From b4d9fdba69c1f8233e0e77d410ef4c124ce77bb0 Mon Sep 17 00:00:00 2001
From: Antonio Andriella <aandriella@iri.upc.edu>
Date: Tue, 27 Oct 2020 08:29:10 +0100
Subject: [PATCH] add policy_load and get_state_action methods
---
src/robot_behaviour/robot_reproducer.py | 28 +++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
diff --git a/src/robot_behaviour/robot_reproducer.py b/src/robot_behaviour/robot_reproducer.py
index f70596a..8733120 100644
--- a/src/robot_behaviour/robot_reproducer.py
+++ b/src/robot_behaviour/robot_reproducer.py
@@ -7,12 +7,14 @@ facial expression and gesture. Every time we check if the action
import rospy
import random
import ast
+import pickle
+import numpy as np
from robot_behaviour.face_reproducer import Face
from robot_behaviour.speech_reproducer import Speech
from robot_behaviour.gesture_reproducer import Gesture
class Robot:
- def __init__(self, speech, sentences_file, face=None, gesture=None):
+ def __init__(self, speech, sentences_file, action_policy_filename=None, face=None, gesture=None):
'''
:param speech: instance of class Speech
:param sentences_file: the file where all the sentences are stored
@@ -23,6 +25,7 @@ class Robot:
self.sentences = self.load_sentences(sentences_file)
self.face = face
self.gesture = gesture
+ self.action_policy = self.load_robot_policy(action_policy_filename)
self.action = {
"instruction": self.instruction,
@@ -52,10 +55,31 @@ class Robot:
"neutral" : self.neutral
}
+
+ def load_robot_policy(self, learned_policy_filename):
+ with open(learned_policy_filename, "rb") as f:
+ loaded_policy = pickle.load(f)
+ return loaded_policy
+
+ def get_irl_state_action(self, state_index, epsilon=0.1):
+ action = 0
+ print("Select it between the following:", self.action_policy[state_index])
+ if random.random() < epsilon:
+ new_list = (self.action_policy[state_index])
+ best_action_index = np.argmax(self.action_policy[state_index])
+ new_list[best_action_index] = 0
+ action = np.argmax(new_list)
+ else:
+ action = np.argmax(self.action_policy[state_index])
+ return action
+
+ def get_random_state_action(self):
+ return random.randint(0, 6)
+
+
def send_to_rest(self):
self.gesture.initial_pos()
-
def load_sentences(self, file):
file = open(file, "r")
contents = file.read()
--
GitLab