Progress on the enclosure semantics integration front

Last week Kevin and Jan continued the ongoing integration work. An initial version of support for plotting of enclosures was implemented. Kevin also implemented support for starting and halting the enclosure solver from within the Acumen IDE, as well as displaying the enclosure solver's progress reports in the IDE console. Jan extended the syntax supported by the enclosure semantics to include higher derivatives. Adam also helped in discussions about how to change the architecture of the Acumen environment to provide common infrastructure for the different semantics.  

In the coming week Kevin and Jan will focus the integration work on completing a minimal set of features needed to provide a demonstration version of the Acumen IDE using the enclosure semantics. In particular, they will implement a trace view for enclosure values at time-segment endpoints, plotting of the so-far computed enclosures when halting mid-computation and support for setting of solver parameters. In the longer term the main aim is to relax the restrictions on control structures currently supported by the enclosure semantics. 

Writing about tools for hybrid systems design

During the past week Adam has been working on a summary of the survey Languages and Tools for Hybrid Systems Design and has worked on developing a problem formulation for a report about hybrid systems tools and languages. In the coming week Adam aims to finish writing the summary and to refine the problem formulations.

Continuing the enclosure semantics integration work

Previously, Adam and Jan have been working on the integration of the enclosure-based hybrid solver with Acumen. The integration work has centered around the extraction of the hybrid automaton representation of Acumen models that is required by the solver. They also made some necessary additions to the Acumen syntax, the GUI and command line front-ends. 

In the coming week Adam and Jan will work on the design and implementation of the enclosure simulation data representation that will interact with the plotter through the abstract interface developed by Kevin and Jan. Adam and Jan will also develop initial solutions for the setting of simulation parameters through Simulator object.

Interfacing the old and the new - how to support reporting of enclosure simulation in Acumen

Last week Kevin and Jan worked on the definition of an abstract interface to the simulation data that the Acumen plotter will use. The change will enable plotting of data generated from the reference (pure) and the parallel semantics as well as the enclosure semantics. Kevin has implemented an initial version of the abstract interface.

In the coming week Kevin will integrate the interface with the current plotting code. Kevin and Jan will continue working with the abstract interface and implement support for plotting of enclosure simulation data. Kevin will also implement support for a simple trace view of enclosure values at time endpoints of the simulation time segments.

Acumen examples for teaching control

This week Walid worked on developing a new collection Acumen examples to illustrate some of the basic concepts of control. These new examples, tentatively named HelloControl.acm, are written to make more use of Acumen objects and of 3D visualization. Objects can help users build physically reasonable models by encouraging them to think about boundaries between different entities and about what is or is not readily available to one object from another. Adding 3D visualization seems fairly natural once we are already using objects. One example is intended to facilitate discussing proportional/differential (or dampening) control. The system being controlled consists of three masses connected by two springs (along a line), and the control force is applied to the outer two masses. Behind each system is a green bar representing the total energy (kinetic+potential) in the system.  The goal is to stabilize the system at a length where the system is at rest.  Here is the 3D output from an example with three different parameter values for the PD controller.

 

Another example is aimed at explaining proportional/integral/differential (PID) control. In this instance there is a single mass, but it is a three dimensional entity subject to gravity and a small, periodic disturbance.  A small ball marks the goal position for each system.  Again, multiple different parameter values are compared.

 

 In the class room setting we presented each example along with its Acumen code, and carefully explain and discuss each example system in each video in detail. When writing formally, it is important to make sure that any video or figure (such as the ones above) are explained in detail in the text that mentions them. For the present short communication, though, you are instead invited to conjecture on which parameters are varied between the different systems to achieve the effects seen above.

Poster for CPS Robot Design project

In preparation for the annual principal investigator (PI) meeting, Walid put together a poster for the NSF CPS Robot Design project.  The current plan is to use a similar poster for the CERES Open Day.  In the mean time, he welcomes any comments or suggestions that you have for the next poster.

Simpler event detection

In the past week Jan found a way to simplify the event detection algorithm of the enclosure based hybrid solver. It is now done by checking that the domain invariant of the current mode is violated. Jan also found a bug in the specification of the solver's event handling, where transitions from modes other than the current one were possible. These improvement have resolved some issues the solver had when simulating a simple two tanks hybrid system. Currently, the top-level splitting strategy is not calibrated correctly which in some cases leads to unnecessary splitting. This issue will be addressed by carefully checking the conditions for splitting in the current implementation, in particular the criterion for measuring the improvement obtained by splitting. In the coming week Jan plans to review the specification of the top level splitting strategy and to continue the integration work.

Anita visits PSU and other universities in October

As part of Anita's long term goals to strengthen collaborations with Portland State University (PSU), she is planning a visit to PSU in October. The purpose of the visit is to discuss and plan how to continue the collaboration between Halmstad University (HU) and PSU. Envisioned activities include the use of Acumen to model and simulate micro-smart-grid components and electric vehicle components. Anita will also visit Rice and Texas A&M Universities; and The University of Texas at Austin where she will talk about the Acumen project and other interesting research projects going on at HU.

Integration and integration

In the previous week Adam and Jan further developed tests, fixed bugs in the enclosure-based solver, and began to look at integrating the solver with Acumen. The user can now choose the enclosure option in a semantics menu in the GUI.  A semantics flag has also been added to the command-line interface (CLI) version.

Jan fixed a bug in the implementation of the integral operator and completed an implementation of multiplication of affine enclosures.

In the coming week the focus will be on continuing the integration effort Adam and Jan began earlier and on developing a suitable representation of the simulator output that will be used to replace the table of time-series that the other older semantics as as the starting point for the plotting functionality.  

A Student's Guide to Maxwell's Equations: An Update

Walid's reading of the book on Maxwell's equations slowed down as he worked to deal with a few project deliverables.  He did, however, run into an interesting short article that lamented Maxwell's modesty and it's apparent negative impact on the pace of advancement of the field of electromagnetics.  It's not everyday that one makes discoveries like Newton and Maxwell, but the article is a good reminder of the importance of explaining, both as often as possible and as clearly as possibly, the utility of your research results.  And, as it happens, The Weekly Paragraph is a great place to start doing that!

The article goes on to present a couple of gems.  The first is a set of observations that explain how Maxwell's equations and the theory that they constitute became a model for much of the innovations that define modern physics, and in particular the two-level structure of the theory in terms of a world that is not directly tangible and another that is tangible but is affected directly by the first world.  This structure, it seems, also appears in relativity, quantum mechanics, etc.  The second gem relates directly to our research in Acumen, and seems to be best captured by a direct quote:  "To see the beauty of the Maxwell theory it is necessary to move away from mechanical models and into the abstract world of fields.  To see the beauty of quantum mechanics it is necessary to move away from verbal descriptions and into the abstract world of geometry.  Mathematics is the language that nature speaks.  The language of mathematics makes the world of Maxwell fields and the world of quantum processes equally transparent."

Hear, hear!