One of the things I’ve noticed while working with the rail industry is that the relationship between the rail and telecommunications sectors often isn’t very good. I believe this is detrimental to both sides. Part of the problem is a simple lack of understanding of how the two sectors operate and how they see the world. I thought it would be useful to write a few blog posts to explain the state of the art to help improve mutual cooperation.
Passenger Internet access on trains (aka WiFi on trains) is very popular with users and demand for capacity keeps growing. Getting enough data bandwidth to a moving vehicle containing hundreds of users is a massive technical challenge. This post is going to look at the technical approaches to perform the data backhaul from the train to the ground communications networks and offer some thoughts on the future direction.
The common architecture for train Internet access is to bring the on-train data to a gateway that then connects via one or more radio technologies to the ground. In rail-speak the gateway is normally called the “mobile communications gateway” (MCG). There are several MCG vendors but the two biggest are Nomad Digital, who pretty much created the market, and Icomera. There are three wireless technology options for train to ground backhaul: satellite, cellular and trackside systems.
In this post I am going to start by looking at cellular backhaul and then talk about satellite and trackside systems another day.
Using cellular systems to perform the backhaul is the path of least resistance in most rail scenarios and is the dominant solution in the market. Most countries have reasonably good cellular coverage and the physical requirements on the train for antennas and other equipment are not too demanding (though one thing you quickly learn is that making physical changes to rolling-stock is never easy). The major challenges with cellular have been dealing with coverage holes, obtaining enough bandwidth and international roaming for cross-border trains.
The approach generally taken to deal with coverage and capacity was pioneered by Nomad Digital. This involves using multiple cellular modems in the MCG each with a different subscription, typically from different mobile operators. The MCG will multiples data over all the available cellular connections and make the net total bandwidth available to the train. This is done transparently to the on-train users. In areas of coverage by more than one operator this provides the maximum bandwidth by aggregating several operators together. In areas of sparse coverage it means that the train can still maintain a connection even if just one of the serving operators has coverage.
This multi-operator aggregation model has done well but it also creates some problems. By allowing the rail industry to “make do” with inadequate cellular coverage and standard cellular subscriptions it has postponed the establishment of a more strategic relationship between rail and telecoms which could create better long-term solutions. The aggregation model also seems to have the potential to conflict with developments in the cellular industry that are starting to appear in standards and deployed networks.
As we have said the MCG makes simultaneous use of multiple network operators as a way of maximising coverage and bandwidth. This works best of there are several independent cellular infrastructures available in the operating region. Now operators are increasing moving towards a shared infrastructure model the number of really independent cellular infrastructures is going to be reduced. Multiple connections to operators that share the same infrastructure do not actually improve coverage or available capacity; they just compete with each other for the same resources. Use of multiple connections is also a way for the MCG to gain access to more than one cellular channel but in developments of LTE we already see that LTE modems can aggregate channels in to a single connection. Again we have the possibility of different subsystems independently chasing the same resources.
Another role of the MCG is to make the best available backhaul connection by combining or choosing between multiple radio technologies where they are available. As the cellular industry starts to adopt the “always best connected” model the cellular modems will increasingly act independently as a broker between multiple radio technologies. The mobile operator will configure selection algorithms for the cellular modems. The MCG will be independently configured with selection algorithms between its modems. The questions of who is in charge and whether the selection algorithms work harmoniously together are interesting.
Lastly, we come to roaming. In the domains of M2M and internet-enabled appliances we have seen a lot of regional or global deals with mobile operators where there is no real cost penalty for roaming. Unfortunately the rail sector seems to almost entirely rely on normal national subscriptions. MCGs may include “geo fencing” techniques to swap networks and subscriptions automatically during cross-border journeys though I gather these have had some technical problems. It is symptomatic of the disconnect between rail and telecoms that the rail sector chooses to solve roaming problems in a way that avoids engaging the mobile operators. It will be interesting to see how the planned removal of European roaming fees changes this model.
So where do we go from here? The MCG is clearly going to remain the hub of train communications and will retain the role of brokering and multiplexing between different radio connections. However as cellular systems become more capable of multichannel and multiradio operation and as the cellular industry moves towards more shared infrastructure then there needs to be more attention paid to ensuring that the behaviours of the different system components are complementary rather than conflicting. To some extent the MCG is a work-around for the difficult relationship between the rail and telecoms sectors. It would be nice to see the telecoms sector contributing to solutions, for example in the roaming area which has already been well addressed for M2M.