> For the complete documentation index, see [llms.txt](https://cctang.gitbook.io/softwarecrafting/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cctang.gitbook.io/softwarecrafting/seminars/2017-mathworks-advisory-board/day2.md). # Day 2: Model-Based Design with Simulink: Modeling and Simulation ## Model-Based Design Keynote - Andy Grace * Technology Trends * Mechanization of Everything * High-speed Wireless Everything * Autonomous Systems in the Age of Aritificial Intelligence * "AI can bring on a second industrial revolution" - Kevin Kelly, Wired Magazine * Deep Learning and Compute Power improvements are driving innovation in AI * Leads to: * Simulation as an imperative * High levels of abstraction necessary * Rigorous processes for safety and security * Model-Based Design * Systematic use of models *throughout* the development process * Scope: Electronic systems that interact with the *physical* world * Three major areas * Modeling & Simulation * Software components * High-level abstraction with Simulink, Stateflow, & MATLAB * Physical components * Provide a suite of tools that allow working at multiple levels of fidelity * Physical systems * One integrated simulation platform that support multiple tools, formalisms, and versions * Software systems * System architecture was the most requested breakout from last year's MAB survey * System composition * Interfaces * Variants * Run-time effects * Scheduling * Modeling requirements * Simulink Requirements (R2017b) * Goal: Top-down and bottom-up system design linked together into one workflow * Performance: Parallel Simulations (R2017a) * Code Generation * Ease of use * Reduced bus wiring (R2017a) * Unbeatable code efficiency * Automatic fixed point conversion * Goal: Support extensive hardware targets from student targets to the highest performing rapid prototyping systems * Vision: Efficient code that easily plugs into any software environment * Test & Verification * V\&V will be separated into three products (R2017b) * Simulink Requirements * Simulink Coverage * Simulink Check * Vision: Automate steps and provide assurance for designs * Deep Solutions * Wireless * Providing more communication standards * Autonomous Systems Technologies * Localization * Planning * Controls * Perception ## From Components to Systems ## Design Patterns for Bottom-Up Component Modeling - Mani Ramamurthy * Traditional workflow bottom-up model-based design * A recent trend toward emergence of the software framework * Areas of need * Modeling abstractions for embedded frameworks * Modeling abstractions for communication middlewares * Modularity, reuse, portability * Continuous-integration * Agile development * Mission: Enhance model-based design for component frameworks * *Design Patterns*, Erich Gamma, Richard Helm * Maximizing reuse * Define components * Parameterize design * Parameterize structure * Effect interfaces * Rich interfaces * Interface management * Interface Objects * Use concepts such as ICDs * Partitioning and Scheduling * Behavior design patterns * Design Partitioning * Scheduling Editor * Composing * Simulating timing * Run-time environment modeling ### Maximizing Reuse - Mike Tocci * Units of reuse * Libraries * Foundation for graphical reuse in Simulink * Dynamic Interface * Support for code reuse as an optimization * ideal for Simscape components * Libaries support Test Harnesses (R2016b) * Model References * Interface is locked * Incremental code generation * Only rebuild if changes are being made * Encapsulate design data * Simulink Variants * Structural Parameterization * Variant propagation (R2016b) * How to start * Build a utility * Put utility into a library subsystem * What is a componenent in Simulink * 1000s of blocks * Generate C/C++ code * Supports composition * Design Data * Keep data definitions and usage as close as possible * Use Model Workspace for all local data * Scope global data using Data Dictionaries * Model Data Editor (R2016b) * Model Mask (R2017a) ### Effective Interfaces - Alphan Ulusoy * Signal interfaces * Ports are not aligned between components * Often buses are used as a workaround * New highlight > create bus shortcut (R2017a) * Separation of interface from implementation (R2017a) * Called Bus Element blocks * Looks a lot like from/goto pattern * Accessible from bus creator hover-over options * Function Interfaces * Multi-instance shared resources (R2017b) * Example of watchdog using a datastore and the startup pattern * Function Ports for Composition * Client/Server pattern * Both use the same data dictionary ### Modeling Software Services - Nirmal Gunaseelan * Driver Services (R2017b) * Example: NVM driver software * Uses software interface * Black-box approach: S-functions * White-box driver services is new feature * Middle services in the software framework * Modeled with global Simulink functions * Visualization is available * Reference implementation for AutoSAR (R2017b) ### Partitioning and Scheduling - Alongkrit Chutinan * Software integration problem * Multiple software components * Integrate components into top-level application * Typical workflow * Create components * Reference components in top level * Buildin simulation environment around components * Generate code from individual component model * Collect entry point for functions for each component * OS has pre-defined entry point functions (partitions) * Integrate code into OS * Model styles * Export function model * The scheduling pattern * Rate-based model * Algebraic loops are * Scheduling pattern with rate-based models (R2017b) * Using scheduling pattern to simulate startup/shutdown * Available to export function models (R2016b) * Extended to rate-based models (R2017b) ### Automatic Partitioning for Multicore Execution - Katalin Popovici ### Scheduling Editor - Andrew Calcutt ### Advanced Run-Time Environment Modeling - Senthil Manickavasagam ### Complementary Patterns Using Stateflow (Enabing Effective Designs with Stateflow) - David Andrade * Enabling patterns * Scheduling * Mode switching * Debouncing using a stateflow keyword called duration (R2017a) * State logic * MATLAB * Directly in the state, or in a function * Child stateflow charts * Atomic subcharts are available * Port mapping dialog (R2015a) * Simulink subsystems within Stateflow (R2017b) * Can initialize state within a SL subsystem before it's enabled ## Top-Down System Engineering ### Framework for Modeling System Architecture - Devdatt Lad * Vision * Seamless model-based workflow all the way from high-level system architecture to detailed implemention * Model-based system engineering * System structure * New canvas tool * Implementation agnostic pipes and connectors * Can specify as design progresses * Validate requirements * Parameterize * Evaluate feasible alternatives * Select optimal architecture * Architecture Tool speaks MATLAB * Plug in analysis algorithms into the diagram properties * Profiles * Allows customization of architecture semantics to the tool * Behaviors * Capture operation scenarios as sequence diagrams * Capture function requirements as activity diagrams * Capture component behavior as state charts * Possible to start with existing Simmulink models * Interface Connectivity View * Impact analysis * Deployment View * Software Architecture View * Custom-defined Views * Boot-strapped implementation as you move from architecture to implementation ### Towards Software Architectures - Tunc Simsek * Bring together bottom-up and top-down design * Explict control over software interfaces and implementation (R2017b) * Control how data is shared * Control how functions are scoped * Hide implementaiton details through graphical encapsulation * Simulate asynchronous messages and functions * Scheduability analysis of software architecture * Continuous integration with software architecutre modeling ## Scaling Up Model-Based Design - Rob Aberg * Size and Scope: * Past: 1 - 2 code generation targets in one model * Simulink caches (R2017a) * Simulink parsim (R2017a) * Simulink is being enhanced to be a full integration platform ### Simulation Integration Platform - Haihua Feng * Flexible, reliable, and scalable multi-platform integration * Challenges * Interfaces available * Today: 100+ connection partner tools provide co-sim gateways to Simulink * Today: S-function * FMI Support Update * Functional Mockup Interface * FMI Import into Simulink * Available for download (R2017a) * Ships with Simulink (R2017b) * Numberical issues across co-simulation boundaries ### Large Model Performance: Simulate It and It Simulates! - Fu Zhang * Multi-domain workflow * Modeling > Configuration > Simulation > Analysis > Feedback * New Modeling Capabilities (R2017b) * Improved algebraic constraint handling * Solver Configuration Simplification * Simulink will automatically pick solver (R2016a) * Pick a solver at the beginning of the simulation * Large Scale Model Execution * Large scale models are usually assembled from components * Model references with local solvers * Benefits of white box co-simulation * Better communication among components * Easier debugging with the same tools * Analysis: Using Solver Profiler (R2016a) ### Simplified Block Authoring Workflows - Kaustubha Govind * Why custom blocks? * Create a new algorithm that doesn't exist * Custom code/simulation * Co-simulation gateway * Host-specific code * Hardware-specific code * Standardization across teams * Block authoring process * Choose block type * Implement algorithm * S-Function builder, for example * Customize interface * Mask editor * Build rich dialogs * New widgets * Add to library * Goal * Streamline * Mistake-proof * Simplify * S-function Quality and Robustness Checking (R2017b) ### Faster Simulations, Bigger Models #### Working Effectively with Field Data - Alex Trubitsyn * Field Data * Fleet * Track * Improved usability of Dataset Object * Integration with variable editor (R2017a) * Working with big data for simulations * Stream directly to disk * Constant run-time memory use * Accessing simulation data on disk (R2016a) * Simulink.SimulationData.DatasetRef * Read-only reference to object on disk * Analysis in MATLAB * Pass to another simulation * SimulationDatastore (R2017a) * Read from disk, so size in memory is minimal #### Massive Parallel Simulations - Rahul Kumar * `SimulationInput` (R2017a) * `parsim` (R2017a) * Without PCT/MDCS, simulations run sequentially * Automatic management of build folders * Better error diagnostics * Progress display * Monitoring multiple simulations with Simulation Manager (R2017b) ### Better Workflows for Bigger Teams - Gavin Walker * Simulink Project increases your agility to try new workflows * Provides environment * Locations for generated files ("slprj") (R2017b) * Tight integration with source control * Simulink Comparison & Merge tool has a new UI (R2017b) * New workflows built on top of Simulink Project ## Editors and Visualization - Jay Torgerson * Vision * Create the world's greatest graphical modeling experience * Most * epxressive * intuitive * powerful * extensible * fast * Theme: Model-centered design * Dashboar blocks * Property inspector dock in canvas * TODO: Investigate this * Symbol viewer in Stateflow (R2016b) * Model data editor (R2015a) * Simulink Finder * Handling complexity * Perspectives * Interface * Test harnesses * Unit test subsystems * TODO: Investigate this * Requirements (R2017b) ### Visualizing Simulation Results - George Quievryn * Simulink Instrumentation * Data * Visualize, Analyze, Manage, Debug, Test, and Understand * Unified logging, spectograms, new comparisons (R2017a) * Stateflow streaming, triggers, more spectral views (R2017b) * Verifying model changes * Compare simulation results between runs * Compare results in simulation and hardware * Simulation dashboards * Shipped in R2015a ### Managing Model Dashboards - Rich Conlan * Flight instruments (R2016a) * Enhanced capabilities ### Annotating and Documenting Models - Simon Greenwold * Types of annotation * Presentation * Emphasis on expressiveness * TODO: Investigate * Documentation * Contextual documentation * Commentary ### Searching and Navigating Large Models - Atul Suri * Signal tracing navigation using arrow keys (R2017b) * Search runs in the background (R2017b) ### Managing Model Data - Paul Jackson * Model Data Editor (R2016b) * Logical organization of design/instrumentation/code properties * Spreadsheet within the canvas * Directly configure the codegen parameters of the outport * R2017b * Signal properties tab * Data Stores tab * States tab * Parameters tab * See referenced variables * Filter contents * Configure parameters in the model without variable in the base/model workspace * Variable Navigation (R2015b) * Discoverable (R2017a) * Restore Missing Variables (R2017b) ### Editing at the Speed of Thought - Dennis Wilkinson * Smart editing * Keep you focused on your model * Bring the tools to you * Eliminate the busy work * Edit as fast as you can think * Recent advances * Default names are automatically hidden (R2017b) * Create subsystem (R2017b) * Pick up signal names automatically when creating a subsystem * Format painter (R2017a) * Pop-up menu * Quick Insert Refinements (R2017a) * Single type and insert will still work * Double click and insert * Automatically differentiates between annotations and block search insert * Automatic port creating by dragging signal to a block * Automatic port naming with signal name * Suggested blocks in Stateflow when transitions are drawn * Stateflow automatically inserts keywords in actions, if forgotten (eg, entry) * Stateflow > Chart > Arrange automatically * Automatically arranges the Stateflow canvas to be more readable * Line routing * Channel router (R2017a) * Channel routing with smart actions integration (R2017b) ## Breakout Sessions ### Speeding Up Simulation - Rajesh Pavan Sunkari #### Massive Parallel SImulations - Rahul Kumar * New workflow * Setup > Run > Monitor > Inspect * `parsim` command * Strealined design-centric workflow * Create a simulation input object * Unclear how to rearchitect test harnesses that utilize input files * Rahul says it can handle input files * One unified interface for running parallel simulations * No need to explicitly set up parpool * No set\_param calls needed * Run in background option (R2017b) * SimulationInput Object (R2017a) * Specify a numerical array, timeseries, or dataset object * PreSimFcn * TODO: Ask Bao to create example * External Inputs * `DatasetRef` * `createInputDataset` * `PreSimFcn` * Once per simulation * Performed on the worker * Use PreSimFcn to offload parameter value computations to parallel workers * `PostSimFcn` * Offload post-processing to parallel workers and reduce the data to be returned back from workers * Benefits of using `parsim` * Mistake-proofing your workflow * Avoiding transparency violations * Use 'TransferBaseWorksapceVariables' * Model dependencies * Automatically picks up model dependencies and sends them to the parallel workers * Only for MDCS clusers, workers on local pool always has access to the host files * Management of build artifacts * Managing logged files * `parsim` automatically brings back logged files * Logging to file or to To File blocks * Makes the file names unique by appending the 'run id' to the file name * Error diagnostics * Alerts whenever a simulation failed * Simulation metadata is returned * Execution time * Any errors that get generated * Status of simulation termination * Simulation Manager (R2017b) * Monitoring multiple simulations #### Scheduling for Multicore - Andrew Calcutt * Parallelize one model for one run across multiple cores * Specify the order of partitions #### Solver Technology (Simulation Performance Improvement) - Fu Zhang * Fu's team is focused on the continuous time solvers * Solver Profiler (R2017b) * Simulink will identify list of candidates to address solver execptions * Command-line interface is available ### Polyspace * Polyspace Improvement - Simplicity * Full support of compilers * Unified user interface * Single configuration * Simplified workflow to launch * Results grouped by product * Auto-detection of concurrency configuration primatives * Removed need for manual configuration * Speeds-up project configuration for multi-tasking applications * POSIX, VxWorks, C++-11 * Polyspace Improvement - Completeness * Code Prover * Easy customization and automation for launching Polyspace and accessing its results with MATLAB API (R2017a) * Improved in-source Annotation workflow (R2017b) * Launch Polyscapce on code generated from MATLAB Coder (R2017b - must request option to expose) * Polyspace results API in MATLAB (R2017a) * Results are in MATLAB Table format * Improved traceability back to Simulink model (R2017a) * Adds a new way to navigate from Polyspace to Simulink * Improved in-source annotations workflow (R2017b) * `/* polyspace MISRA-C:13.2 */` * By default, Polyspace will mark it as "Not a defect: Justified" * Old syntax is still supported for compatibility * Improve Floating Points support * Detect Integers in Floating Points (R2017a) * Benefits * Provide more information to Polyspace users * Reduce oranges * Stack size computation (R2017b) * Provide a metric of estimation of stack size for each function * Only available to Code Prover * Bug Finder * MISRA Checker * MISRA-C:2012 Support (R2015b) * Amendment 1 for security (R2017a) * Security Standards * Full support of ISO-17961 * Improved support of CERT-C * Cybersecurity - Industry Activities & Standards * Coding standards & practices * CERT-C * ISO/IEC TS 17961 (R2017b) * CWE - Common Weakness Enumeration * MISRA-C:2012 Amendment 1 * Checkers * Add new checkers and improve existing ones * Additional programming and security checks (R2017b) * Review of race conditions and deadlocks * Identifying conflicting Read/Write accessess (R2016b) * Polyspace Improvement - More Performant * 50% Code Prover improvement from R2016a -> R2016b, no appreciable changes in R2017a and R2017b * Incremental analysis when integrated into Eclipse * Closing results will release the Polyspace license ### Getting Data In and Out of Simulink - Alex Truitsyn; Stacey Gage; David Boissy * Creating and Connecting Model Inputs * Create input signals for Simulink model * Signal Editor (R2017a) * Create data through a tabular form * Command-line API (R2017b) * Usability Improvements (R2018a) * Connect external inputs to Simulink model * Dataset Signal Plot (R2016b) * Initial release of creatInputDataset (R2017a) * Support for Timetables in plot (R2017b) * Additional data type and format support (R2018a) * Root inport mapper (R2014b) * Data input/output > connect input * Right from the inport block parameters > connect input * TODO: Investigate * Hetergenous mapping modes in Root Inport Mapper (R2016b) * Strong data typing with spreadsheets in Root Inport Mapper (R2017a) * Command line mapping of inputs (R2013b) * `getRootInportMap` * Timetable will be replacing timeseries * Logging in timetable format (R2017a) * Dataset element format * Load timetable format (R2017b) * Simulation Output as Big Data (R2016a) * Stream to .mat file on disk * Won't run out of memory * SimulationDatastore (R2017a) ## Followups * Simulation artifact sharing - Rob Aberg * Build once, everyone else on the same architecture can utilize them * Location of Simulink Project configuration folder - Gavin Walker * Requirements Perspective on canvas - Jay Torgerson * External documentation link or storage so it doesn't get appended to model file - Simon Greenwold * TODO: Investigate logic analyzer block