Extending an isotropic virtual environment model with geometrical information in model-mediated teleoperation

i-Botics is an open innovation centre for Research and Development in Interaction Robotics. The independent research centre has been founded by TNO and University of Twente and aims at developing knowledge and technology for value adding Robotic solutions. Industry partners in the full value chain and Governmental involvement is crucial for the development and successful implementation of these solutions. Knowledge is aggregated from various domains and researchers from our partners work together with us in the open innovation centre i-Botics.

One of the topics on which research and development is taken further within i-Botics, is telepresence (remote “being”) and telemanipulation (remote “operation”, “handling”). A typical scenario is that a complex task needs to be performed in an environment where a human cannot be safely (for example, deep sea or at a nuclear site). Yet, for the complex task human abilities as judging a situation based on perception and dexterity are essential. In this, time-delays in the communication between operator and robot have a significant effect on the ability of the operator to achieve a complex task remotely.

A virtual world could potentially be a middle ground between robot and operator in which the operator is able to work delay-free. This virtual environment would be updated by an active array of robotic sensors that gather information at the remote site. In addition, such a remote world could be used effectively for training purposes.

There is currently a system that can estimate dynamic properties of an environment online using RLS, and immediately assign it to a virtual environment, provided it is fully isotropic (i.e. the parameters uniquely describing the environment do not vary in space). However, in general, an environment consists of objects that each have their own dynamic properties. So, the dynamics of an environment may vary according to its geometry.
Also, the current model for the virtual environment is currently a body with inertia, linear viscous damping, and a Kelvin-Voigt contact model to account for flexible dynamics. This needs to be extended to include static friction to be accurate.

The goal of this assignment is to design a system that can adequately assign appropriate dynamics from an online estimator to multiple objects in an environment. Practically, this will involve:
* Integrating visual information with information from the slave manipulator in order to build up a picture of the geometry of the environment.
* Improving the RLS estimator to include stiction.
* Building an observer capable of determining which object in the environment the manipulator is in contact with and updating the estimate of its parameters.