Limits...
Marginal probabilistic modeling of the delays in the sensory data transmission of networked telerobots.

Gago-Benítez A, Fernández-Madrigal JA, Cruz-Martín A - Sensors (Basel) (2014)

Bottom Line: This paper studies marginal probability distributions that, under mild assumptions, can be a good approximation of the real distribution of the delays without using knowledge of their dynamics, are efficient to compute, and need minor modifications on the networked robot.Since sequences of delays exhibit strong non-linearities in these networked applications, to satisfy the iid hypothesis required by the marginal approach we apply a change detection method.The results reported here indicate that some parametrical models explain well many more real scenarios when using this change detection method, while some non-parametrical distributions have a very good rate of successful modeling in the case that non-linearity detection is not possible and that we split the total delay into its three basic terms: server, network and client times.

View Article: PubMed Central - PubMed

Affiliation: Systems Engineering and Automation Department, University of Málaga, Campus Teatinos, Boulevard Luis Pasteur s/n, Málaga 29071, Spain. anagagobenitez@gmail.com.

ABSTRACT
Networked telerobots are remotely controlled through general purpose networks and components, which are highly heterogeneous and exhibit stochastic response times; however their correct teleoperation requires a timely flow of information from sensors to remote stations. In order to guarantee these time requirements, a good on-line probabilistic estimation of the sensory transmission delays is needed. In many modern applications this estimation must be computationally highly efficient, e.g., when the system includes a web-based client interface. This paper studies marginal probability distributions that, under mild assumptions, can be a good approximation of the real distribution of the delays without using knowledge of their dynamics, are efficient to compute, and need minor modifications on the networked robot. Since sequences of delays exhibit strong non-linearities in these networked applications, to satisfy the iid hypothesis required by the marginal approach we apply a change detection method. The results reported here indicate that some parametrical models explain well many more real scenarios when using this change detection method, while some non-parametrical distributions have a very good rate of successful modeling in the case that non-linearity detection is not possible and that we split the total delay into its three basic terms: server, network and client times.

No MeSH data available.


Related in: MedlinePlus

Illustrative histograms and fittings of the models that do not reject one or both hypothesis tests according to Table 4.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3958226&req=5

f5-sensors-14-02305: Illustrative histograms and fittings of the models that do not reject one or both hypothesis tests according to Table 4.

Mentions: After running the significance tests, we see in Table 4 that, as expected from the multimodal nature of most scenarios, only a few of them can be explained by parametrical models: #3, #2 and #18 (corresponding to the B class and to a percentage of all delay values in all the scenarios below 12%). The rest of the scenarios do not support the hypothesis. In the case of non-parametrical models, kernel and spline provide the best results, with above 40% of delay values explained. Non-rejected scenarios are displayed in Figure 5 for illustrative purposes. These results show that, for a parametrical choice, the lognormal or erlang distributions could be preferred when we are not able to detect outliers, bursts and regimes, but these are not good enough and might be applicable to very few scenarios. For a non-parametrical option, the kernel distribution should be the best election. In general non-parametrical models are able to explain scenarios from all classes (A, B, C), given their closest fitting to the particular data in the sample (which leads, on the other hand, to the previously mentioned problem of overfitting). The SoG model obtains worse results than the kernel due to its restriction in the number of modes, while the spline is also worse than the kernel due to its greater overfitting, that produces very good models of the data used for fitting but not in its goodness-of-fit with the delays reserved for hypothesis testing.


Marginal probabilistic modeling of the delays in the sensory data transmission of networked telerobots.

Gago-Benítez A, Fernández-Madrigal JA, Cruz-Martín A - Sensors (Basel) (2014)

Illustrative histograms and fittings of the models that do not reject one or both hypothesis tests according to Table 4.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3958226&req=5

f5-sensors-14-02305: Illustrative histograms and fittings of the models that do not reject one or both hypothesis tests according to Table 4.
Mentions: After running the significance tests, we see in Table 4 that, as expected from the multimodal nature of most scenarios, only a few of them can be explained by parametrical models: #3, #2 and #18 (corresponding to the B class and to a percentage of all delay values in all the scenarios below 12%). The rest of the scenarios do not support the hypothesis. In the case of non-parametrical models, kernel and spline provide the best results, with above 40% of delay values explained. Non-rejected scenarios are displayed in Figure 5 for illustrative purposes. These results show that, for a parametrical choice, the lognormal or erlang distributions could be preferred when we are not able to detect outliers, bursts and regimes, but these are not good enough and might be applicable to very few scenarios. For a non-parametrical option, the kernel distribution should be the best election. In general non-parametrical models are able to explain scenarios from all classes (A, B, C), given their closest fitting to the particular data in the sample (which leads, on the other hand, to the previously mentioned problem of overfitting). The SoG model obtains worse results than the kernel due to its restriction in the number of modes, while the spline is also worse than the kernel due to its greater overfitting, that produces very good models of the data used for fitting but not in its goodness-of-fit with the delays reserved for hypothesis testing.

Bottom Line: This paper studies marginal probability distributions that, under mild assumptions, can be a good approximation of the real distribution of the delays without using knowledge of their dynamics, are efficient to compute, and need minor modifications on the networked robot.Since sequences of delays exhibit strong non-linearities in these networked applications, to satisfy the iid hypothesis required by the marginal approach we apply a change detection method.The results reported here indicate that some parametrical models explain well many more real scenarios when using this change detection method, while some non-parametrical distributions have a very good rate of successful modeling in the case that non-linearity detection is not possible and that we split the total delay into its three basic terms: server, network and client times.

View Article: PubMed Central - PubMed

Affiliation: Systems Engineering and Automation Department, University of Málaga, Campus Teatinos, Boulevard Luis Pasteur s/n, Málaga 29071, Spain. anagagobenitez@gmail.com.

ABSTRACT
Networked telerobots are remotely controlled through general purpose networks and components, which are highly heterogeneous and exhibit stochastic response times; however their correct teleoperation requires a timely flow of information from sensors to remote stations. In order to guarantee these time requirements, a good on-line probabilistic estimation of the sensory transmission delays is needed. In many modern applications this estimation must be computationally highly efficient, e.g., when the system includes a web-based client interface. This paper studies marginal probability distributions that, under mild assumptions, can be a good approximation of the real distribution of the delays without using knowledge of their dynamics, are efficient to compute, and need minor modifications on the networked robot. Since sequences of delays exhibit strong non-linearities in these networked applications, to satisfy the iid hypothesis required by the marginal approach we apply a change detection method. The results reported here indicate that some parametrical models explain well many more real scenarios when using this change detection method, while some non-parametrical distributions have a very good rate of successful modeling in the case that non-linearity detection is not possible and that we split the total delay into its three basic terms: server, network and client times.

No MeSH data available.


Related in: MedlinePlus