Machine_learning_2023/AI_brain/rotate_and_go_a_star.py
2023-05-05 15:12:48 +02:00

139 lines
4.2 KiB
Python

import math
import queue
from dataclasses import dataclass, field
from typing import Any
from domain.world import World
class State:
def __init__(self, x, y, direction=(1, 0)):
self.x = x
self.y = y
self.direction = direction
def __hash__(self):
return hash((self.x, self.y))
def __eq__(self, other):
return (self.x == other.x and self.y == other.y
and self.direction == other.direction)
class Node:
def __init__(self, state: State):
self.state = state
self.parent = None
self.action = None
self.g = 0
self.h = 0
@dataclass(order=True)
class PrioritizedItem:
priority: int
item: Any = field(compare=False)
def __iter__(self):
return iter((self.priority, self.item))
def action_sequence(node: Node):
actions = []
while node.parent:
actions.append(node.action)
node = node.parent
print(node.g)
actions.reverse()
return actions
class RotateAndGoAStar:
def __init__(self, world: World, start_state: State, goal_state: State):
self.world = world
self.start_state = start_state
self.goal_state = goal_state
self.fringe = queue.PriorityQueue()
self.enqueued_states = {}
self.explored = set()
self.actions = []
def search(self):
h = abs(self.start_state.x - self.goal_state.x) ** 2 + abs(self.start_state.y - self.goal_state.y) ** 2
self.fringe.put(PrioritizedItem(h, Node(self.start_state)))
while not self.fringe.empty():
priority, elem = self.fringe.get()
self.enqueued_states.pop(elem, 0)
if self.is_goal(elem.state):
self.actions = action_sequence(elem)
return True
self.explored.add(elem.state)
for (action, state) in self.successors(elem.state):
next_node = Node(state)
next_node.action = action
next_node.parent = elem
next_node.g = (elem.g
+ abs(elem.state.x - state.x)
+ abs(elem.state.y - state.y)
+ self.world.get_cost(state.x, state.y)
)
next_node.h = abs(state.x - self.goal_state.x) + abs(state.y - self.goal_state.y)
self.print_fringe()
f = next_node.g + next_node.h
if state not in self.enqueued_states and state not in self.explored:
self.fringe.put(PrioritizedItem(f, next_node))
self.enqueued_states[state] = f
elif self.enqueued_states.get(state, -math.inf) > f:
self.add_existed(next_node, f)
self.enqueued_states.pop(state, 0)
self.enqueued_states[state] = f
return False
def print_fringe(self):
elems = []
while not self.fringe.empty():
e = self.fringe.get()
elems.append(e)
print(str(e.item.state.x) + ":" + str(e.item.state.x))
for e in elems:
self.fringe.put(e)
def add_existed(self, node: Node, f: int):
old = []
while not self.fringe.empty():
e = self.fringe.get()
if e.item.state == node.state:
break
old.append(e)
self.fringe.put(PrioritizedItem(f, node))
for e in old:
self.fringe.put(e)
def successors(self, state: State):
new_successors = [
# rotate right
("RR", State(state.x, state.y, (-state.direction[1], state.direction[0]))),
# rotate left
("RL", State(state.x, state.y, (state.direction[1], -state.direction[0]))),
]
if self.world.accepted_move(state.x + state.direction[0], state.y + state.direction[1]):
new_successors.append(
("GO", State(state.x + state.direction[0], state.y + state.direction[1], state.direction)))
return new_successors
def is_goal(self, state: State) -> bool:
return (
state.x == self.goal_state.x
and state.y == self.goal_state.y
)