To help visualize the distributed architecture of the Internet of Things (IoT), we explore specific concrete examples of what distributed intelligence looks like in the smart factory, smart building, precision agriculture, and smart mine.
Full Article Below -
The Internet of Things has an inherently distributed, multi-layered architecture. While a few IoT applications consist of a single simple device connected to a cloud application, most have multiple layers of cooperating local devices or subsystems communicating to multiple layers of interconnected cloud applications, as described in Distributed Intelligence in the IoT. In this article we explore this concept further by looking at some illustrative real world examples.
The Smart Factory
Manufacturing plants provide one of the richest and most mature examples of distributed intelligence for the Internet of Things. Consider an automotive plant. Parts and materials from literally thousands of suppliers are delivered to the assembly plant. There are numerous complex automated steps for fabricating and assembling the chassis and the body. Let’s examine just a single one of those steps, the welding of the frame. Welding robots use numerous sensors, such as lasers, cameras, pressure sensors, ultrasound, x-ray, infrared thermography, current and voltage sensors, and much more to precisely control the welding robots’ movements, apply the proper amount of electricity at the right time, and inspect the weld for defects.
Figure 1 - Automated Welding of Automobile Chassis
These systems have a multi-level architecture as shown in the highly simplified schematic in Figure 2. Here information from a CAD system feeds the cell controller with instructions, which in turn controls the robot and welding controllers. Each of those use multiple sensors to precisely do its job. There can actually be more layers of processing on the lower layers than is shown here. For example, machine vision systems have pattern recognition engines incorporating powerful subsystems with lots of compute power.
Figure 2 - Multi-level Architecture of a Robotic Welding System
Sitting above the multiple layers of intelligence in the welding system are production line control systems that control the pace of the lines, provide mass customization, production equipment instructions, and other input to the cell controllers, and monitor the overall factory. There are plant-level software systems controlling all of this, that coordinate with other enterprise systems such as warehouse, materials, sourcing, production planning and scheduling, and other systems. This is just a small glimpse at the multiple levels of complexity for these plants and their IoT components.
Figure 3 – Example Automotive Plant Production Control (Source: Hitachi)
There is a lot of intelligence that goes into sending an elevator up and down. Encoders detect the motor’s rotation speed to control the motion and position of the elevator car. Current sensors sense the working current to adjust power. Leveling sensors transmit the floor number of each car to the control center which activates the motor and brake mechanisms to send the elevator to each desired level. Closed-loop systems continually monitor the speed and position of the elevator doors applying varied force as needed. As buildings become taller and taller, higher and higher speeds are required to ascend in a reasonable time. This requires new innovations such as Hitachi’s system that sends data about acceleration speeds to the guide rails, adjusting pressure to ensure a smooth, wobble-free ride.
The Smart Building
Modern buildings are filled with many increasingly intelligent systems such as HVAC,1 security, surveillance, lighting, elevators and escalators, communications network (phone and data), fire and safety (smoke detector, alarm/evacuation, sprinkler), plumbing, and more. In big office buildings, each is a complex system on its own.
Today’s elevators are already bristling with sensors and processors (see Figure 4 and side bar). Some newer elevators use a destination control system in which the user enters their destination floor before getting on the elevator. Some require each user to present their RFID card, which enables them to go to certain floors, but not others. These systems thus know which floor the user is currently on, which can be used in case an evacuation is needed.
Figure 4 - Simplified Diagram with Example Sensors and Microprocessors/Controllers in an Elevator (Source: Farnell element 14)
In the past, there was little communication between the elevator’s control systems and other building systems. But this is changing rapidly. Some destination selection elevators are connected to the building’s security turnstiles to direct VIP visitors or mobility-impaired tenants to their elevator automatically. There are increasingly sophisticated ‘building operating systems’ that track the movement of people into and out of the building to optimize resources, such as preheating or pre-cooling the building. One can imagine a security system that becomes aware of individuals as well, to accommodate their movements and habits. Occupancy sensors can be used not just to control the lighting, but to adjust the HVAC when a space is unoccupied, turn off displays and speakers, and alert the security system if unauthorized movement is detected. Access control systems may talk to video surveillance—for example, cameras can start recording upon door contact sensor opening, or a camera detecting suspicious or dangerous activity might lock down an area. Thus all of the systems in a building are increasingly being interconnected.
Connecting the Smart Building, Smart City, and Smart Grid
Smart buildings optimize their heating and cooling rhythms based on learning from past patterns of occupant arrivals and departures, combined with weather forecasts. This intelligence could be further enhanced by awareness of other external factors. This could include events within the surrounding city, such as the shutdown or major delay of transit systems, or major traffic problems, or a big event occurring. Similarly, the smart city might be very interested in data from the smart building, such as knowing when people are leaving the building, perhaps to better coordinate the availability of public transit resources, police, or other public resources. We already see integration of building systems to the smart grid in the form of demand response programs in which the smart grid tells the smart building when there is a current or upcoming period of high demand and the building automatically curtails its use of electricity during those periods (and gets paid for it). Thus the multi-layered architecture extends beyond the building itself to coordination with the surrounding environment and utilities.
Similar multi-level architectures are seen in most other IoT applications. The agricultural machines used in precision farming are sophisticated and complex. They have extremely precise GPS guiding the positioning of the machine, while soil nitrogen content sensors, biomass sensors, and other sensors provide real-time data to controllers deciding the rate of dispensing for seed, fertilizer, and pesticide. These controllers continually adjust boom height, nozzle flow, droplet size, and other mechanisms of the machine. The whole operation is guided by a ‘farm operating system,’ which coordinates the multiple machines, irrigation systems, and human operations, and is typically connected to weather information services, commodity exchanges, and other external systems. (See Figure 5).
Figure 5 - Precision Agriculture Example (Source: eSpatially New York)
The modern mine is becoming increasingly autonomous. There are autonomous continuous mining machines such as Joy Global’s Longwall System which uses 7,000 sensors and many actuators and controllers to coordinate its shearer with its armored face conveyor, belt conveyor, powered roof support system, and pump stations. The system can extract a mind-boggling 2,200 tons per hour while it continually transmits data with other pieces of equipment and systems both below and above ground. This includes communicating with underground haulage vehicles, as well as autonomous mining trucks and autonomous trains on the surface, all connected to control centers that could be hundreds of miles away. Again you have complex multi-layered machines integrated together in a connected and coordinated environment of many interlocking systems.
Figure 6 - Longwall Shearer
Implications for Platforms, Interoperability, and Standards
This distributed architecture has implications for the development of IoT applications and the standards needed to support them. In the past, many IoT systems were built from scratch. Increasingly they are being built from preexisting components, systems, and IoT platforms with an emphasis on speed of development and not reinventing the wheel. This will place an increasing emphasis on standards for interfaces between various devices, subsystems, modules, and software components. Because the architecture of IoT is multi-layered, these interfaces exist at multiple levels.
Figure 7 - Variety of Interfaces at Different Layers within Multi-Level IoT Architectures
In some areas, such industrial automation protocols for factory floors and other settings, there are a number of mature legacy standards in place.2 But there are also numerous newer standards under development and to-be-developed. They will also include countless domain-specific standards, such as for specific systems (elevators, automated mining trucks) and the environments they run in (digital building platforms, mining system platforms, etc.). Interoperability testing and certification standards are needed as well. Higher level platforms, such as Building Operations Systems and Plant Floor Control Systems are needed not just to provide the higher level application functionality, but also to bridge the integration gap with legacy devices and systems.
As we have seen, the concept of distributed intelligence and the tools to enable it comprise a complex and rich area for discussion and research. These represent some of the major challenges for IoT going forward and we expect to see more focus on this in the coming years.