Among many things presented by Dr.Filippo Malandra, there was this point that the IoT devices meet the normal LTE network for access to Internet. IoT devices do not have their own networks. This can easily add more load to an existing base station (eNodeB). Problem with the eNodeB not able to withstand this kind of incoming data communication and resource request can in turn demand a higher resources capable base station upgrade. This is a possibility.
IoT Devices can be predicted to infiltrate the urban landscape more than a rural area, because the wireless ecosystem is pre-exisitng for these devices to thrive. Also, according to United Nations 68% of world population will be in urban areas.
North America is less urban than other places in the planet
Hence, it may be easier to deploy and control IoT devices that are expected to be in urban areas. For example the San Francisco Bus Signs, which are LTE enabled but due to its display eats more power for working that an IoT expected to. The wonder of having all the data in place is the tech-enablement achieved, for example with some effort it is a do-able thing to have the bus stop sign board at home. But the growing urban areas in the world are not tech-enabled? And who owns the right data in all these upcoming and existing non-tech enabled highly urban areas. If the taxpayer does not have access to public utility then innovations for a smart city may be limited.
It is also seen that the area of urban areas keep increasing. In an urban environment we can expect anything from bus stops, traffic lights, street lights, smart meters for water, gas, electricity etc (may be electrical vehicle charging stations also) which are stationary and GPS trackers and other vehicular accessories like Electronic Logging Devices, Vehicle Gateways and so on which are mobile. A traffic passing through the city can be expected to have atleast an UE ( User Equipment, a.k.a cell phone with LTE internet access , in common parlance) to navigate and / or communicate. There are mobile and stationary users at any given point in a network with IoT and UEs. IoTs are mostly, low power communication devices that periodically / at a set schedule / at an alert trigger communicate with the base stations (a.k.a cell towers). If a planned roll out of IoT to enable smart cities, that can be expected in most smart city projects, is performed then the data traffic is mostly predictable with current urban cell usage patterns. To give an example, Smart cities M2M (Traffic Characterization and Performance Analysis) was demonstrated by Dr Malandra for the city of Montreal, Canada. If this is for a city Dr. Amarjeet Singh is working at a building scale for building utility monitoring.
However, lack of co-ordination brings headless chicken phenomenon, as elucidated in this article “Your Navigation App Is Making Traffic Unmanageable“. The workaround is possibly a collaboration with urban decision makers and technology provider. Here, IoT vs IoH (Internet of Humans) pose a management challenge. Technologically it is feasible to have different networks for IoT and separate it from IoH (Internet of Humans). Like other technologies that were initially restricted to specific groups, the glory of the stupid wars, that propelled technology race and gave everyone access to cell phone and internet, an open to public use network is preferable. Money will follow. Services will improve. But unlike, posing a direct challenge to the current traffic management system like the article mentions above, IoT and IoH can co-exist for a symbiotic relationship, with a feedback loop to enhance and enrich each other. IoT enables services for the IoH and IoH can provide data to be mined for patterns. These patterns of traffic, can then be studied for predictive modeling. Event alerts of scheduled programs such as AT&T Park Baseball event still needs to gather the incoming traffic patterns to cater to the imminent users of the local network. Where more traffic is expected to congest network, network congestion mitigation programs are already available in terms of mobile eNodeBs to cater to the increased need.
Machine learning and AI predictions on data mined help some prediction as to what would be the ideal time to start operating the base station towers. It is economically important due to the energy usage of these stations. Based on subscribers quality of service experience base stations may be added or removed. User needs may be directly related to the population density and assuming market penetration data availability and existing number of registered user nodes in an area. Data-driven approach helps in putting up towers. Assuming all of this is taken into account. What is the best data on urban population? Is it census survey, count number of registered subscribers and then predict the variance in subscribers and requirement? That task seem to ask for in-depth analysis and optimization. Alright, then that is Internet of Human Things. Ever dynamic and moving. Machines are used to do repeated tasks without boredom, so machines must be predictable in data usage patterns. It must be fairly certain to predict M2M traffic data.
Consider the case where an incoming fan-car having 4 users connected to same network looking forward to watch their team’s match. When they start from their social media data, it can be predicted that they will be going to the stadium. Their start and endpoints are known. Number of users entering the new base station is known. From their drive speed and distance, time of entering the new base station is predictable. Their resource requirement (4 user equipment) is predictable, because 4 UEs traveling together at same speed must be in same vehicle. Location is also predictable by cell signal and received signal strength indicator (RSSI values).
With multiple data sources channeled and used for training and testing on-demand base stations can be optimally brought in started. From multiple datasets such as current traffic data, previous data, IoT network data, and on demand LTE data, data-driven optimization can be arrived at. Optimization problems can fall under optimizing each or all of the following – (1) Time of travel (2) Distance to travel (3) Optimize maximum permissible number of users on a road , just like rooms in university have Fire Marshal‘s code. (4) Decide who will be routed normally and who will be re-routed. This provides an interesting service in terms of (5) priority routing for Government heads, VIPs or possibly commercialize the priority routing through E-ZPass kind of system. This will also enable transportation managers to look at improving and expanding routes and roads based on data and have better road-routes and better traffic usage. Thanks to all the network connected devices feedback information on usage patterns and effectiveness of implementation will be an API function call away.
Multi modal transportation scheme was explored in Smart Columbus project. But the fun of being human is to be unpredictable. There may be group think. What if everyone decided to use bikes and congest the bike traffic lanes instead of cars? Internet of Human Things can disrupt traffic and cause confusions in machine predictions. As mentioned in this article Your Navigation App Is Making Traffic Unmanageable. Human dependence of traffic apps will cause problems when it does not take into account number of users of the app and physical road traffic bandwidth, on top of which the city was never planned to have much traffic in residential areas. So, how do we know which is better human learning or machine learning.
Machine learning is only as good as the the quality of data fed in and the amount of training it can undergo. It is a good argument when a self driving AI car can claim that it has undergone training with more number of hours of driving than humanly possible. A robotic traffic pattern enabled by self driving cars will definitely add safety to the roads. There is no random acceleration or breaking. There is no random decisions to visit places in yet another random fashion like humans do. Good urban infrastructure solves the problem that it created in first place, by bringing people in without planning. In planned cities if resident population is controlled then IoT traffic and IoH traffic may look very similar due to its limited mobility. It will not be like a random platoon of motorbikes passing through the city to celebrate “motorbike day”. Model predictions will fail miserably if it tries to follow traffic patterns of college student’s motorbikes in India. Just like a new born, machine learning needs a cradle to learn and what is better than letting it run on less dense cities in North America. It will no doubt grow better everyday in predictions and be adopted elsewhere as other urban areas start to adopt this “humanly-challenged” machine learning algorithms and associated platforms. When Internet of Things meet Internet of Humans on the same network challenges are interesting and solutions give rise to new services.