In few words
In telepresence, the human operator will immerse into a remote environment and control a tele-operated device on a motion / force level. By means of a multimodal human machine interface (HMI) the human perceives and acts as in the real world. Generally, the modalities vision, audio and haptics play a superior role in human perception, when manipulation is required. From the robotics and control point of view, the haptic channel is of major interest, since it couples a human operator with a remote / distant teleoperator by energetic means.
Bilateral Control is the discipline which investigates the closed haptic loop created between the human operator and the remote / unreachable environment. Special control methods are hereby applied in order to stabilize an often very large loop whose communication delay, package loss inclusion, unavoidable nonlinearities and the inclusion of a human operator in the control loop make it especially challenging to tackle.
After stability, the main goal of a telepresence system is transparency, meaning in its ideal form that the user is not able to distinguish remote presence to local presence. The pursuit of stability often compromises transparency once the system constraints are established. This trade-off is a common denominator in every single approach dealing with bilateral control and signifies one of the main research areas within our telepresence framework. In this sense, one of the most accounted issues in haptic telemanipulation scenarios is the aforementioned time delay present in the communication channel. This often forces to design conservative control laws in order to achieve the unconditional system stability, which in turn often results in system transparency losses.
Among the most remarkable approaches in dealing with time-delayed telepresence, are those based on passivity criteria. The appeal of passivity-based approaches is the property that system passivity (and therefore stability) is granted by the passivity of its subsystems, without the need of a precise knowledge of the subsystem’s internals.
In this framework, we study three main methods to preserve passivity and the subsequent stability:
- Wave variable -based Methods for varying time delay and package loss: the wave variables transformation uses the power conjugated variables of force and velocity to define wave variables, the same way voltages and currents are related to energy waves in transmission lines. A wavy system has the interesting feature that passivity is preserved even with the time delay incurred by the data to travel from one point to the other. System stability is thus not altered in presence of communication delays. We have researched and extended this method to make it feasible for time-varying and package loss scenarios, such as the Internet or satellite -based communications.
- Bilateral Passivity Controller: the Time Domain Passivity (TDP) provides the mean to analyze passivity in real-time (opposed to the previous frequency domain -based approach), and bases its control on the real-time observability of the system energy. Thereby, passivity is no longer a design constraint which fixes controller rules (and thus compromising transparency), but rather an on the fly observation and correction of the passivity of the system.
- Bilateral Energy Transfer: the previous method is extended to introduce the bilateral energy transfer: to convey the energy introduced in the master side robot toward the slave robot (and vice versa) as faithfully as possible, with regard to the system passivity. Thus, two main characteristics are sought in the design of the control architecture:
- to limit unneeded energy dissipation and
- to convey the energy, even in the presence of unavoidable transmission delays.
Measuring performance becomes a tricky aspect in haptic contexts due to the fact that human perception is part of the system.
We have checked two of the few methods in the literature that tackle the problem using a velocity-force haptic telepresence system testing for different control parameters and different time delays in the channel:
- an analytical model-based method based on the Yokokohji's Transparency Indexes and
- an empirical approach based on the Z-width concept, which allows to measure transparency without the need of a precise knowledge of the system.
Furthermore, psycho-physical evaluation using the expertise from psychologists has been conducted to measure the cognitive impact upon an array of sixty test people. Participants had to actively perform as human operator on the OOS testbed explained in the section below dedicated to the applications.
Remarkable outputs of this study are:
- A psycho-physical transparency coefficient as a function of bilateral control parameters and the communication time delay.
- The conclusion that telepresence is a feasible approach for manned on-orbit servicing, up to delays of 650ms.
On-Orbit Satellite Servicing
One of our demonstrators is the on-orbit servicing simulator, where the user moves one of the arms of our HMI to telemanipulate a teleoperator through a delayed communication channel with adjustable delay. The teleoperator is a LWR-II, as shown in the picture below. The on-ground astronaut grabs first the target satellite. Once the satellite is stabilized, tasks based on activities similar to the ones conducted by astronauts for maintenance-purposes are executed: plugging / unplugging of umbilical connectors, opening and closing a small door of a control box, etc. The target satellite emulator is based on another LWR-II, which simulates the orbital movement and dynamics of a real satellite, reacting to forces and torques applied by the teleoperator. This second robot has a test board where the aforementioned tasks can be performed. Besides the haptic channel, visual (video captured by a stereocamera) feedback is as well provided to the human operator. It is further worth to mention, that in this setup three robots become energetically coupled to the human, who operates the system. A video of the application, here.
On-Orbit Satellite Servicing with the geostationary satellite ARTEMIS
Several experiemnts have been performed in collaboration with the LRT (Lehrstuhl für Raumfahrttechnik der Technischen Universität München, Garching, Germany) and the ESA (Redu, Belgium) to prove the feasibility of telepresence as solution for on-orbiting servicing using a relayed geostationary satellite-based communication. Relayed communications based on geostationary satellites allow for a wider contact window between the on-ground station (where the human operator commands the system) and the target satellite (upon which, the maintenance service is to be applied). The geo-satellite is used as a streaming mirror which reflects the data between on-ground station and target satellite. This comes at the price of an increased communication delay. The mean average measured during these experiments was up to 650ms.
The setup used the previously mentioned DLR-OOS tested. The communication link between master and slave systems included the ESA ARTEMIS geostationary satellite and the ESA ground station in Redu (Belgium), which was used as a mirror to send the data back to ARTEMIS and hence down to the slave system. Therefore, although master and slave devices were physically separated two meters from each other, the data packages traveled a distance of more than 130,000Km.
Multimodal Dexterous Fullbody Telepresence
Bimanual Dexterous Telepresence is here presented with the humanoid robot from DLR, Justin. Our HMI is used (1) to move the humanoid towards a certain target and (2) to teleoperate its arms, hands and fingers. Moreover, by means of a Head Mounted Display (HMD), the user sees in real-time by stereoscopic means that what Justin's eyes see. Further, by using of an optical tracking system, the user's head is tracked and further used to move Justin's head and torso. The human operator is thus highly immersed in the remote environment and is therefore able to execute complex tasks.As shown in the video below, using the grasping force controllers we are able to (dis)connect a bayonet nut remotely telemanipulating Justin's arms and hands, and always with visual and haptic feeback. A video of the application, here.