Skip to content

Reinforcement learning #5

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sersangn wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Reinforcement learning #5

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8304318
failed attempt at lvl 2
sersangn Oct 16, 2023
a80f81c
FINISHED RULE BASED
sersangn Oct 16, 2023
00bbdf0
FINAL FINAL RULE BASED
sersangn Oct 16, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 17 additions & 30 deletions Rulebased.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,39 +4,29 @@
import cv2 as cv
import numpy as np

# code for locating objects on the screen in super mario bros
# by Lauren Gee
# Rule-Based Agent implementation for Gym Super Mario Bros World 1 Level 1

# 23599356 Sersang Ngedup
# 22957747 Kirsty Straiton

# Template matching is based on this tutorial:
# https://docs.opencv.org/4.x/d4/dc6/tutorial_py_template_matching.html

################################################################################

#Screen Height, width and match threshold constants
SCREEN_HEIGHT = 240
SCREEN_WIDTH = 256
MATCH_THRESHOLD = 0.9

################################################################################
# TEMPLATES FOR LOCATING OBJECTS

# ignore sky blue colour when matching templates
MASK_COLOUR = np.array([252, 136, 104])
# (these numbers are [BLUE, GREEN, RED] because opencv uses BGR colour format by default)

# You can add more images to improve the object locator, so that it can locate
# more things. For best results, paint around the object with the exact shade of
# blue as the sky colour. (see the given images as examples)
#
# Put your image filenames in image_files below, following the same format, and
# it should work fine.

# filenames for object templates
image_files = {
"mario": {
"small": ["marioA.png", "marioB.png", "marioC.png", "marioD.png",
"marioE.png", "marioF.png", "marioG.png"],
"tall": ["tall_marioA.png", "tall_marioB.png", "tall_marioC.png"],
},
"enemy": {
"goomba": ["goomba.png"],
Expand All @@ -46,12 +36,8 @@
"block": ["block1.png", "block3.png"],
"floorblock":["block2.png"],
"stair":["block4.png"],
"question_block": ["questionA.png", "questionB.png", "questionC.png"],
"pipe": ["pipe_upper_section.png", "pipe_lower_section.png"],
},
"item": {
"mushroom": ["mushroom_red.png"],
}
}

def _get_template(filename):
Expand Down Expand Up @@ -109,7 +95,6 @@ def _locate_object(screen, templates, stop_early=False, threshold=MATCH_THRESHOL
if stop_early and locations:
break

# [((x,y), (width,height))]
return [( loc, locations[loc]) for loc in locations]

def _locate_pipe(screen, threshold=MATCH_THRESHOLD):
Expand Down Expand Up @@ -187,9 +172,12 @@ def make_action(screen, info, step, env, prev_action):

# List of locations of blocks, pipes, etc:
block_locations = object_locations["block"]


################Above code was written by Lauren Gee and modified by us##########################
################Below code is our rule based agent##########################


################ Below code is our rule based agent ##########################

#DEFAULT ACTION
action = 3
Expand All @@ -198,13 +186,11 @@ def make_action(screen, info, step, env, prev_action):
##Code for jumping according to enemy locations
if len(enemy_locations)>0:
for enemy in enemy_locations:
if enemy[0][0]>mario_locations[0][0][0]:
if enemy[0][0]>mario_locations[0][0][0]: ##dont need this
if(enemy[0][0]-mario_locations[0][0][0]<40) and (enemy[0][1]>=180) and mario_locations[0][0][0] <= enemy[0][0]:
print(enemy)
print(mario_locations)
action=4

#-------------------PIPE CODE-------------------#
#-------------------PIPE CODE------------------------------#
pipeindex = 0
pipelist = []
pipe = False
Expand All @@ -216,11 +202,9 @@ def make_action(screen, info, step, env, prev_action):
#Code for jumping with pipe
if pipe:
for pipeindex in pipelist:
# print("pipe", pipeindex, block_locations[pipeindex])
if block_locations[pipeindex][0][0]<mario_locations[0][0][0]:
continue
if(block_locations[pipeindex][0][0]-mario_locations[0][0][0]<30):
print("PIPE!!!")
action=2

#-------------------DEALING WITH GAPS CODE-------------------#
Expand All @@ -231,21 +215,24 @@ def make_action(screen, info, step, env, prev_action):
botlevel.append(block)
if block[2]=="stair":
stairs.append(block)
#got some exception error but program continued... might introduce errors later

if len(botlevel) <= 13:
prevblock = botlevel[0]
for block in botlevel:
if ((block[0][0] - prevblock[0][0]) !=0 )and ((block[0][0] - prevblock[0][0]!=16)):
if prevblock[0][0] <= mario_locations[0][0][0] and mario_locations[0][0][0] <= block[0][0]:
print("GAP!!!")
# print("GAP!!!")
action= 2
prevblock= block

#-------------------DEALING WITH STAIRS CODE-------------------#

if len(stairs) > 0:
action = 4
for stair in stairs:
if stair[0][1] == 192 and stair[0][0]>mario_locations[0][0][0]:
if(stair[0][0]-mario_locations[0][0][0]<40) and mario_locations[0][0][0] <= stair[0][0]:
action=4

if step % 20 == 0:
if prev_action == 4:
action = 0
Expand All @@ -263,7 +250,7 @@ def make_action(screen, info, step, env, prev_action):
#Stops mario from flying off screen when he's at the top of the stairs
if mario_locations[0][0][1] <= 66:
action = 3
print(action)

return action

####################Implementing the rule based agent#################################
Expand Down