The Cognitive Digital Twin: Architectural Framework

The new design and architecture mindset

CDTs are massive smart modular yet highly integrated and collaborative systems, where biologically-inspired architectures, autonomous cognition, and controlled evolution will be fundamental characteristics and capabilities.

"Inspired by the 11 billion years long brain and the neocortex evolution, Distributed Cognitive Architecture (DCA) is the future of smart systems and machines"

The design, implementation, and operation of cognitive digital twins will not be an easy task at least over the next few years. Nevertheless, the current systems and system-of-systems (SoSs) engineering theories and frameworks will fail short and can’t efficiently help us in building such systems.

"Model-based Systems Engineering (MBSE) techniques and tools should gain cognitive and evaluation capabilities that they can support the semi or, in some cases, full autonomous evolution of the digital and physical twins design and functionalities."

The advances in neuromorphic computing, which replaces the older emphasis on separate modules for input, output, instruction-processing and memory, showed us that machine intelligence should be addressed by both the physical as well as the digital design and architecture.

"A combination of traditional and neuromorphic computing would be soon the dominating computing architecture until a fully biologically-inspired computing architecture, hardware and software is mature enough"

Keeping this concept in mind while designing the physical and digital systems and machines will solve many problems we currently have in AI / ML. It'll also accelerate its patchy progress we’ve since decades. You can’t write couple of algorithms that you think are capable of doing smart things, train them with large set of data then upload them to a machine and expect that this machine will suddenly be smart. Therefore, a major shift in how we envision, architect and build systems and machines is required.

“We should abandon the Von Neumann’s model and principles of designing and building binary machines that can just compute. We should also free up ourselves  from the mentality of Isaac Asimov of building robots that can do repetitive tasks for us (salves). Instead, we should start building biologically-inspired smart machines that will be our future counterparts, colleagues and possibly our adversaries”

However, over the next few years, some will be trying to stitch existing technologies together, extend protocols and standards that they can quickly implement workarounds and get the first generation of cognitive digital Twins designed and implemented. That is fine as a temporary approach as long as it doesn’t distract us from shifting toward the new approach of building such cognitive digital physical twins.

Cognitive Digital Twins : Key architectural building blocks

For the first wave of Cognitive Digital Twins, the following key architectural building blocks will be required:

- Cognitive Digital Twin Core (CDTC) – The Core

The digital core would be the initial digital representation of the “Physical Twin” including its main initial states “as-designed”, “as-built”, “as-sold” and “as-configured”. Also, it’ll contain the initial states and core capabilities of the digital twin itself including the required systems and tools.

"The CDTC will have different metadata, core self-defense mechanisms as well as self governing rules to guarantee that the whole CDPT will function, evolve and used as intended"

The initial core will usually be kept unchanged over time. However, to enable continuous evolution of both physical and digital twins, minor additions or updates will be required and allowed based on specific sets of predefined rules.

- Cognitive Digital Twin Anchors (CDTA) – The Data Workers

Real-time, historical and extrapolated data will drive the actions and the evolution of the CDPTs. Anchors will continually monitor and collect specific parameters, statuses and contents of interest from different local as well as external distributed data sources. They'll use data mining techniques to initially prepare the collected data. CDTA will use different cognitive capabilities (AI/ML …) and techniques to dynamically define the data structures and detect suitable sources.

"CDTA will use different techniques to extrapolate and generate own version of the reality based on varieties of parameters and rules such as; time, experience, context, situation, and self and/or environmental awareness - machine perception"

CDTA will be able to adapt over time that they can effectively deal with changes, which might happen to data structure, contents as well as the sources. They should use local and cross-systems capabilities to evolve its data models in a non-destructive way over time that the right data will be collected, structured, enriched, possibly converted to useful information and made available for usage.

CDTA are required for neuromorphic computing. Our current computing systems collect and store data based on predefined formats and rules. This data will be made available to different processes and tools on request. On the other hand, CDTA will collect data either based on predefined rules, dynamically defined rules or triggered by specific events. Usually, data will be sent directly to the destination(s) in a format that will enable faster reasoning and if needed immediate actions. This key architecture principal is a dominating architecture in the biological species. It is essential for cognitive smart systems and machines and will be made possible by CDTA - The Observing Machines.

- Cognitive Digital Twin Surrogates (CDTS) – The Knowledge Workers

Smart “Things” will have functions to be performed, characteristics to be maintained and interactions with digital and physical worlds to be made.

"Based on situational & context-aware cognitive rules, CDT surrogates are experts in a specific domain and will continuously create and update their knowledge to keep up with the evolution of the overall CDPT and its environment"

To do so, they’ll apply different knowledge discover and machine learning techniques on the data and information available to the CDPT from different sources. This up-to-date knowledge will enable the CDT to take more complex decisions and evolve the cognitive capabilities of the overall CDPT over time.

Additionally, CDTS will communicate their knowledge with different systems and subsystems of the CDT, its physical twin as well as other CDTs (e.g. within its digital Swarm). Complex sets of cognitive rules and skills will determine what, when and how to share knowledge and with whom. CDTS will continuously work with other CDT’s subsystems to update the rules used to gather, create and share knowledge across the lifetime of the CDPT.

- Cognitive Digital Twin Bots (CDTB) – The Makers

CDTBs are creative highly specialized digital workers that can carry out complex and hybrid tasks either in the digital space (e.g. software algorithms) or in the physical space (e.g. to control a Humanoid on another planet or a machine on a smart factory floor).

"CDTBs are smart digital or digital-physical sub-systems designed and continuously learning to do specific tasks repetitively and intelligently. They will leverage the knowledge they gained to get smarter and share it back with their Cognitive Digital Twin"

Those CDTBs will evolve over time through learning by doing as well as the knowledge made available by the surrogates (CDTS). CDPTs will be able to autonomously define, create and train new CDTBs as needed.

- Cognitive Digital Twin Perspectives (CDTP) – The Interfaces

CDT perspectives will be responsible for multi-directional multi-channel communications and interfaces including communications with the physical Twin, other Digital Twins as well as Humans or non-human subjects and objects. A fundamental redesign of some of the current software engineering concepts such as APIs is needed.

"The current APIs have to be redesigned and extended to be able to support CognitiveDigital Physical Twin ProgrammingInterface (CDPTPI or simply TPI)"

TPIs will create one of the major new economies of the 21st century. They'll be the main interface to smart digital and physical systems and machines. To do so, we’ll also have to redefine M2M and M2H interfaces. This is mainly to compensate for the weakest point in such new smart digital world, the Humans, who can lack attention or mental clarity or just bored or biased for example.

"CDPTs will learn to adapt and evolve their communications capabilities, techniques, and styles with humans, other machines, as well as processes"

- Cognitive Digital Twin Self-Management (CDTSM) – The Administrator

Cognitive digital physical twins will usually be large complex and distributed systems. Some digital components will be embedded in the physical twins and some physical components will be controlled by the digital twin. Managing the life cycle of such systems will be too complex and near impossible to do for humans. Therefore, a smart self-management system should be in place to manage the digital physical twins.

Managing the lifecycle phases of CDPTs will require a combination of advanced sets of software systems and cyber-physical management capabilities. In addition, they’ll at least partially manage the physical Twin as well. CDTSM will apply its cognitive capabilities to develop sophisticated project and program management skills similar to what humans possess today. Self-monitoring, proactive and event-driven self-diagnostic as well as robust digital and if possible physical self-healing should be key capabilities of these systems.

"CDTSM will manage the functions and evolution of both digital and physical twins to guarantee compliant actions and purposeful evolution"

CDTSM systems will evolve and be able to define, implement and enforce new self-management rules and tasks. The metadata and core functions embedded in the core of the CDT would define some fixed boundaries and rules for the self-management as well as self-evolution.

- Cognitive Digital Twin Defense System (CDTDS) – The Guardians

Every Digital twin will introduce unprecedented vulnerability and risks to both Digital as well as physical worlds. Because of such possible devastating consequences, the current cyber security techniques such as intrusions detection and response are not enough to secure Cognitive Digital and Physical Twins.

Therefore, CDPTs should be designed so that danger and even serious intents of intrusions will be proactively detected and blocked through a bottom-up digitally and physically secure design. It should be able to use intelligent digital and physical techniques and tools to define and implement self-defense plans aiming to protect itself against any kind of harm, misuse or unauthorized use.

Context and environmental awareness will play crucial roles as well. Therefore, CDTDS will be able to override any external control including human control if needed, especially in massive digital or physical attacks situation with high-risk consequences where humans could be slow to comprehend and react appropriately on time.

"CDPTs will use advanced cognitive techniques to develop and evolve their self-defense plans and capabilities including reasonable preemptive actions against other CDPT(s) if needed"

The Cognitive Digital Swarm (CDS)

Definition

CDTs would socialize, collaborate and partner with another CDTs, which for example share the same interest or provide specific services they need. Such relationships and interests will make it beneficial for the CDTs to create Digital Swarms using specific rules and requirements defined by each individual CDT. Such rules will help the CDTs to identify, find, join and as needed leave digital swarm(s) on-time safely and efficiently. This should be part of the CDT’s core capabilities and be refined over time through continuous learning and balanced evolution.

“Cognitive Digital Twins will be like humans, they’re smarter, safer and more efficient if they communicate, share and collaborate with others”

Cognitive Digital Swarms will be able to create and share their collective intelligence and capabilities based on specific rules defined by every individual member as well as by the swarm collectively. “Swarm Intelligence” will make every swarm member capable of taking smarter decisions and carrying out more sophisticated tasks efficiently and in a more secure way.

"The cognitive digital swarms will use their collective cognitive and situational awareness capabilities to define own rules and identify new swarm members across the whole accessible IoE digital ecosystem and possibly invite them to join the swarm as needed"

Once created, the digital swarms will have their own sets of governance and collaboration rules, which will manage their joint activities together. The accessibility to the physical twins of digital swarm members will be in many cases exposed to more strict rules than the rules used to access the joint digital resources.

CDTs will be allowed to have simultaneous relationships with different digital swarms. However, rules should exist to guarantee that conflict of interests will be avoided and risks to the members will be minimized or preferably eliminated.

In some situations, digital swarms will need to join capabilities and abilities with other swarms to achieve larger common goals. Therefore, swarms will create “Digital Clusters” under specific circumstances. Imagine some digital swarms within a smart city joining capabilities and resources to enable the expected higher quality of life for its residents and visitors.

Key categories of cognitive digital swarms

For the first waves of digital swarms, I’m proposing 4 major categories, which will be helpful in increasing the values, stability, resilience as well as the security of Cognitive Digital Physical Twins and their Swarms as follows:

- Strategic Digital Swarms

The participating CDTs will dynamically share the relevant digital and physical resources, take over tasks on behalf of the swarm or each other. They’ll make specific experiences available to each other that they can jointly achieve their individual and collective goals.

"Strategic digital swarms will not have detailed predefined specific goals. However, they’ll have common causes such as keeping the traffic flowing in a city or protecting the airports of a country or even run a company"

They’ll potentially define specific goals over time and driven by the current situation they're dealing with. The collective cognitive and physical capabilities, as well as the requirements of the swarm, will play a major role in defining and implementing the short and long term common goals.

For example, all the safety-related Cognitive Digital Twins of some or even all the US Airports will be by default and continuously a member of the national aviation safety strategic digital swarm. While every member will do its individual tasks of monitoring and securing its own airport, they’ll share security and safety relevant information instantaneously and in a smart way so that the unexpected security situations will be dealt with appropriately and proactively.

The DS members will also be able to carry out tasks on behalf of the whole DS or specific member such as accessing the physical twin, taking over control over specific equipment, whole building or soon commanding Robotic Security Guards (RSG), which are in fact CDTs themselves, to deal with a security situation in an airport or a school.

- Joint Venture Digital Swarms

Under some conditions, two or more cognitive digital twins will collaborate to achieve a common predefined goal(s). They will work together to put a plan together as they go. For example, several smart factories are assembling the components and systems required to build a specific aircraft model based on the conditions specified in suppliers’ contracts. The different interdependencies between those factories such as What? When? Where and How? will dictate a temporary joint venture relationship between the CDTs of those involved factories across different vendors and geographies. They will work together to fulfill the contract conditions and deliver on time, budget, quantity, and quality. A swarm of the involved supply chain management systems will assist this manufacturing swarm.

"JV swarms would extremely streamline the manufacturing processes and lead to unprecedented levels of manufacturing automation, quality of products and cost reduction"

Once goals have been achieved, JV swarms will not be needed anymore. However, the collective experience should be shared and possibly used to jump start and optimize another joint venture to produce same product or even other products by the same JV swarm or the CDTs themselves within another swarm.

"CDTs will evolve through doing that they can provide higher efficiency, learn to do smarter and more complex tasks in smarter ways over time - continues unsupervised machine learning"

- Outsourcing Digital Swarms

Every Cognitive Digital Twin and its physical Twin will have its strengths, weaknesses, and limitations by design. However, to guarantee focus and optimal use of the digital and physical resources, some limitations will continue to persist or new ones might occur temporary or even permanently due to different circumstances.

"CDTs will be able to learn by doing and even get rid of some of the design weaknesses and limitations"

To make Digital Twins as efficient as possible; outsourcing or crowdsourcing of some tasks will be essential capabilities for the digital twins as well as digital swarms. The relationship between the members of this type of digital swarms will be weak and temporary. Sharing information as well as physical resources will be limited to the bare- minimum required to do the common tasks.

- Spatial Digital Swarms

The spatial locations of some or all elements of Cognitive Digital Physical Twins will make it beneficial in specific cases to establish a spontaneous or planned digital swarm based on current location or spatial requirements. In some cases, the spatial locations of the physical twins will not have to be near to each other for them to be part of a spatial swarm.

For instance, the CDT of my car will be continuously in a strategic alliance with my Human Cognitive Digital Twin and my smart Home Cognitive Digital Twin. Those Twins will construct my Personal Cognitive Digital Swarm. My PCDS will be the foundation for my Personal Cognitive Digital Assistant - CDA. My CDA will dynamically join spatial swarms with the relevant Public Cognitive Digital Twins and swarms to deliver the best and safest experience to me on the road, at home, at work or anywhere else all the time.

- Anonymous Digital Swarms

Like humans, some CDTs will anonymously be a part of a digital swarm for a limited time until they achieve specific short-term goal(s). In this case, not much of information nor physical resources will be shared or at least it should be optional.

For example, the Cognitive Digital Twins of all cars at a specific traffic light will temporary and anonymously join a spatial DS including the CDT of the traffic lights controller itself as well as other related smart city public CDTs. They will work together to optimally and instantly control the traffic flow around this specific spatial location. This will be done within the larger context of optimal traffic management of the nearby areas as well as the whole smart city.

"Anonymous Digital Swarms will play a major role in the era of Cognitive Digital Physical Twins (CDPTs) especially because of the much-needed extra safety and security layers they provide by design."

The short-term goal(s) will be achieved without future commitments or sharing knowledge and experience. Also, access to the Physical Twins should be limited or completely avoided. However, for instance, in the above example, in some situations it would make sense that the swarm controller will be granted access to the breaking systems of one or more cars to avoid collisions or guarantee compliance with current traffic rules.

(Continue reading this article here)

This article was originally posted on LinkedIn.