It is often difficult to create autonomous robotic capabilities that match what can be achieved via teleoperation. Even though it is mechanically possible for a humanoid robot such as Robonaut 2 to perform complex coordinated tasks such as tying a knot, exchanging objects between end effectors, plugging in connectors, unscrewing a cap, opening a door, or grasping large objects with two hands, our lack of planning algorithms makes it difficult to control these behaviors autonomously. The lack of planning and control algorithms also impedes human-robot interaction as it is difficult for manipulation robots to plan arm trajectories in real-time using active sensing to avoid collisions with humans. This proposal is to develop a suite of planning and control algorithms that will enable NASA robots to perform complex manipulation behaviors in a coordinated way. This work would benefit NASA by making NASA robots more capable and useful during autonomous tasks, by enabling remote supervisors to command more complex tasks, and by enabling NASA robots to operate safely alongside humans during shared tasks.