text.skipToContent text.skipToNavigation
background-image

Overcoming the Network Problem in Telepresence Systems with Prediction and Inertia von Clarke-Griebsch, Stella (eBook)

  • Verlag: Utz Verlag
eBook (PDF)
33,99 €
inkl. gesetzl. MwSt.
Sofort per Download lieferbar

Online verfügbar

Overcoming the Network Problem in Telepresence Systems with Prediction and Inertia

Production engineering is of central importance to the continuous development of our industrial society. The performance capacity of manufacturing companies is highly dependent on the utilised resources, the applied production techniques, and the implemented organisation strategies. Guaranteeing a company's success entails an ideal combination of technology, organisation and workforce management. Achieving an optimal configuration of cost, time and quality is a highly complex and intricate task which requires production strategies to be continuously monitored, reviewed and enhanced. This involves both a reduction in, and a sound understanding of, the complexity of products, manufacturing systems and operations.

Produktinformationen

    Format: PDF
    Kopierschutz: AdobeDRM
    Seitenzahl: 164
    Sprache: Englisch
    ISBN: 9783831607013
    Verlag: Utz Verlag
    Größe: 2157 kBytes
Weiterlesen weniger lesen

Overcoming the Network Problem in Telepresence Systems with Prediction and Inertia

6 Conclusions and Future Work (p. 117-118)

This thesis focused on methods to overcome the network problem in telepresence systems. Telepresence applications include all scenarios which require a task to be completed in an environment which is, for some reason, inaccessible to humans, but requires human intelligence. Examples include remote surgery, manual microassembly, disposal of explosives, deep sea exploration, and satellite repair in outer space. All of these scenarios necessarily involve a communication mechanism between the human operator and the remote robot or teleoperator. As with all networks, this communication layer will contain network delays which can adversely affect the operation.

The research in this thesis concentrated on overcoming or dealing with such delays through the use of simulated inertia and prediction algorithms. Two main research questions were addressed. The first question focused on the use of simulated inertia as a supporting factor in telepresence scenarios, with and without network delays. The second question aimed to find an optimal prediction algorithm for a typical telepresence scenario.

To answer the first question, an experiment was conducted to determine the influence of simulated inertia in the haptic device on operator performance and immersion under both delayed and non-delayed network conditions. To create a delayed network, the real international network layer between the Technical University of Munich, Germany, and the University of Wollongong, Australia, was utilised. This experiment involved thirty-six human subjects completing a total of 216 experiments under varying simulated inertia and network conditions. In answer to the first research question, simulated inertia in a haptic device was found to have no significant influence on operator performance or sense of presence.

This finding stands in contrast to the author's hypothesis that inertia could act as a supporting mechanism through increasing the stability of the input device and decreasing the choppiness of the operator's movements. Nevertheless, the fact that inertia had no significant influence on important operator performance measurements means that inertia can be implemented when it is expected to yield other secondary advantages. Specifically, inertia in the haptic device has a smoothing effect on the dynamic input data. Smoother data is easier to predict, which is directly relevant to the second research question.

To answer the second research question, a strategically selected group of prediction algorithms was chosen for analysis. The four chosen algorithms were Double Exponential Smoothing, the Kalman Filter, Neural Networks, and Support Vector Regression. Each algorithm was optimised against dynamic data extracted from a typical telepresence scenario. The optimal algorithm was then implemented in two schemas. The first schema involved using the algorithm to directly compensate for network delays through the prediction of future movement data. By sending these predicted val ues across the network, the delay can be partially or fully compensated for.

The second schema involved implementing the prediction algorithm together with the socalled deadband approach to reconstruct previously compressed data. In this schema, the prediction essentially contributes to a compression algorithm. In answer to the second research question, Double Exponential Smoothing was found to be the optimal prediction algorithm for a given typical telepresence scenario. This algorithm was implemented into a prediction schema, where it was shown to improve operator performance by up to 32%.

Weiterlesen weniger lesen

Kundenbewertungen