Wireless networking

PAUL HAVINGA: Hello. My name is Paul Havinga. In this lecture, I will talk about network aspects of smart logistics. Smart logistics uses smart objects like sensors, localisation identification, so in this section, I will discuss about the basic principles involved in networked smart objects, with the focus on routing. How can you transmit messages from one device, one object, to another object. But first I’ll talk about the computing evolution. As you all know that computing started in the ’60s, a bit before, with mainframe computing, large computers to execute big data, big processing applications. And then the computing evolved into desktop computing and the internet in the ’80s and ’90s– one computer at every desk to do business and personal activities.

At the start of the century, then your biggest computing and wireless networks evolved. And then computing and networking became, basically, part of daily life objects like smart phones, smart running shoes, et cetera. And you have numerous computing devices in every place and every person. And it becomes an invisible part of the environment. Millions of desktops versus the billions of these embedded processors, embedded network processors. So this technology, the hidden technology, is now entering logistics. And this is a big building block for smart logistics. So what do we have? Well, basically, you’ve got a network that is formed when a small- set of small sense devices are deployed in an application and they cooperate for sensing a certain physical phenomenon.

So basically, you combine a sensor or identification with some smartness, a processor, and a virus communication device. And this then becoming a smart object, sensing, processing, and communication. In smart logistics, in wireless sense networks, Ubix computing, it’s about cheapness. It has to be more bits per URL, per nodule, so energy. And long range it’s not important. It’s about you need to connect to your neighbour. So short range is sufficient. Also reliability is not really the crucial part. One connection can fail, but overall system should work. And luckily, since we’ve got many smart objects, you have redundancy. So you can live with some failures, with some faults in the network itself. So the network concept is also a bit different.

It’s not one base station that helps all the others. Now it’s about a multi-hop network. So you’ve got all smart objects and they communicate to each other and they help each other. And this is also one of the basic mechanisms. So the limited ways of various communication to save energy is basically also the mechanisms that will be used. So this is a multi-hop network. So what you do is you send packets to an intermediate node and then this intermediate node will forward this packet until it reaches the final destination. So this is a store and forward multi-hop network. This sounds simple, isn’t it? No, it’s not, because those links are not high quality links.

They are time varying, they’ve got a low bandwidth, they are possibly mobile. And also, all these smart objects are power constrained. So what do you need to do with this multi-hop network? Well, we need to profile connectivity. But you also need to deal with the locals of memory, bandwidth, power. And also the number of smart objects is high, so it has to be scalable. And you also have to make sure that a link failure will cause that whole system to fail. So busy it’s all about routing. So we need to route a message from one place to another place. Now, the fairly simple thing to use is just simple broadcasting. Just flood the whole network and then it will receive.

But a more advanced mechanisms, like gossiping, hop count, on demand, pro-active into your capital. And I will address those very briefly in the next. So flooding– that was the basic mechanism, the easy one. So in flooding, every node that receives the packets simply retransmits it to all its neighbours. So it’s simple, but it’s not efficient because many urgent messages are propogated unnecessarily in a distant scale. A fairly easy end of this is called gossiping. So not broadcast to everybody, but only with a certain probability you will retransmit your data to your neighbours. And this retransmission can be based on simple probabilistic mechanisms, but also based on priority or importance of the message.

So this makes it that overhead is really reduced and it also scales better. And this works very well with any unstructured network in which you don’t have an infrastructure, but your use depends completely on your neighbours, on your neighbouring smart objects to forward the data. But since it’s based on probability, it’s non-deterministic, so there are no guarantees that the message will be delivered at the end. To be sure that you will have some guarantees, you need to have some more advanced routing mechanisms. In pro-active routing, you establish routing tables in every node in the network. And you continually evaluate the routes.

So basically, the tables contain the routes or the path from a node to the gateway or even to another node. And this can be based on different objective, functions, so like the shortest path, or the most reliable path, or the most energy efficient path. So this is fairly advanced and it gives you a more first-on network mechanisms, so you can apply different objective functions. This pro-active routing is also very fast, because whenever a route is needed, it is already there. It is already in the tables so you can use it immediately. It’s also much more efficient because there’s no need for frequent broadcasting. In case the network is not static, you can also use an on-demand mechanisms.

This is called reactive. So you build a route on demand, so if there’s a need for it to submit the data, you discover the route and then you transmit it. So you’ve got an on demand route discovery, and there’s no need for periodic route advertisements. The advantage is that this is more suitable for changing dynamic networks with mobile nodes, but also, on the other hand, it requires a longer delay. Of course, you need to build the routing first. So in lecture, you’ve learned about some networking aspects of smart logistics and about smart objects. And I concentrated on routing mechanisms because you assume that there will be many smart objects in any smart logistic situation.