Smart Industry

This will collect information about the application domain of smart industry.

Smart Manufacturing

The twentieth century was dominated by the growth of sophisticated mass production. In the twenty first century, advances in technology means that the stage is now set for the emergence of mass customisation of products tailored to each customer's specific needs. Mass production will still be needed for standard components that are needed in bulk, however, customers will increasingly expect bespoke products that are produced locally and quickly to order. This emphasises the need to be able to rapidly reconfigure manufacturing equipment to produce bespoke products at low cost, making optimal use of resources to fulfil the order book. This requires vertical integration for flexibility, and horizontal integration across the value chain, along with open markets of services. The Web of Things as a platform of platforms is well suited for realising these aims by enabling the integration of existing platforms and protocols.

The traditional view of digital automation for production has three levels:

Informational
Control
Field

The field level deals with the materials and the sensors and actuators associated with the tools that operate on them, for instance motors, valves, relays, position sensors, bar codes, RFID tags and so forth. There are many field level communication technologies, and the costly investment in the equipment and expertise make it important to be able to integrate them as part of smart manufacturing systems. The sensors and actuators are embedded as part of field level equipment such as programmable logic controllers (PLC), robots with articulated arms and fixed or interchangeable tools, and computer numerically controlled tools (CNC) e.g. mills and lathes. New techniques including 3D printing and laser cutting.

The control level monitors and controls devices at the field level. This includes dealing with alarms e.g. when a sensor reports a temperature exceeding a critical level, or a limit detector sensing something at the end of its travel. The control systems need to be able to respond in real time to ensure safe and efficient operation. The informational level deals with higher level concerns, e.g. enterprise resource planning (ERP). This can include virtual representations of the plant, and the ability to model the various components as needed to plan how to optimally reconfigure the equipment for particular purposes, and the means to run simulations, to carry out data intensive analytics and so forth.

Web of Things and Smart Manufacturing

The things in the Web of Things stand for physical and abstract entities. Things are exposed to applications as software objects with properties, actions and events. Things also have names in the form of URIs as part of W3C's Resource Description Framework (RDF). This is used for expressing the semantics of things, the data models that things expose to applications, the data and metadata, including the relationships with other things. When an application wants to interact with a thing, it asks the server, that the application is running on, to construct a software object given the URI for the thing's data model. The thing could be local to that server or it could be remote, in which case the server will construct the object as a proxy for the remote thing. Applications are decoupled from the underlying protocols and communication patterns. This allows servers to use the protocol which best fits the specific requirements. The metadata allows servers to determine how to communicate with each other. The architecture scales from microcontrollers to cloud-based server farms.

Industrie 4.0 Components = Things in the Web of Things

In principle, Web of Things servers could be integrated into the microcontrollers for the sensors and actuators at the field level in manufacturing plant. In the near term, we can expect that companies will stick with the existing field level hardware and protocols, leaving the integration of the Web of Things at the control level and information level. Further study is needed to examine how the Web of Things can support the variety of field level protocols. OPC-UA should be a good fit as its information model is analogous to RDF, and it supports properties, actions and events, albeit with slightly different terminology. OPC-UA currently supports a binary protocol, and Web Services over XML & HTTP. However, you can provide your own custom protocol bindings. OPC Foundation specifications are only available to its members.

Automated Negotiation of Contracts

The idea here is to support automated negotiation of contracts in a value chain for markets of services. An example is 3D printing a replacement part for a faulty aircraft. The supplier of the part is say in Europe, whilst this particular aircraft is grounded in an airport in Australia. The supplier locates a company near the airport that is capable of printing the part to the required quality. To protect the supplier’s IPR, the printer has to guarantee to discard the part design once it has been printed. This is analogous to protecting the content of a streaming movie.

The value proposition is that it would be quicker to produce the part locally rather than ship it across the world. Automated contract negotiation covers the price, terms and conditions, and is key to enabling value chains in open markets of services where manual negotiation would introduce unacceptable extra costs and delays.

One approach is to set up an automated auction where the aim is to select the best supplier from a set of competing suppliers that bid for the contract. Another approach is to model negotiation between a given supplier and a consumer as a sequence rounds, where the aim is to find the best solution that is acceptable to both parties. In principle, the two approaches could be combined for an auction involving a sequence of rounds that progressively prune the set of suppliers.

This is ripe for the application of game theory. Players have their own goals, some of which are disclosed, either up front or during a specific negotiation round, and others which are kept confidential. There could be “red lines” for things that are not open for negotiation, for instance particular liability clauses, or a minimum price offered by a supplier. Other factors could be variable and subject to a model that scores a combination of choices, e.g. for alternatives for materials, finishes, colours, shipment and so forth. The sequence of rounds ends with the consumer being satisfied with an offer by a supplier, or a failure to reach an agreement. The approach could involve human-computer cooperation, e.g. asking a human to rank alternatives. The need for the human involvement could be progressively reduced through machine learning over a series of negotiations, i.e. by learning the evaluation criteria based upon what the choices that humans made.

Curiously, when people try taking the part of a supplier or consumer in a negotiation with a machine, humans tend to lie to the automated agent, in order to pretend to be tougher than they really are! However with properly designed agents, it would always be a mistake to tell such lies. In essence, most people would be unlikely to be as effective negotiators as automated agents whose performance has evolved over many negotiations.

A caveat is that automated agents would have problems evaluating offers where the supplier has included something that the agent is unfamiliar with, and hence unable to score. Humans can win here since we have access to the very broad set of knowledge about the world known as “common sense”. This motivates human-machine collaboration as a basis to learn about new features.

Negotiation can be considered as a non-cooperative game where players are assumed to act rationally. This makes the Nash equilibrium relevant — the situation where each player uses their best strategy given what the other players are doing. This removes the unilateral incentive to change since if would lead to a worse outcome for that player. A google search on “nash equilibrium and automated negotiation” turns up several interesting hits.