Відео

There is already a video on Kalman Filter: ua-cam.com/video/gnJqY7s2Lhk/v-deo.html.

Unfortunately, the PPT slides are available from within the TU Dresden's domain.

The slides are based on the book I co-authored. The book is entitled: Fundamentals of Wireless Sensor Networks: Theory and Practice (W. Dargie and C. Poellabauer, Wiley and Sons, 2011).

In a wireless sensor network, the predominant direction of traffic is directed towards the sink or the base station. All nodes must send their data to the sink. A malicious pretending to be a sink and deceiving nodes to direct their packet towards it causes a sink redirection and the attack is labelled as a sinkhole attack. It is one of the most dangerous types of attacks. This attack takes places when the malicious node propagates a false routing update request pretending to be the authentic sink. Sybil attack is in general the attempt by several malicious nodes into the network. These nodes, if admitted, can congest the network by overwhelming intermediate nodes to forward spurious packets, sabotage the network by dropping packets which they are requested to forward; etc.

You should note two distinct steps we have made in getting representative samples: In the first, we expressed the posterior probability in terms of three components, namely, the prediction, transition, and measurement probabilities. Then we transform this expression into an iterative weight function, but still a posterior probability (17:30). In the second, we discretise the weight function using the Dirac delta function. Any continuous signal can be expressed by the Dirac delta function, which is what we did in the last step. This enables us to get a countable number of samples which represent the continuous (and often a complex) posterior function.

​@@waltenegus Thank you so much for the help. I am confused about the transformation from steps 1 to 2. In particular, I wonder how the expression of posterior probability P(X_{0:k} | D_k) is transformed to a expected function E{f(x)}. It seems to me in video(14:33) that P(x_{0:k} | D_k) = E{f(x)} in some way, but I can make sense of it myself and would much appreciate it if you could clarify it further.

Another question in 21:23, why q(x^i_k, x^i_{0:k-1 | D_k) equals q(x^i_k, x^i_{0:k-1 | D_{k-1}) ? I noticed the measurements D change from D_k to D_{k -1} in the denominator and wonder how to derive it.

This expression refers to the number of routes leading to the node and yes, it can be (indeed, it must be) greater than one. To condition it between 0 and 1, there must be a normalisation factor.

Thanks for your answer!

If you are intending to monitor such parameters as relative humidity, temperature, moisture, radiation, and so on, you may consider RTL or any of the tree-based routing protocols, which are relatively easy to deploy. Most of them are implemented for the Contiki platform (github.com/contiki-ng/contiki-ng). The question is how big is your network? Alternatively, you may cluster the network (ieeexplore.ieee.org/abstract/document/9082144) and use the cluster-heads to collect data ifficiently.

Part I and II are on Minimum Mean Square Estimation (linear and nonlinear), which are the basis for Kalman Filter. You will find these lectures in the same playlist.

You are right; I apologise. I am planning to release a new version this semester. Bear with me. But if you have any question, let me know.

@@waltenegus Great! Thanks Professor.

Unfortunately this is available only from within the university's website. The slides are based on the book: Fundamentals of Wireless Sensor Networks: Theory and Practice.

Difference between MAC and Routing: Suppose you are in a class room, with many students. You would like to address your professor. What do you do? You wait until all are silent (otherwise, there will be interference). This action is equivalent to the decision of a MAC protocol. Nodes first listen to communicate. Suppose now you wish to pass a book to one of your peers sitting at another end of the room. What do you do? You pass it to one of your neighbour (which you choose) and your neighbour passes it to one his neighbour (which he chooses) and so on, until the book arrives at the destination. That is equivalent to routing -- choosing a neighbour and passing the book. Two decisions at once.