import numpy as np from cliport.tasks.task import Task from cliport.utils import utils class BallOnBoxOnContainer(Task): """Pick up each ball and place it on the corresponding colored box, which are located in specific positions on a container.""" def __init__(self): super().__init__() self.max_steps = 15 self.lang_template = "place the {} box on the container" self.lang_template_2 = "place the {} ball on the {} box" self.task_completed_desc = "done placing balls on boxs and box on container." self.additional_reset() def reset(self, env): super().reset(env) # Add container. container_size = (0.2, 0.2, 0.06) container_pose = self.get_random_pose(env, container_size) container_template = 'container/container-template.urdf' replace = {'DIM': container_size, 'HALF': (container_size[0] / 2, container_size[1] / 2, container_size[2] / 2)} container_urdf = self.fill_template(container_template, replace) env.add_object(container_urdf, container_pose, 'fixed') # Define colors. ball_colors = ['red'] box_colors = ['blue'] # Add boxs. box_size = (0.04, 0.04, 0.06) box_template = 'box/box-template.urdf' boxs = [] replace = {'DIM': box_size, 'HALF': (box_size[0] / 2, box_size[1] / 2, box_size[2] / 2), 'COLOR': ball_colors[0]} box_urdf = self.fill_template(box_template, replace) box_pose = self.get_random_pose(env, box_size) box_id = env.add_object(box_urdf, box_pose) boxs.append(box_id) # Add balls. ball_size = (0.04, 0.04, 0.04) ball_urdf = 'ball/ball.urdf' balls = [] ball_pose = self.get_random_pose(env, ball_size) ball_id = env.add_object(ball_urdf, ball_pose, color=box_colors[0]) balls.append(ball_id) # Goal: place the box on top of the container self.add_goal(objs=[boxs[0]], matches=np.ones((1, 1)), targ_poses=[container_pose], replace=False, rotations=True, metric='pose', params=None, step_max_reward=1/2, language_goal=self.lang_template.format(box_colors[0])) # Goal: place the ball on top of the box language_goal = self.lang_template_2.format(ball_colors[0], box_colors[0]) self.add_goal(objs=[balls[0]], matches=np.ones((1, 1)), targ_poses=[container_pose], replace=False, rotations=True, metric='pose', params=None, step_max_reward=1/2, language_goal=language_goal)