🚀 go-pugleaf

See: http://amp.ece.cmu.edu/ECESeminar/
Prof. Tsuhan Chen, Host, 2000 ECE Seminar Series


                                 ECE SEMINAR SERIES
                         --------------------------------
                           THURSDAY, FEBRUARY 24, 2000
                                     4:00 PM
                             Refreshments: 3:30 PM
                        Scaife Hall Auditorium, Room 125
                       __________________________________


     ``Hybrid Models for Automatic Powertrain Systems: Revising a Vision''

                                 Dr.  Ken Butts
                           Powertrain Control Systems
                            Ford Research Laboratory
                        Ford Motor Company, Dearborn, MI


                                    BIO

Dr. Butts leads a research team at the Ford Research Laboratory that has
been creating and deploying a model-based software development
environment for automotive powertrain controllers. This environment is
now used in Ford Motor Company's production processes to improve the
quality and to reduce the development time of ~100 reusable software
components. After full deployment, these components will be composed in
various combinations to service ~50 concurrent vehicle applications. Dr.
Butts has more than eighteen years experience in automotive powertrain
control research and advanced development. He has focused on model-based
powertrain controller development since 1994. Dr. Butts received degrees
in Electrical Engineering from General Motors Institute (B.E.E.-1980),
University of Illinois (M.S.E.E-1982), and University of Michigan (Ph.D.
EECS-Systems-1993).

                                 ABSTRACT

Due to the persistent need to develop increasingly complex systems with
improved quality and reduced development effort, automotive manufacturers
are employing model-based development approaches wherever sensible. This
is particularly true for powertrain control system development, as domain
relevant computer-aided control system design tools have become commercially
available. It is now possible to model and simulate the powertrain system
dynamics in closed-loop with detailed behavioral models of the control
algorithm. These control algorithm models capture nominal, initialization,
diagnostic, and failure-mode-effects management modes of operation to the
extent that simulation-based validation and verification procedures can be
employed. These procedures help to ensure that the algorithm design and
its associated software realization meet the system requirements with
quality.

Simulation-based development (design, validation, and verification) methods
only evaluate the system's behavior under the initial conditions, input
scenarios, and parameter values as defined in the simulation test-suite.
Thus, comprehensive validation and verification is expensive and time
consuming. (Of course, exhaustive system validation and verification is
impossible in a simulation-based development approach.) Importantly, given
that powertrain models are being created to support mainline development,
we now have an opportunity to go beyond simulation by employing systems
analysis methodologies in the design process. The purpose of this talk is
to describe hybrid systems analysis queries that, if answered in an
efficient and intuitive way, would be a boon to the powertrain controller
development community.

We begin by stating two analysis tool objectives that are derived from our
experiences in using these models in a production development environment.
First, wherever possible, we desire that our analysis methods be based on
commercial tools. Second, we desire to analyze the models in the styles
(one for physical plant models and another for control algorithm models)
that they are currently built. We desire to analyze the models in the
accepted styles because the model preparation requires significant engineering
effort and it is unlikely that the organization could support the additional
expertise, training, and effort required to specially prepare alternative
analysis models.

Next we describe the modeling styles that are used in the production
development process and provide an example based on an automatic transmission
control system. The physical plant model is comprised of a two-state engine
model, a quasi-static torque converter model, the transmission dynamics for
first gear, second gear and the one-to-two gearshift, and a simple
longitudinal vehicle dynamics model. The discussion focuses on the
hybrid nature of several components within the system. The associated
control algorithm model is comprised of an abstract transmission shift
scheduler and simplified shift control logic for the one-to-two gearshift
and the one-to-two-to-one ``change-of-mind'' gearshift.

We list analysis queries that would enhance the powertrain controller
development process if they were available. We also discuss the specific
application of these queries to the transmission control example. These
analyses include stability within a mode of operation, modal transition
integrity, safety, and liveness.

We conclude with the remark that we have a wealth of new information
available in the automotive powertrain controller development process:
formal and detailed models of the system's behavior. We hope to be able
to fully exploit these models through analysis.