Context Robotics of the future must be more collaborative and interactive with the human to accomplish more complex, tedious and repetitive tasks for the human but also complex in terms of decision-making for the robot.
Indeed, human-robot collaboration and interaction emerge in several robotic applications, such as industrial and assistance robotics and exoskeletons. This is made possible by the ability of some type of robots, commonly called "cobots", to work in a real interaction with humans, by supporting and helping them in achieving tasks, by improving ergonomics, hardness and security criteria, while taking full advantage of the human intelligence .
One of the main tasks is objects co-manipulation for displacement, positioning in space or co-manipulation of tools purposes .
This type of task involves problems known in the literature as physical Human-Robot Interaction pHRI , which is not enough explored in the collaborative robotics. It particularly involves a continuous interaction Phd thesis robotics is even less explored, due to the complexity that it brings compared Phd thesis robotics occasional interaction , where the man and the robot share the workspace.
Handling of these measurements, for an adapted reference generation and the implementation of control laws, must allow a better achievement of the collaborative task by an adapted robot movement programming. The main objective of this proposal is to explore robotic control strategies that allow co-manipulation of different sizes and weights objects.
This physical interaction must be continuous while ensuring the stability of the interaction as well as the safety and comfort of the operator.
For this, the human operator must be considered as a "leader", due to his intelligence and decision-making capabilities, and the robot as a "follower" which analyzes and read the operator actions to adapt his control .
This allows guiding the execution of the task by the operator, by giving a "reactive" behavior to the robot. Present works will explore the robotics force control  and compliant control less rigid behavior laws in order to achieve our objectives.
For this purpose, we will explore two categories of force control: A comparative evaluation between these two strategies is planned. Therefore, these control laws require the use of the robot dynamic model in contact with its environment .
Furthermore, these works should be accomplished though several steps with different task knowledge levels. Firstly, a partial knowledge of the operator's intention rotational or translational movement of the object will be assumed without quantitative knowledge of its measurements movements and effortswhich will be integrated as references for robot movement generator.
Secondly, it is planned to ignore all a priori operator intention knowledge and to establish a global intention prediction strategies based on models that should be explored and defined .
The considered control law must give the robot a compliant behavior, which compliance levels should be considered by defining some compromise.
Indeed, more compliance presents a safety advantage but with less accuracy, while less compliance presents more precision but with uncomfortable and more risky interaction for the operator. Unlike the main work in this topic, where the robot compliance is predefined and does not change during the task, we will explore in the current thesis an original approach that consists of varying the robot's compliance throughout the task, taking into account evolution in robot and operator states.
In the first step, our methods and control strategies will be tested on simulation software. Then it will be adapted and implemented for an experimental validation on the Kuka iiwa collaborative robot, available at LGP laboratory of Tarbes France.
Scientific issues In order to achieve the thesis objectives, some scientific obstacles have to be addressed and overcame.
The first main obstacle of this thesis is about the synthesis of robot control laws that are adapted for the continuous human-robot physical interaction pHRI context.
These laws must be both compliant for the operator and efficient for the task accomplishment and its accuracy.Rensselaer Artificial Intelligence and Reasoning Laboratory.
The Rensselaer Artificial Intelligence and Reasoning (RAIR) Laboratory is located in the 3rd and 4th floor of the historic Winslow building. To reach the Winslow Building by GPS, search for 'Winslow Building Troy NY'. Theses and dissertations are an excellent source of in-depth technical information often not published in detail elsewhere.
To review theses and dissertations by subject area, try these links, chosen for their wide appeal and timeliness. Find and compare PhD programmes and postgraduate doctorate studies from top universities worldwide: search thousands of programmes to do research abroad or at home.
Thesis In order to obtain an MSEE degree all students must complete a Graduate Paper or a Graduate Thesis. Of the minimum 30 credit hours needed to earn the degree, a typical student earns 24 to 27 credit hours from course work and the remaining credit hours from the Graduate Paper or MS Thesis.
This thesis describes the creation and control of soft robots made of deformable elastomer materials and powered by fluidics. We embed soft fluidic actuators into self-contained soft robotic systems, such as fish for underwater exploration or soft arms for dynamic manipulation.
MS Thesis Robotics / Control Curriculum Six of those credits will be MS Thesis credits (MCEN ). Students should aim to complete at least 15 of the remaining 24 credits in Mechanical Engineering and may take up to 9 credits outside of the department.