8 (-120 / 15)
""" """
with temp_primaite_session as session: with temp_primaite_session as session:
session.evaluate() session.evaluate()