Systems and Software engineering with/out Simulation: State of the - PowerPoint PPT Presentation

Systems and Software engineering with/out Simulation: State of the Art and Way Forward Sept 25, 2019 Bernard P. Zeigler Prof. Emeritus U. Arizona Chief Scientist, RTSync Corp. IEEE 2019-2021 Distinguished Speaker

Outline • Modeling and Simulation (M&S) represents a core capability needed to address today’s complex, adaptive, systems of systems (SoS) engineering challenges. • The limitations of Model-Based Systems Engineering (MBSE) include limited capability to develop multifaceted models, as well as their analysis with computationally powerful and correct simulation engines. • Software engineering has become a primary implementation for SoS development so it must also be brought into the discussion. • We discuss potential for closer integration between the three streams.

Bernard P. Zeigler

Trusted Technical Documents 5

• • • • • • • 6

Student Scholarships $40,000+ Available Each Year Distinguished Visitors Program • Provides professional and student chapters top technology leaders and innovators from every sector of the computing industry. • New speakers and new topics are added on an ongoing basis. 7

Modeling and Simulation in support of Systems and Software Engineering Systems of Systems (SoS) Engineering MBSE Software Engineering Architectural Design Modeling and Simulation Test and evaluation, Conflict management, V&V

Modeling and Simulation in support of Systems and Software Engineering Smart cities, AI-based technology Cyber-physical systems Systems of Systems (SoS) SysML Engineering Internet of Things MBSE Software SoSADL Engineering UML Architectural Design DEVS Modeling and Simulation Test and evaluation, Conflict management, V&V

History of MBSE and DEVS A. Wayne Wymore A Mathematical Theory of Systems Engineering: The Elements, 1967 B.P. Zeigler Theory of Modeling and Grady Booch, et al Model-Based Systems Simulation,1976 Unified Modeling Engineering, 1993 Discrete Event System Language (UML), 1995 Specification (DEVS) Multi-formalism Modeling Systems Modeling Language (SysML), 2001 

Theory of Modeling and Simulation Levels of System Specification Level Specification What we know at this level Example: A Person in a Name Conversation 0 Observation How to stimulate the The person has inputs and Frame system with inputs; what outputs at the usual cognitive variables to measure and level, such as streams of words how to observe them over a time base; 1 I/O Behavior Time-indexed data For each input that the person collected from a source recognizes, the set of possible system; consists of outputs that the person can input/output pairs produce 2 I/O Function Knowledge of initial state; Assuming knowledge of the given an initial state, every person’s initial state when input stimulus produces a starting the conversation, the unique output. unique output response to each input. 3 State Transition How states are affected by How the person transits from Basic Entities in M&S inputs; given a state and an state to state under input input what is the state after words and generates output the input stimulus is over; words from the current state Basic Entity Definition Example: A Person in a what output event is generated by a state. Conversation 4 Coupled Components and how they A description of a person’s I/O source system real or artificial source of Participants ’ in a conversation Component are coupled together. The behavior in terms of neural components can be components and their data specified at lower levels or interaction by spikes is at this behavior collection of gathered data I/O Behavior as in levels of can even be structure level. database system specification systems themselves – leading to hierarchical experimental specifies the conditions Observation of participants’ structure. frame under which system is stream of words in a observed or experimented conversation with model instructions for generating Coupled model of Finite state data generator and recognizer implemented in neural form simulator computational device for Discrete Event Simulation generating behavior of the Environment model

Equivalences at each Level of System Specification Modeling Simulation

DEVS Model of Spiking Neural Net Atomic Model: Leaky Integrate and Fire (LIF) Neuron <X,S,Y, δ int , δ ext , λ, ta> Coupled Model: Neural Net X={p} δ ext (s < Th,p) = s+1 Internal couplings: S={0,1,2,..,Th} Y = {p} δ int (Th) = 0, δ int (s>0) =s-1, δ int (s=0) =0 I d is the set of influencers of d, I d ⊆ D, d ∉ I d ta(T)=tFire, ta(s>0) = tLeak , ta(s=0)=∞ Z j : i ∈ I Y i → X j λ( s>0)=p, λ( s=0)= φ Z j (…y i …) =F(<w i* y i >) IO Behavior Coupled (NN) X p Atomic (Neurons) φ λ y i S δ ext (s,p) Th w i x i λ y i δ ext (s,p) F w i Y p λ y i δ ext (s,p)

A behavior required by self-organizing NN

A behavior required by self-organizing NN https://www.youtube.com/watch?v=UPCAq HP9GYo

A seemingly simple behavior that spiking LIF neuron model can’t display – we need suitably configured higher level unit to implement it s0 p x y One neuron per state One neuron per state S1 λ y i δ ext (s,p) y p x s2 S3 X x S2 neuron persists S2 neuron persists p in state in state λ y i x δ ext (s,p) S2 neuron activates S2 neuron activates Y S3 neuron S3 neuron S1 x y x p S s3 p y λ y i δ ext (s,p) λ y i s2 δ ext (s,p) activate s1 s0 s0 s2 s3 x and y are coded as x and y are coded as y spike and cannot take s0 spike and cannot take s0 to different non- to different non- equivalent states as equivalent states as required required

DEVS Systems-based Simulation is applicable to wide spectrum complex systems • Theoretical – supporting application – Closure under coupling, universality, uniqueness, relation to other formalisms – Hierarchical Model Construction supports complex system development – Supports the correctness of the algorithms and validation of the executing models, e.g., time management is rigorously defined – Discrete Event System Specification (DEVS) formalism can be easily expanded beyond discrete- event world to continuous characteristics of system (Hybrid-DEVS) • In Application – Models, Simulators and Experimental Frames are distinct entities with their own software representations – Precise and well-defined mathematical representation – Models/Experiments are developed systematically for interoperability – Repositories of models/experiments created and maintained systematically – Components can be easily reused for constructing new models – Discrete-event basis improves performance (e.g. no need for a global clock to control timing) – DEVS software can be deployed on distributed computing environments and interact with heterogeneous M&S system. 17

Application Example – European IOT Software Project

MBSE and DEVS: Development and Test of New Message Set for Tactical Data Link Standard Requirements for new Protocol message set Distributed Interactive Simulation (DIS)/Standard Interface for Multiple Platform Link Evaluation (SIMPLE Simulation based Design using testing in net-centric Translation into UML/SysML environment DEVS and SES standards State charts, equivalents conformance Class diagrams and testing Interaction diagrams MS4 Me Rational DEVS Rhapsody IDE Development Remote testing of Environment implementations over the Web Semi-automated test suite design using inverse modeling methods

Automated Testing of Message Protocol Standards Automated Test Case Generation: Goals and Approach Goal: Goal: Increase the productivity and effectiveness Increase the productivity and effectiveness Approach: of conformance for multi-participant of conformance for multi-participant Automate Testing scenarios scenarios Translate from UML To DEVS Formalized approach for converting standards Formalized approach for converting standards documents into test models to run directly against documents into test models to run directly against Create a system, automating the process to the extent a system, automating the process to the extent Express participant Test Models possible possible Systems as DEVS Using Inverse (dynamic,stochastic) models Construction DEVS simulator executes System models to induce Test Model PASS/FAIL behavior in SUT Under Test (SUT) DEVS Simulator Middleware Socket (example) Interact with SUT Network over Middleware

Message Protocol Standard Specification IO Pair Message Name State Name Specification Name 1 Prepare Prescribes 2 DoTask TransmitTask 3 ReceivedTask Acknowledge 4 PerformTask Document Input Output message message System Messages can be: Messages can be: • • Queries Queries • • Commands Commands state • • Responses Responses • • Information Information