Data preparation

[aopolin-lv]

Hello, ggik is a nice work. I have a question that need you help.
If I have a urdf file of a specified robot arm, and many joint-ee pairs data. How can I train a ggik model using that data? Do you have suggestions?

[Olimoyo]

Hi, thank you for your interest in our work! If your robot is custom, you would need to represent your robot graphically using this repository: https://github.com/utiasSTARS/GraphIK.

Then you can generate data by following the script here. In particular, you can see that we can generate data by loading a few different robots in lines:

generative-graphik/generative_graphik/utils/dataset_generation.py

Lines 154 to 185 in 9263453

	for robot_name in robots:
	# generate data for robot like ur10, kuka etc.
	if robot_name == "ur10":
	# randomize won't work on ur10
	# robot, graph = load_ur10(limits=None, randomized_links = False)
	fname = graphik.__path__[0] + "/robots/urdfs/ur10_mod.urdf"
	q_lim_l = -np.pi * np.ones(6)
	q_lim_u = np.pi * np.ones(6)
	elif robot_name == "kuka":
	# robot, graph = load_kuka(limits=None, randomized_links = params["randomize"], randomize_percentage=0.2)
	fname = graphik.__path__[0] + "/robots/urdfs/kuka_iiwr.urdf"
	q_lim_l = -np.pi * np.ones(7)
	q_lim_u = np.pi * np.ones(7)
	elif robot_name == "lwa4d":
	# robot, graph = load_schunk_lwa4d(limits=None, randomized_links = params["randomize"], randomize_percentage=0.2)
	fname = graphik.__path__[0] + "/robots/urdfs/lwa4d.urdf"
	q_lim_l = -np.pi * np.ones(7)
	q_lim_u = np.pi * np.ones(7)
	elif robot_name == "panda":
	# robot, graph = load_schunk_lwa4d(limits=None, randomized_links = params["randomize"], randomize_percentage=0.2)
	fname = graphik.__path__[0] + "/robots/urdfs/panda_arm.urdf"
	# q_lim_l = -np.pi * np.ones(7)
	# q_lim_u = np.pi * np.ones(7)
	q_lim_l = np.array([-2.8973, -1.7628, -2.8973, -3.0718, -2.8973, -0.0175, -2.8973])
	q_lim_u = np.array([2.8973, 1.7628, 2.8973, -0.0698, 2.8973, 3.7525, 2.8973])
	elif robot_name == "lwa4p":
	# robot, graph = load_schunk_lwa4p(limits=None, randomized_links = params["randomize"], randomize_percentage=0.2)
	fname = graphik.__path__[0] + "/robots/urdfs/lwa4p.urdf"
	q_lim_l = -np.pi * np.ones(6)
	q_lim_u = np.pi * np.ones(6)
	else:
	raise NotImplementedError