At a metropolitan train station

A metropolitan train - station with three main terminals and 1000 LTE user devices moving around an 1 km2 area. Incelligent enables 68% faster downloads, 18% less energy consumption and 20% more capacity.

FigureTrainStation small

Figure 1: Train station network topology

Situation

A small portion of a LTE (4G) cellular network (1 km2 area urban area type) and ~1000 LTE user devices of passangers. These devices are being served by a tri-sector LTE eNodeB macro base station and also by 3 Pico-cell hot spots (each of the hotspots having 9 pico-cells, thus resulting in 27 pico-cells in total) that are placed in a triangular manner at the center of each eNodeB sector, as shown in the figure.

In this use case ~60% of the users are passengers, therefore their mobility is highly influenced by train arrivals and departures in each of the sectors. The rest of the users are not influenced by the trains therefore moving arround the area more ore less randomly. The users are moving at a speed varying from 3 to 10 Km/h. 

In this situation, a periodic mobility phenomenon appears with part of the population moving in-between the pico-cell hotspots. Unless properly handled, this event may lead to a congestion problems appearing in a round robin fashion to the 3 sectors because of the high concentration of users that request for internet access.

Results

Incelligent learns the user mobility and network usage patterns and within some minutes it reaches very accurate predictions of them. After this point it knows how the network state will evolve and how to optimize service quality and network operation. In this use case Incelligent applies traffic steering between the pico - cells and the relevant macro - sector. The knowledge-based optimization ends up with the following achievements:

Less handovers and greater stability: the knowledge-based SON function has the ability to prepare the network proactively and shortly before the predicted phenomenon takes place (fastest convergence).

Less power consumption: power (and thus energy) consumption of the network is reduced by 18% (a saving of approximately 4kWh per macro cell per day) due to the more efficient utilization of cell layers in time and space.

Enhanced QoS/QoE: the predictive offloading yields higher average UE downlink throughput by 20%, thus faster download (less average file download session duration).

Moreover, the knowledge-based optimization enables a more stable operation, avoiding oscillation among transmission parameter values.