Complementing its CTO-class consulting services, Stratum 4 offers clients an advanced computational tecchnology framework – an asynchronous, event-driven, distributed computing architecture appropriate for designing and implementing decentralized industrial organizations (DIO), cognitive computing systems dedicated to the automation and control of industrial value production processes.
Note: Material presented here is treated in greater detail in a book, the second edition published in 2012, entitled Creating Rational Organizations - Theory of Enterprise Command and Control. Hardcopy is available here. A downloadable PDF version is available in the Download section.
Cognitive (“cybernetic”) systems provide networked computational platforms supporting analytic and artificial intelligence applications. They enable sophisticated algorithms capable of sensing, analyzing and maintaining awareness of the evolving states and behaviors of real or synthetic processes, including the environments within which those processes operate, and the system’s own operational performance. Operating alone or in concert, cybernetic systems continuously monitor, analyze and regulate the processes responsible for value production.
Supporting our economy’s inexorable march towards improved productivity through automation, intelligent automation systems provide the technical means to achieve increased levels of situational awareness and the ability to react in a timely manner with predictable performance under a wide range of operating conditions.
Such reactive behavior requires a process governance model (DIO) qualified by policies (ethics, rules of engagement, smart contracts) and availability of shared resources (dynamic resource management). As a critical element of flexible automation (industrial robotics, building automation, traffic management), adaptive but stable behavior is essential in tracking and [machine] learning the essential states and behaviors of evolving processes.
Consequently, there is a continuous drive for increased level of automation of identifiable and quantifiable units of value production on behalf of oorganizations in such diverse segments as manufacturing, transportation, civil and information infrastructure, healthcare, military command and control, smart buildings and the management of the environment.
As intelligent systems proliferate, the volume and variety of data they produce and consume expands exponentially. As a result, the science, technology and professional practice of cognitive computing is driving and is aided by advancements in data science – analytics, high-performance computing and high-fidelity modeling, simulation and visualization services.
Applied analytics is an essential capability underlying advances in intelligent automation and is a key driver of innovation in the digital economy – as disruptive to organizations in the next decade as social media was in the last.
As a professional consultancy, Stratum 4 provides clients with analysis, design and development services needed for creating, validating and deploying intelligent products and service systems. This is the domain of applied computational science, supporting applications of embedded and distributed computing technologies, data science, and creation of mathematical models needed to characterize, analyze and develop agile, aware, adaptive, secure, scalable, policy-based and highly available systems.
Cognitive systems include, among their more salient features, the requirement to establish and maintain situational awareness of the evolving states and behaviors of processes responsible for value production. In this view, an enterprise is a computation defining a “quantifiable unit of value production,” however denominated and whether in the form of a service (e.g., accounting), a robot manufacturing cell, a crewless (“ghost”) cargo ship, an IoT field instrument, or an organizational management entity (e.g., a commercial business unit).
Viewed as a computation, enterprise value production takes place in cyberspace-time, an abstract 4-dimensional computational space having cyberspatial and temporal extents. Cyberspace is a 3-dimensional subdomain delineated by geospatial, infospatial, and sociospatial coordinates. Geospace is typically expressed in earth-centered coordinates (e.g., GPS, Zip or postal addresses). Infospace is typically expressed in Internet coordinates (e.g, IPv6 service access points, Twitter and Facebook accounts). And sociospace is typically expressed in terms of enterprise accountability hierarchies (e.g., organizational charts and authorities). Enterprise may be located anywhere in cyberspace. Consequently, enterprise value production is an inherently distributed computation. In this technology framework, instances of value producing computations equate to the work of specific cyberspatial objects (CSO).
The value of work produced by cyberspatial objects is predicated on their meeting objectives summarized in specific value propositions offered to other CSO. CSO govern value production by exercising their sociospatial authority to take decisions at the intersection of two value chains, a supply chain connecting to allied CSO producers and consumers and an asset chain connecting superior and subordinate CSO into an accountability hierarchy. Neighboring CSO thus connected form a federated enterprise, defined by a meta CSO, with a higher level or compound value proposition.
CSO may contribute to multiple federations. Federation membeship requires that CSO have unique identities and that their value propositions be discoverable in cyberspace. When discovered, CSO may bind (i.e., be contracted), either statically or dynamically, to provide their services. This requires CSO to strive to achieve their own and the federation’s objectives. If a CSO is a member of multiple federations, the resultant structure is that of a 3D lattice. As a group, a federated enterprise trades through multiple supply chains, typically intersecting along a single asset chain in each federation. To meet value production objectives in each federation, participating CSO must reliably context-switch among their respective obligations, scheduling to meet multiple deadlines while preserving the integrity of individual and shared value propositions.
A single CSO may operate in multiple federated enterprises, context switching between the value propositions defining its role in each.
As a sovereign and semi-autonomous unit of value production, a CSO may choose to cooperate with other CSO, collaborating through messages typically delivered via Internet protocols (e.g., TCP/IP). In this cyberspatial framework, a CSO may be stationary or mobile in one or more dimensions. If mobile, the services (c0mputations) it offers other CSO may be time varying according to its position, velocity or acceleration. For example, the value proposition offered by an intelligent IoT device designed to provide continuous measurements of temperature, pressure and humidity may vary depending on its direction, elevation and velocity through geospace. If in motion, its infospatial address may also change (e.g., satellite server address).
If a CSO is stationary, its output are defined in a relatively static contexts, and readily validated. If in motion, however, the value of its outputs are subject to more stringent curation and calibration to compensate for absolute or relativistic changes as it moves through geo-, info- or possibly sociospatial regions. In both cases, a CSO’s inputs, taken from other moving CSO associated with supply and asset chains, may require special filtering and analysis to compensate for motion effects. From a cybersecurity perspective, determining the identity and privileges of a CSO in motion presents additional technical challenges. Further, the supply and asset chain services offered to allied CSO may become restricted or unreliable for periods of time when motion carries a CSO to certain cyberspatial locales.
Security in Cyberspace
Security in cyberspace, required of Cyber-Physical Systems (CPS), is a challenging multi-level system architecture and implementation problem. There are important issues of security at the physical level, where sensors and actuators, attached to physical processes, directly measure and affect the states and behaviors of physical processes protected to various degrees by physical security. There are critical security concerns at the automation level, where a CPS must maintain predictable performance while providing distributed time-critical, high-availability and safety-critical process control. And there are challenging compound security issues arising from multiple interconnected CPS, linked through supply chain networks, that must operate as an integrated System-of-Systems (SOS), typically under decentralized control policies. At each level, security architecture (i.e., policy and mechanism) have functional, structural and performance requirements.
The basic computations of a CSO involve the monitoring and control of real (e.g., physical) or synthetic (e.g., economic) value production processes. This regulatory activity is traditionally implemented as a feedback-control loop involving achieving and maintaining awareness of the states of a process and its encapsulating environment and to identify any changes to situations affecting its performance – situation assessment, SA. Subsequent to identifying a change in the process’s situation, selecting and resourcing an appropriate response – response plan generation, PG. The final stage includes authorizing, scheduling, executing and monitoring the performance of active response plans – response plan execution, PE.
Considered an always on regulatory process, the cyclic CSO computation defines an information processing pipeline, or CSO service system. Opening the feedback loop and serializing its three primary services defines a CSO Information Processing Pipeline.
As a service system, CSO exhibit several technically challenging design requirements. It requires processing multiple input data streams sourced from multiple entities (e.g., IoT data steams, web services), each producing different data types, in different quantities, arriving at different rates, with some synchronous and some asynchronous with respect to the production process under control. The SA service must process these separate input streams, making them coherent in time, space and data quality, and within the resultant dataset to search for and identify the occurrence of events of interest.
Situational awareness requires time-critical detection of specific event signatures, patterns from within and across input streams. Recognition results in selection of predefined, albeit generalized response plans, or courses of action (COA). COA are prototypical, developed in conjunction with the authoring of event signatures, as general responses to specific events, defining essential plans and tasks, but with only generalized policy specifications and resourcing requirements. Subsequently, selected COA are sent to the PG service where context-specific policies (i.e., rules of engagement, or operational ethics) are established and needed resources are acquired and assigned.
The PG service takes COA as input and, following policy compliance checks and assignment of adequate resources, makes them executable as plans of action (POA), subsequently forwarding POA to the PE stage for execution. PE is responsible for overseeing all currently executing plans in order to optimally schedule and authorize the new POA, converting and launching it as an executable plans of record (POR). Optimal scheduling seeks to guarantee maximum overall value production by executing plans to meet their individual and ensemble deadlines. In execution, POR and their respective tasks directly or indirectly affect other value production processes under the control of the CSO. The impacts of executing POR on value production, seen in the output steams emitted by the service system, are subsequently observable in SA input streams, thus closing the regulatory control loop.
The value produced by a CSO action (e.g., completion of a PE task) is a function not only of what is done, but when it starts and when it completes. This is imperative in time-critical systems, where a value production is considered real-time to the degree CSO computations meet specified deadlines as scheduled in the CSO’s PE stage.
For a given situation, the decision to respond presumes the POR completes by a deadline that maximizes the value produced. Otherwise the response may have minimal value, the situation may have passed, or been preempted by a more critical situation. For the purpose of “optimal” scheduling, the time-value of CSO actions must be computable and predictable in advance.
Shown here is a generalized time-utility function, where the utility U(t) of a computation ranges in value from Umin at time t_start, grows to Umax at t_critical, then decays to Umin by t_end. Clearly, to achieve maximum utility, a CSO value proposition would need to complete by the deadline at t_critical.
As shown below, there are many possible time-value functions. (d) is perhaps the most intuitive, specifying that a computation’s value is zero unless it completes between t2 and t3. (b) specifies that a computation ‘s value is maximum only at time t3. And (f) shows that a computation’s value increases over time, but is governed by different, increasingly intense value functions.
Clearly, scheduling a single CSO in response to a single situation with a single computation (value proposition), one defined in a static context, is relatively straightforward. The situation is espcially challenging when multiple CSO, allied in a federated enterprise, must respond to many possible situations with computations defined in dynamic cyberspatial contexts, all while competing for shared resources and while trying to achieve maximum utility.
More information related to this Technical Framework is available in hardcopy and PDF versions. The Cyberspatial Mechanics paper was peer-reviewed and published in 2008 in the IEEE Transactions on Systems, Man and Cybernetics – Part B. It was followed in 2009 by a book entitled Creating Rational Organization. Both are available in the Download section of this site.
Note: The automation framework presented here is predicated in part on US Patents 7,181,302 B2 and 7,835,931 B2. Together they define “a method and systems for distributed, real-time command and control of an enterprise.”
(c) 2018 Stratum 4 LLC, All Rights Reserved