MIT/JSC/UMass/CSIRO AM Project Abstract

Site Map

Project Information:

To date there has been very little practical success in enabling robots to autonomously grasp and manipulate objects.

Our project is predicated on autonomous robotics because the communication delay makes it impossible to provide any timely feedback to the tele-operator. For example, a dropped tool would fall well before the drop would be observed. The approach will rely on tasking a robot via functional directives and then the robot taking the responsibility to carry out any necessary actions to meet the requirements of the directive. The robot will be responsible for dexterously handling tools, structures, and objectives with situational fluidity, dealing with unpredicated situations which often require instantaneous compensation. This will allow the robot to perform directives over the time scale of hours.

Massachusetts Institute of Technology (MIT), University of Massachusetts at Amherst (UMass), and the Commonwealth Scientific and Industrial Organization (CSIRO) in Brisbane Australia, propose collaborating with the Ambrose group to develop new algorithms that will run on existing NASA, MIT, UMass, and CSIRO hardware, with some upgrades, to enable more remote manipulation operations aimed at both in flight assembly and maintenance, as well as autonomous planetary surface operations. The work will aim to develop autonomous dexterity such that human supervision can occur at a coarser temporal scale. The supervisory commands will be able to lead by hours rather than seconds, and the robots will be relied upon to autonomously handle the details of the manipulation tasks.

We will address specific technical challenges in all components of autonomous manipulation. These components include positioning sensors, perception (both visual and tactile), placing the robot body optimally for the task, grasping, force operations, object and material transfer, disengaging from an object, and detecting failures.

We will develop a new generation of force-based hardware which will radically inrease the manipulation capabilities of robots. We will demonstrate the technology with remote directives to robots across the planet Earth.

The proposed technology will have application to at least two parts of the Exploration Program. It will be beneficial to assembly, maintenance, and servicing of vehicles in space even when they are remote from human presence. The technology will enable autonomous robots to assemble units together in space, connecting cables and pipes, and to maintain and service vehicles in space by opening and closing access panels, swapping out boxes, and reconnecting cables.

Additionally, the proposed technology will have application to lunar and planetary surface operations. It will enable robots to be used for site preparation and assembly of habitats from multiple components that need to be mated and to have cables and pipes connected. It will enable robots to prepare the surface by manipulating unstructured surface materials, e.g., digging trenches and then covering habitat modules with regolith to protect them from radiation. This will be achievable by many robots, but with relatively low bandwith requirements back to Earth (even from the lunar surface) and only a small oversight team of human controllers directing the overall flow of the work, rather than providing input for detailed motions. At greater distances, e.g., on the surface of Mars, it will enable a tremendous speedup in large scale surface operations, perhaps by a factor of 100 over today's remote supervision technology. It will thus make plausible the idea of sending robots ahead to prepare the habitats for the later arrival of humans.