left_pos = action.left_param_action([0., 0., 0., 0., 0., 0., 0.])
# move the joints to the new positions.
limb_right.move_to_joint_positions(right_pos)
limb_left.move_to_joint_positions(left_pos)
# update the current joint angles (could do this by using the positions
# concern, there may be variability with movement, especially if the robot is holding
# a weighty object, might need to reconsider to speed up processing.
# DONE determine if getting updates significantly slows and if there is a difference between calculated angles
action.action_update(limb_right.joint_angles(), limb_left.joint_angles())
# get the endpoint pose for reward calculation
right_reward.update_gripper(limb_right.endpoint_pose())
left_reward.update_gripper(limb_left.endpoint_pose())
# get the rewards and status
right_distance, right_done = right_reward.euclidean_distance()
left_distance, left_done = left_reward.euclidean_distance()
print "Reward for iteration " + str(i)
print right_reward.euclidean_distance()
print left_reward.euclidean_distance()
if left_done or right_done:
# block_sample.delete_gazebo_models()
quit()
block_sample.delete_gazebo_models()
quit()
class Baxter(object):
def __init__(self):
# initialize baxter node in ROS
rospy.init_node('baxter_node')
# create instances of baxter_interfaces's limb class
self.limb_right = baxter_interface.Limb('right')
self.limb_left = baxter_interface.Limb('left')
# load the gazebo model for simulation
block_sample.load_gazebo_models()
# create a new action for both arms
self.action = Action(self.limb_right.joint_angles(),
self.limb_left.joint_angles())
self.neutral()
# initialize the reward with goal location, param 2 is block location in sample
# TODO add a 2nd block, have each hand work towards their own goal.
# TODO use input from robot vision to get goal position
self.right_reward = Reward(self.limb_right.endpoint_pose(),
(0.6725, 0.1265, 0.7825))
self.left_reward = Reward(self.limb_left.endpoint_pose(),
(0.6725, 0.1265, 0.7825))
def neutral(self):
# set the arms to a neutral position
self.limb_right.move_to_joint_positions(
self.action.right_home_position())
self.limb_left.move_to_joint_positions(
self.action.left_home_position())
def neutral_left(self):
# set the arms to a neutral position
self.limb_left.move_to_joint_positions(
self.action.left_home_position())
def neutral_right(self):
# set the arms to a neutral position
self.limb_right.move_to_joint_positions(
self.action.right_home_position())
def step_left(self, action):
start_pos = self.left_state()
# get an incremental update to the joint positions.
left_pos = self.action.left_param_action(action)
# move the joints to the new positions.
self.limb_left.move_to_joint_positions(left_pos)
# update current state of robot
self.action.action_update(self.right_state(), self.left_state())
# get the endpoint pose for reward of robot
self.left_reward.update_gripper(self.limb_left.endpoint_pose())
reward = self.left_reward.euclidean_distance()
done = self.left_reward.is_done(reward)
# return the reward and status
# print reward
return start_pos, left_pos, reward, done
def step_right(self, action):
start_pos = self.right_state()
# get an incremental update to the joint positions.
right_pos = self.action.right_param_action(action)
# move the joints to the new positions.
self.limb_right.move_to_joint_positions(right_pos)
# update current state of robot
self.action.action_update(self.right_state(), self.left_state())
# get the endpoint pose for reward calculation
self.right_reward.update_gripper(self.limb_right.endpoint_pose())
reward = self.right_reward.euclidean_distance()
done = self.right_reward.is_done(reward)
# get the rewards and status
return start_pos, right_pos, reward, done
def random_step_left(self):
init_arm_pos = self.left_state()
item = 0
new_arm_pos = init_arm_pos
value = [-0.1, 0, 0.1]
arm_update_value = []
for i in range(7):
arm_update_value.append(random.choice(value))
for key, value in new_arm_pos.iteritems():
# sets new arm position to initial position + a random value
new_arm_pos[key] = init_arm_pos[key] + arm_update_value[item]
item += 1
self.limb_left.move_to_joint_positions(new_arm_pos)
self.action.action_update(self.right_state(), self.left_state())
self.left_reward.update_gripper(self.limb_left.endpoint_pose())
reward = self.left_reward.euclidean_distance()
done = self.left_reward.is_done(reward)
return new_arm_pos, arm_update_value, reward, done
def random_step_right(self):
init_arm_pos = self.right_state()
item = 0
new_arm_pos = init_arm_pos
value = [-0.1, 0, 0.1]
arm_update_value = []
for i in range(7):
arm_update_value.append(random.choice(value))
for key, value in new_arm_pos.iteritems():
# sets new arm position to initial position + a random value
new_arm_pos[key] = init_arm_pos[key] + arm_update_value[item]
item += 1
self.limb_right.move_to_joint_positions(new_arm_pos)
self.action.action_update(self.right_state(), self.left_state())
self.right_reward.update_gripper(self.limb_right.endpoint_pose())
reward = self.right_reward.euclidean_distance()
done = self.right_reward.is_done(reward)
return new_arm_pos, arm_update_value, reward, done
def right_state(self):
# returns an array of the current joint angles
return self.limb_right.joint_angles()
def left_state(self):
# returns an array of the current joint angles
return self.limb_left.joint_angles()
def close_env(self):
block_sample.delete_gazebo_models()
self.neutral()
quit()
def observation_space(self):
low = np.array(
[-1.7016, -2.147, -3.0541, -.05, -3.059, -1.5707, -3.059])
high = np.array([1.7016, 1.047, 3.0541, 2.618, 3.059, 2.094, 3.059])
return [low, high]
def action_space(self):
return [[-0.1, 0, 0.1], [-0.1, 0, 0.1], [-0.1, 0, 0.1], [-0.1, 0, 0.1],
[-0.1, 0, 0.1], [-0.1, 0, 0.1], [-0.1, 0, 0.1]]
def action_domain(self):
low = -0.1
# may want to change this, add in a bigger value like .2
none = 0.0
high = 0.1
return low, none, high
def reset(self, arm):
self.neutral()
if arm == "right":
state = self.right_state()
elif arm == "left":
state = self.left_state()
else:
print "non-valid arm parameter: enter 'left' or 'right'"
state = []
state_list = list(state.values())
return state_list
def reset_random(self, arm):
if arm == "right":
self.limb_right.move_to_joint_positions(
self.action.right_random_position())
state = self.right_state()
elif arm == "left":
self.limb_left.move_to_joint_positions(
self.action.left_random_position())
state = self.left_state()
else:
print "non-valid arm parameter: enter 'left' or 'right'"
state = []
state_list = list(state.values())
return state_list
def random_start(self):
# set the arms to a random start position
self.limb_right.move_to_joint_positions(
self.action.right_random_position())
self.limb_left.move_to_joint_positions(
self.action.left_random_position())