1. University of Pennsylvania 1 SDRL CHARON SDRL and GRASP University of Pennsylvania Funded by DARPA ITO

2. University of Pennsylvania 2 SDRL What is Charon? Charon is a high-level modeling language and a design environment for embedded systems reflecting the current state of the art both in formal and object oriented methods (UML).

3. University of Pennsylvania 3 SDRL The Hybrid Systems Group at UPenn Professors:  Rajeev Alur (CIS)  Vijay Kumar (ME)  Insup Lee (CIS)  Max Mintz (CIS)  George Pappas (EE)  Harvey Rubin (Med) Postdocs:  Eric Aaron  Thao Dang  Rafael Fierro  Hyoung Hong  Supratik Mukhopadhyay  Oleg Sokolsky Ph.D. students:  Calin Belta  Joel Esposito  Yerang Hur  Franjo Ivančić  Pradyumna Mishra  Usa Sammapun Others:  Radu Grosu  Salvatore La Torre  Valentina Sokolskaya  Paulo Tabuada

4. University of Pennsylvania 4 SDRL Overview  Charon language  Architectural Hierarchy (Agents)  Behavioral Hierarchy (Modes)  Charon toolkit  Syntax-directed editor  Parser and type checker  Global simulator  Plotter  Simulation in Charon  Conclusions

5. University of Pennsylvania 5 SDRL Language Summary  Individual components described as agents  Composition, instantiation, and hiding  Individual behaviors described as modes  Encapsulation, instantiation, and scoping  Support for concurrency  Shared variables as well as message passing  Support for discrete and continuous behavior  Well-defined formal semantics

6. University of Pennsylvania 6 SDRL Architectural Hierarchy in Charon Each agent can be represented as a parallel composition of sub-agents Agent Robot1 Robot2 sensor actuator processor Input Port Output Port sensor actuator processor

7. University of Pennsylvania 7 SDRL Behavioral Hierarchy in Charon main awayTargetatTarget controlsensing Entry Port Exit Port Modes  Each agent consists of modes or behaviors  Modes can in turn consist of submodes

8. University of Pennsylvania 8 SDRL Modes main awayTargetatTarget controlsensing  Each agent consists of modes or behaviors  A mode describes flow of control inside an agent  Modes contain  Control points (entry points, exit points)  Variables (private, input, output)  Continuous dynamics  Invariants  Transitions  Nested submodes

9. University of Pennsylvania 9 SDRL Continuous Behavior in Charon  Differential Constraints  write Position robot_Pos;  diff diffStop {d(robot_ Pos.x)=0.0; d(robot_ Pos.y)=0.0;}  Algebraic Equations  write real robotEST;  alge contEST { robotEST < foo(x) + bar(x); }  Invariant Constraints in Modes  inv invTUCost { lub <= x <= gub; }

10. University of Pennsylvania 10 SDRL Charon Toolkit  A prototype implementation of Charon development environment is available  Contains  syntax-directed editor  parser and type checker  global simulator  plotter (adapted from Ptolemy@Berkeley)  Implemented in Java  Open internal representation allows interfacing with other tools

11. University of Pennsylvania 11 SDRL Charon Toolkit – cont.

12. University of Pennsylvania 12 SDRL  Work to date  GUI  Syntax-directed editor  Charon semantics  Parser and Type-Checker  Internal representation  Global simulation  Plotter  Current work  Graphical input language  Modular simulation  Distributed simulation  Model Checker  Qualitative reasoning Current Implementation Status Charon Parser Simulator Generator Simulator Generator Control Code Generator Control Code Generator Model Checker Syntax Tree Internal Representation Generator Type Checker Type Checker Internal Representation Charon SpecificationVisual Specification HyPix Interpreter Trace Plotter Simulation Files

13. University of Pennsylvania 13 SDRL Simulation in Charon  Simulation is the main analysis technique for hybrid systems  Global simulation is currently implemented  Advanced simulation techniques exploit the structure in Charon programs:  behavioral hierarchy is used by modular simulation for the computation of continuous behavior  architectural hierarchy is used by distributed simulation to distribute the simulation workload

14. University of Pennsylvania 14 SDRL Simulation in Charon – cont.  In the current approach, a program-specific simulator is generated from the Charon program  Each object of the Charon program is converted into an executable Java object  Together with a program-independent core, these objects implement behavior of the program

15. University of Pennsylvania 15 SDRL Simulator Generation  Note the similarity to code generation! CHARON program modes and agents external objects simulation core generated classes for each mode and agent I/O, mathematical libraries, etc. JVM generate map

16. University of Pennsylvania 16 SDRL ?Conclusion?Final Outlook?Links?  Not sure what to say here, other than:  More information about the Hybrid Systems Group at UPenn available at  http://www.seas.upenn.edu/hybrid  More information about Charon and download of current version available at  http://www.cis.upenn.edu/mobies/charon