Wednesday, 23 January 2019

AI and Analytics Based Network Designing & Planning

Recently I blogged about how Deutsche Telekom is using AI for variety of things. The most interesting being (from this blog point of view), fiber-optic roll-out. According to their press release (shortened for easy reading):

"The shortest route to the customer is not always the most economical. By using artificial intelligence in the planning phase we can speed up our fiber-optic roll-out. This enables us to offer our customers broadband lines faster and, above all, more efficiently," says Walter Goldenits, head of Technology at Telekom Deutschland. It is often more economical to lay a few extra feet of cable. That is what the new software-based technology evaluates using digitally-collected environmental data. Where would cobblestones have to be dug up and laid again? Where is there a risk of damaging tree roots?

The effort and thus costs involved in laying cable depend on the existing structure. First, civil engineers open the ground and lay the conduits and fiber-optic cables. Then they have to restore the surface to its previous condition. Of course, the process takes longer with large paving stones than with dirt roads.

"Such huge amounts of data are both a blessing and a curse," says Prof. Dr. Alexander Reiterer, who heads the project at the Fraunhofer IPM. "We need as many details as possible. At the same time, the whole endeavour is only efficient if you can avoid laboriously combing through the data to find the information you need. For the planning process to be efficient the evaluation of these enormous amounts of data must be automated." Fraunhofer IPM has developed software that automatically recognizes, localizes and classifies relevant objects in the measurement data.

The neural network used for this recognizes a total of approximately 30 different categories through deep learning algorithms. This includes trees, street lights, asphalt and cobblestones. Right down to the smallest detail: Do the pavements feature large pavement slabs or small cobblestones? Are the trees deciduous or coniferous? The trees' root structure also has a decisive impact on civil engineering decisions.

Once the data has been collected, a specially-trained artificial intelligence is used to make all vehicles and individuals unidentifiable. The automated preparation phase then follows in a number of stages. The existing infrastructure is assessed to determine the optimal route. A Deutsche Telekom planner then double-checks and approves it.

In the recent TIP Summit 2018, Facebook talked about ‘Building Better Networks with Analytics’ and showed off their analytics platform. Vincent Gonguet, Product Manager, Connectivity Analytics, Facebook talked about how Facebook is using a three-pronged approach of accelerating fiber deployment, expanding 4G coverage and planning 5G networks. The video from the summit as follows:

TIP Summit 2018 Day 1 Presentation - Building Better Networks with Analytics from Telecom Infra Project on Vimeo.

Some of the points highlighted in the video:
  • Educating people to connect requires three main focus areas, Access, Affordability and Awareness – One of the main focus areas of TIP is access. 
  • 4G coverage went from 20% to 80% of world population in the last 5 years. The coverage growth is plateauing because the last 20% is becoming more and more uneconomical to connect.
  • Demand is outpacing supply is many parts of the world (indicating that networks has to be designed for capacity, not just coverage)
  • 19% of 4G traffic can’t support high quality videos today at about 1.5 Mbps
  • Facebook has a nice aggregated map of percentage of Facebook traffic across the world that is experiencing very low speeds, less than 0.5 Mbps
  • Talk looks at three approaches in which Facebook works with TIP members to accelerate fiber deployment, expand 4G coverage and plan 5G networks.
  • A joint fiber deployment project with Airtel and BCS in Uganda was announced at MWC 2018
  • 700 km of fiber deployment was planned to serve over 3 million people (Uganda’s population is roughly 43 million)
  • The real challenge was not just collecting data about roads, infrastructure, etc. New cities would emerge over the period of months with tens of thousands of people 
  • In such situations it would be difficult for human planners to go through all the roads and select the most economical route. Also, different human planners do thing in different ways and hence there is no consistency. In addition, its very hard to iterate. 
  • To make deployments simpler and easier, it was decided to first provide coverage to people who need less km of fiber. The savings from finding optimal path for these people can go in connecting more people.
  • It is also important for the fiber networks to have redundancy but it’s difficult to do this at scale
  • An example and simulation of how fiber networks are created is available in the video  from 07:45 – 11:00.
  • Another example is that of prioritizing 4G deployments based on user experience, current network availability and presence of 4G capable devices in partnership with XL Axiata is available in the video from 11:00 – 14:13. Over 1000 sites were deployed and more than 2 million people experienced significant improvement in their speeds and the quality of videos. 
  • The final example is planning of 5G mmWave networks. This was done in partnership with Deutsche Telekom, trying to bring high speeds to 25,000 apartment homes in a sq. km in the center of Berlin. The goal was to achieve over 1Gbps connection using a mixture of fiber and wireless. The video looks at the simulation of Lidar data where the wireless infrastructure can be deployed. Relevant part is from 14:13 – 20:25.
Finally, you may remember my blog post on Automated 4G / 5G Hetnet Design by Keima. Some of the work they do overlaps with both examples above. I reached out to Iris Barcia to see if they have any comments on the two different approaches above. Below is her response:

“It is very encouraging that DT and Facebook are seeing the benefits of data and automation for design. I think that is the only way we’re going to be able to plan modern communication networks. We approach it from the RAN planning perspective: 8 years ago our clients could already reduce cost by automatically selecting locations with good RF performance and close to fibre nodes, alternatively locations close to existing fibre routes or from particular providers. Now the range of variables that we are capable of computing is vast and it includes aspects such as accessibility rules, available spectrum, regulations, etc. This could be easily extended to account for capability/cost of deploying fibre per type of road. 

But also, we believe in the benefit of a holistic business strategy, and over the years our algorithms have evolved to prioritise cost and consumers more precisely. For example, based on the deployment needs we can identify areas where it would be beneficial to deploy fibre: the study presented at CWTEC showed a 5G Fixed Wireless analysis per address, allowing fibre deployments to be prioritised for those addresses characterised by poor RF connectivity.”

There is no doubt in my mind that more and more of these kinds of tools that relies on Analytics and Artificial Intelligence (AI) will be required to design and plan the networks. By this I don’t just mean 5G and other future networks but also the existing 2G, 3G & 4G networks and Hetnets. We will have to wait and see what’s next.

Related Blog Posts:

Wednesday, 16 January 2019

5G Slicing Templates

We looked at slicing not long back in this post here, shared by ITU, from Huawei. The other day I read a discussion on how do you define slicing. Here is my definition:

Network slicing allows sharing of the physical network infrastructure resources into independent virtual networks thereby giving an illusion of multiple logically seperate end-to-end networks, each bound by their own SLAs, service quality and peformance guarantees to meet the desired set of requirements. While it is being officially defined for 5G, there is no reason that a proprietary implementation for earlier generations (2G, 3G or 4G)  or Wi-Fi cannot be created.

The picture above from a China Mobile presentation, explain the slice creation process nicely:

  1. Industry customers order network slices from operators and provide the network requirements, including network slice type, capacity, performance, and related coverage. Operators generate network slices according to their needs. Provide the network service requirement as General Service Template (GST).
  2. Transfer GST to NST (Network Slice Template)
  3. Trigger Network Instantiation Process
  4. Allocate the necessary resources and create the slice.
  5. Expose slice management information. Industry customers obtain management information of ordered slices through open interfaces (such as number of access users, etc.).

For each specific requirement, a slicing template is generated that is translated to an actual slice. Let's look at some examples:

Let's take an example of Power Grid. The picture below shows the scenario, requirement and the network slicing template.
As can be seen, the RAN requirement is timing and low latency while the QoS requirement in the core would be 5 ms latency with guaranteed 2 Mbps throughout. There are other requirements as well. The main transport requirement would be hard isolation.

The Network requirement for AR Gaming is high reliability, low latency and high density of devices. This translates to main RAN requirement of low jitter and latency; Transport requirement of Isolation between TICs (telecom integrated cloud) and finally Core QoS requirement of 80 ms latency and 2 Mbps guaranteed bit rate.

More resources on Network Slicing:

Monday, 7 January 2019

The business case of densifying LoRaWAN deployments

LoRaWAN has recently emerged as one of the key radio technologies to address the challenges of Low Power Wide Area Network (LPWAN) deployments, namely power efficiency, long range, scalable deployments, and cost-effectiveness.

The LoRa Alliance has had an exponential growth with 500+ members with the recent arrival of heavyweight members such as Google, Alibaba, and Tencent joining the alliance.

The first wave of LoRaWAN was primarily focused on large country-wide deployments led by operators such as KPN, Orange, Swisscom and many more. However, the next wave that is already coming is the arrival of private LoRaWAN deployments from large enterprises and enabling roaming for inter-connection amongst public/private networks (esp. for use cases which involve LPWAN Geolocation [8] [9]]). As the IoT deployments grow in both the densification and geographical footprint, it is inevitable that network design becomes one of the important factors ensuring long-term success and profitability of both operators and end-customers relying on LoRaWAN connectivity for their IoT use cases.

A typical example is the recent 3 million water meter contract awarded by Veolia Birdz to Orange [12]: such large-scale projects require careful network planning to achieve the required densification and quality of service while optimizing costs.

A Closer Look at Densification techniques for LoRaWAN

LoRaWAN deployments use a star topology with a frequency reuse factor of 1 which allows simplicity in network deployment and ongoing densification: there is no need for frequency pattern planning or reshuffling as more gateways are added to the infrastructure.

Compared to mesh technologies, the single hop to network infrastructure minimizes power consumption as nodes do not need to relay communication from other nodes. Another advantage is that gradual initial network deployment in sparse mode with low node density is possible, compared to mesh which requires minimum node density to operate. Even more importantly, LoRaWAN is immune from the exponential packet loss suffered by multi-hop RF mesh technologies in presence of increasing interferers and noise floor power.

Another unique feature of LoRaWAN networks is that messages in uplink can be received by any gateway (Rx macro-diversity), and it is the function of a network server to remove duplicates in uplink and select the best gateway for downlink transmission based on the uplink RSSI estimates. This allows enabling of features such as geolocation to be easily built into LoRaWAN deployment and enables uplink macro-diversity that significantly improves network capacity and QoS (Quality of Service).

LoRaWAN also supports features such as Adaptive Data Rate (ADR) that allows network server to dynamically change parameters of end-devices such as transmit power, frequency and spreading factor via downlink MAC commands. Optimization of theses settings is key to increase the capacity and reduce the power consumption of end-devices.

The optimization of LoRaWAN parameters along with densification can lead massive amounts of capacity increase in the network. In fact, the LoRaWAN capacity of the network can scale almost indefinitely with densification.

Figure 1: Actility Webinar - Designing LoRaWAN network for Dense Deployment  [1] [2] [3]

The future of LoRaWAN networks, particularly in urban environments where the noise floor is expected to get higher due to increased traffic, goes towards micro-cellular networks

How does densification lead to lower TCO for Enterprise deployment?

As the network is densified by deploying more LoRaWAN Gateways and adaptive data rate and power control algorithms are applied intelligently in the network, this leads in dramatic reduction of power consumption of end-device and thus reduction in Total Cost of Ownership (TCO) of end devices. The figures below show clearly that densification can lead to upto 10X savings in both power consumption and overall reduction in 10-year TCO for enterprise deployment. Changing the batteries require manual labor and is the cost that can significantly dominate 10-year TCO of large-scale enterprise deployment (for ex. Smart gas/water meters).

Figure 2: Battery Lifetime Improvement with densification [1] [2] [3]

Figure 3: Impact on 10-year TCO due to densification [1] [2] [3]

Densification leads to very dramatic reduction in power consumption of the end-devices thus reducing overall Total Cost of Ownership (TCO)

LoRaWAN offers disruptive Deployment Models

LoRaWAN is generally deployed in unlicensed spectrum which allows anyone to roll-out IoT/LPWAN network based on LoRaWAN. This allows three deployment models:

1. Public Operator Network: In this traditional model, the operator invests in a regional or nation-wide network and sells connectivity services to its customers.

2. Private/Enterprise Network: In this model, enterprise customers typically setup LoRaWAN gateways on private premises (e.g. an airport), and either have these gateways managed by an operator, or use their own LoRaWAN network platform.

This mode of deployment is a game changer for dense device use cases, as network capacity and enhanced QoS can be provided at marginal increased cost. It becomes possible because LoRaWAN runs in unlicensed spectrum and gateways are quite inexpensive and easy to deploy.

3. Hybrid model: This is the most interesting model that LoRaWAN allows due to its open architecture.

This is not possible or rather difficult in other competing LPWA technologies or Cellular IoT (due to licensed spectrum and absence of roaming/peering model between private and public networks). There are initiatives like CBRS and MulteFire from 3GPP Players but they are still in progress and far from maturity for large scale IoT deployments (esp. for use cases that demand 10-15 years+ battery lifetime).

In hybrid model, operator provides light country-wide outdoor coverage, but different stakeholders such as private enterprises or individuals help in densifying the network further based on their needs on their premises, via managed networks. This model enables a win-win private/public partnership in sharing the costs and revenues from the network and densify the network where the applications and devices are most present.

This model is possible because multiple gateways can receive LoRaWAN messages and network server removes duplication. In the cases where different operators/enterprises run their networks, LoRa Alliance already has approved roaming architecture in “LoRaWAN Backend Interfaces 1.0 Specification” [6] [7] to enable network collaboration.

This model significantly reduces the operator investment and offers a disruptive business model to build IoT capacity where it is mostly needed.

Figure 4: LoRaWAN Hybrid Deployment Model (source : Actility)

LoRaWAN enables Public-Private deployment that allows disruptive model for cost/revenue sharing and densifying the network where it is needed most, depending on IoT application needs

LoRaWAN densification: A Key driver for reduction in Operator TCO

When designing and deploying a LoRaWAN network, the system operator must balance the cost of a dense network (and it's served sensors) against the cost of a sparse network (and it's served sensors).

Traditional vs Opportunistic network designs

In the traditional deployment model, the operator deploys LoRaWAN gateways on telecom towers. This entails leasing the space from the tower owner, purchasing a waterproof outdoor gateway, climbing the tower to hang the gateway, and perhaps paying for additional power, zoning, permitting, and backhaul. The operator does the detailed RF propagation study and hangs enough gateways to provide coverage for the sensor locations required to provide the services he wants to provide.

Another option is to opportunistically deploy “femto” gateways in devices that the operator is already fielding. The gateways are stateless, and thus do not add much complexity to the hosting device. An 8-channel LoRaWAN reference design is mated to the host device using either USB or I2C. The options here are quite diverse. The operator can embed a simple 8 channel gateway into ongoing WiFi hotspots, power supplies, amplifiers, cable modems, thermostats, virtual assistants, or any mass-produced device that already has backhaul. The Bill of Materials adder is quite modest, the power consumption and heat dissipation are less than 3 Watts, and the size delta is roughly 7 cm by 3 cm.

Calculating the number of opportunistic gateways to provide adequate coverage for a given deployment can be challenging. The height of the gateways has a large impact on the coverage of the gateway. A gateway deployed in a 20th story of an apartment building has a much better coverage pattern than the same gateway deployed in the basement of a single-family home. Gateways deployed in WiFi hotspots mounted on power poles have a different coverage area than a gateway deployed on light poles. So, the actual number of gateways deployed in each scenario varies widely. When you complete the detailed design of each network type, you typically find that an opportunistic deployment model allows the operator to cover a given area by deploying roughly 100 times as many gateways for roughly 1/10th of the cost (when compared to the traditional 3rd party leased tower model).

Example use-case with water meters

For the rest of this analysis, we will assume that the operator needs to deploy a LoRaWAN network to service 100K water meters. Water meters represent a difficult RF propagation model. They are installed at or below ground level, must last 20 years, and suburban meters tend to have accumulations of grass and dirt collect over time. Let’s assume a North American deployment model, and we have the option of using a high power (27dBm) or a low power (17dBm) meter.

One possible design is to use a tower-based approach. In a tower-based approach, the operator typically ends up deploying high power water meters in order to reduce the number of (expensive) tower leases. In order to run at high power, the North American regulations require the sensor to send across 50+ channels, which drives the operator to deploy 64 channel gateways. Let’s assume that the average distance between a water meter and a tower-based gateway is ~3km and the sensors need to send one reading per day. Many of the meters thus operate at SF10 at 27dBm. The sensor designer includes a high-power RF amplifier, calculates the energy requirements over the life of the sensor, and sizes the battery appropriately.

Another possible design is to opportunistically deploy thousands of femto gateways into the area. The question boils down to “How many femto gateways do I need to cover the desired area?”. Working backwards from the densest possible deployment, most MSOs (Multiple-System Operator) serve 1/3 of the households in their footprint. In many urban environments, the average distance between a given operator’s subscribers is 30 meters. If such an operator could opportunistically deploy in most of those sites, they would have inter-gateway distances as small as 30 meters. For the purposes of this analysis, let’s say that the average distance between the sensor and the closest gateway is reduced from 3000 meters to 100 meters. When a sensor is 100 meters from a gateway, it can typically operate at SF7 at 17dBm (or lower). Clearly, the network designer must account for a distribution of distances between a given sensor and its closest gateway, but the overall power savings is significant.

It is also instructive to compare the overall capacity of a tower-based LoRaWAN network to the overall capacity of the opportunistic LoRaWAN network. Remembering that 100 eight channel opportunistic gateways cost about 1/10th of a single 64 channel gateway, we realize that we get ~13 times as much network capacity for 1/10th of the cost. As the sensor density increases, we could deploy additional opportunistic gateways and get ~130 times as much network capacity for the same cost as a tower-based network.

When we compare the cost to build a sensor designed to last 20 years using SF10 at 27dBm to the cost to build a sensor designed to last 20 years using SF7 at 17dBm, we find that we can save more than $10 per sensor by deploying the denser network.

So, in addition to saving a significant amount of capital by opportunistically deploying the gateways, the operator can save more than $10 per water meter by opportunistically deploying a dense network. This saves more than $1M on the 100K water meter deployment. When one layers in additional use cases, the dense LoRaWAN network provides sensor savings on each additional set of sensors. Most of the sensors do not have the 20 years requirement and thus do not save the same amount of money, but batteries are one of the primary drivers for any sensor’s cost.


This analysis is somewhat simplified, and a very large-scale deployment may require a certain amount of traditional gateway placement to provide an “umbrella” of coverage that is then densified using opportunistic methods. By densifying the network, the overall sensor power budget is decreased significantly. One could also envision a deployment model in which an opportunistic gateway is deployed in conjunction with a set of services. The operator would add IoT based services to an existing bundle (let’s say voice/video/data, thermostat control or personal assistant) and know that the sensors would be co-resident with the gateway.

What is the future of LoRaWAN?

LoRaWAN exhibits significant capacity gains and massive reduction in power consumption and TCO when ADR algorithms are used intelligently in the network. We showed how LoRaWAN networks are deployed for coverage and how network capacity can be scaled gracefully by adding more gateways.

There are already 16 channels in EU, but there have been recent modifications of the regulatory framework to relax the spectrum requirements and increase transmit power, duty cycle and number of channels [22].

Moreover, Semtech released the latest version of LoRa chipsets [23] with the following key features:
  • 50% less power in receive mode
  • 20% extended cell range
  • +22 dBm transmit power
  • A 45% reduction in size: 4mm by 4mm
  • Global continuous frequency coverage: 150-960MHz
  • Simplified user interface with implementation of commands
  • New spreading factor of SF5 to support dense networks
  • Protocol compatible with existing deployed LoRaWAN networks

The above LoRaWAN features and upcoming changes to EU regulations will allow significantly scaling of unlicensed LoRaWAN deployments for years to come to meet the needs of IoT applications and use cases. LoRaWAN capacity depends indeed on the regional and morphology parameters. As we have showed in the above results, if the network is deployed carefully and advanced algorithms such as ADR are used, there can be dramatic increase in network capacity and massive reduction in TCO. This will be one of the main factors that will determine the success of LoRaWAN deployments as the demands and breadth of IoT applications scale in future.

We also showed earlier how LoRaWAN offers innovative public/private deployment model in which operators can build capacity incrementally and supplement with extra capacity by leveraging gateways deployed from private individuals/enterprises. Typically, for cellular networks there can be anywhere from 5-10% IoT devices on cell-edge which are in outage [10]. This applies especially to deep indoor nodes (for example, smart meters with additional 30 dB penetration loss). Such nodes can only be covered by densification of cellular network which is expensive considering it is being done only for 5-10% of IoT devices. One way to address this problem is deploying private LoRaWAN on cell-edge and using multi-technology IoT platform that combines both LoRaWAN and Cellular IoT [11].

On the other hand, LoRaWAN offers a cost-effective way to augment network capacity where it's needed most. LoRaWAN gateways are very cost-effective and can be deployed using Ethernet/3G/4G backhaul with minimal investment in comparison to 3GPP small cells. This allows building IoT network in cost-effective manner and scale it progressively based on the application needs. We believe that his deployment model has dramatic effect on ROI for IoT connectivity based on LoRaWAN.

The LoRa Alliance has standardized the roaming feature, which enables multiple LoRaWAN networks to collaboratively serve IoT devices. Macro-diversity used across deployments enables operators/enterprises to jointly densify their networks, hence providing better coverage at lower costs. The future of LoRaWAN as shown below will be private/enterprise network deployments and disruptive business models through roaming with the public networks [4] [5] [6] [7].

Figure 5: Future of LoRaWAN deployments

LoRaWAN does provide horizontal connectivity solution to address wide-ranging needs for IoT applications for LPWAN deployments. However, these benefits are only possible with intelligent network server algorithms proprietary to network solution vendors

For any questions, contact the author below,

[1] Actility webinar Replay: Designing a LoRaWAN Network for Dense Deployment,

[2] Actility webinar slides: Designing a LoRaWAN Network for Dense Deployment,

[3] Actility Whitepaper: Designing a LoRaWAN Network for Dense Deployment,

[4] Actility webinar slides: Industrial IoT - Transforming businesses today with LoRaWAN,

[5] Actility webinar Replay: Industrial IoT - Transforming businesses today with LoRaWAN,

[6] Actility webinar slides: LoRaWAN Roaming Webinar,

[7] Actility webinar Replay: LoRaWAN Roaming webinar,

[8] Actility webinar slides: Multi-technology IoT Geolocation,

[9] Actility webinar Replay: Multi-technology IoT Geolocation,


[11] Actility Whitepaper: How to build a multi-technology scalable IoT connectivity Platform,


Thursday, 3 January 2019

Nice short articles on 5G in 25th Anniversary Special NTT Docomo Technical Journal

5G has dominated the 3G4G blog for last few years. Top 10 posts for 2018 featured 6 posts on 5G while top 10 posts for 2017 featured 7. In makes sense to start 2019 posting with a 5G post.

A special 25th Anniversary edition of NTT Docomo Technical Journal features some nice short articles on 5G covering RAN, Core, Devices & Use cases. Here is some more details for anyone interested.

Radio Access Network in 5G Era introduces NTT Docomo's view of world regarding 5G, scenarios for the deployment of 5G and also prospects for further development of 5G in the future. The article looks at the main features in 5G RAN that will enable eMBB (Massive MIMO), URLLC (short TTI) and mMTC (eDRX).

Interested readers should also check out:

Core network for Social Infrastructure in 5G Era describes the principal 5G technologies required in the core network to realise new services and applications that will work through collaboration between various industries and businesses. It also introduces initiatives for more advanced operations, required for efficient operation of this increasingly complex network.

This article also goes in detail of the Services Based Architecture (SBA). In case you were wondering what UL CL and SSC above stands for; UpLink CLassifiers (UL CL) is a technology that identifies packets sent by a terminal to a specific IP address and routes them differently (Local Breakout) as can be seen above. It is generally to be used to connect to a MEC server. Session and Service Continuity (SSC) is used to decide if the IP address would be retained when the UE moves to a new area from the old one.

Interested readers should also check out:
Evolution of devices for the 5G Era discusses prospects for the high-speed, high-capacity, low-latency, and many-terminal connectivity features introduced with 5G, as well as advances in the network expected in the future, technologies that will be required for various types of terminal devices and the services, and a vision for devices in 2020 and thereafter.

According to the article, the medium term strategy of R&D division of NTT Docomo has three main themes: 5G, AI and Devices. In simple terms, devices will collect a lot of data which will become big data, 5G will be used to transport this data and the AI will process all the collected Big Data.

NTT Docomo has also redefined the devices as connecting through various technologies including cellular, Wi-Fi, Bluetooth & Fixed communications.

Interested readers should also check out:

The final article on 5G, Views of the Future Pioneered by 5G: A World Converging the Strengths of Partners looks at field trials, partnerships, etc. In fact here the embedded video playlist below shows some of these use cases described in the article

In addition there are other articles too, but in this post I have focused on 5G only.

The 25th Anniversary Special Edition of NTT Docomo Technical Journal is available here.