[comp.robotics] Mathematical Models of Automobile Dynamics

yamauchi@cs.rochester.edu (Brian Yamauchi) (01/16/91)

As part of my thesis research in behavior-based robotics, I'm going to
be building a simulated robot control system for freeway driving, and
I'd like to include reasonable assumptions about automobile dynamics
in the simulator.

In particular, I'm interested in the following functions for typical
automobiles:

	1) Maximum acceleration as a function of speed
	2) Maximum deceleration as a function of speed
	3) Maximum change in heading as a function of speed
	4) Interactions between turning and max acceleration
	5) Interactions between turning and max deceleration

Of course, the actual values will depend on whether you're driving a
Ferrari or a Yugo, but I'm interested in the general shape of the
curve for your average coupe or sedan.

A subject search in our library shows scores of references on
automobile industry & trade, automobile workers & labor unions,
environmental effects of automobiles, social aspects of automobiles,
and even a book of poetry about automobile factories, but nothing
about automobile performance or dynamics.

				Thanks in advance,
--
_______________________________________________________________________________

Brian Yamauchi				University of Rochester
yamauchi@cs.rochester.edu		Computer Science Department
_______________________________________________________________________________

laforce@krypton.arc.nasa.gov (Soren LaForce) (01/17/91)

In article <YAMAUCHI.91Jan15180754@heron.cs.rochester.edu>, yamauchi@cs.rochester.edu (Brian Yamauchi) writes...
> 
>As part of my thesis research in behavior-based robotics, I'm going to
>be building a simulated robot control system for freeway driving, and
>I'd like to include reasonable assumptions about automobile dynamics
>in the simulator.
> 



            stuff deleted



> 
>				Thanks in advance,
>--
>_______________________________________________________________________________
> 
>Brian Yamauchi				University of Rochester
>yamauchi@cs.rochester.edu		Computer Science Department
>_______________________________________________________________________________



Ok, you asked.

The department of transportation (DOT) funds research in a variety of
vehicle properties, i.e. dynamics and the simulation of dynamics.

I have in front of me the final report of DOT contract
number: DOT HS-7-01715.  The report number is: DOT-HS-805-370.  The 
report is titled: Improvement of Mathematical Models for Simulation
of Vehicle Handling (Vol 7: Technical Manual for the General Simulation).

Does this sound like what you need?

The document is available to the U.S. public through the National
Technical Information Service, Springfield, Virginia, 22161

The report is 278 pages.  It is a **RIGOROUS** mathematical model.

I suspect that it is more detailed than what you need.

If you are interested, I may be able to provide a simplistic vehicle
model, i.e. no tire model, and about 1 page of alegbra.  I will have to
make sure it is ok to release information before sending it.

If you are interested, please e-mail me directly at:

      laforce@krypton.arc.nasa.gov

Best,  Soren LaForce

hollombe@ttidca.TTI.COM (The Polymath) (01/17/91)

In article <YAMAUCHI.91Jan15180754@heron.cs.rochester.edu> yamauchi@cs.rochester.edu (Brian Yamauchi) writes:
}As part of my thesis research in behavior-based robotics, I'm going to
}be building a simulated robot control system for freeway driving, and
}I'd like to include reasonable assumptions about automobile dynamics
}in the simulator.

There are several insurance companies, and consultants to them, who do
computerized automobile accident simulations to present as evidence in
court cases.  The software is probably proprietary, but they might be
willing to tell you their sources for the information you want.

Sorry, I don't have any specific names.

-- 
The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@ttidca.tti.com)
Head Robot Wrangler at Citicorp(+)TTI             Illegitimis non
3100 Ocean Park Blvd.   (213) 450-9111, x2483       Carborundum
Santa Monica, CA  90405 {csun | philabs | psivax}!ttidca!hollombe

jennings@svax.cs.cornell.edu (Jim Jennings) (01/18/91)

In article <YAMAUCHI.91Jan15180754@heron.cs.rochester.edu> yamauchi@cs.rochester.edu (Brian Yamauchi) writes:
>
>As part of my thesis research in behavior-based robotics, I'm going to
>be building a simulated robot control system for freeway driving, and
>I'd like to include reasonable assumptions about automobile dynamics
>in the simulator.
[info wish list deleted]

If you really only want to make some guesses about maximum
accelerations, velocities, and such, buy a few issues of
Road and Track magazine.   When they review a car, they
give you these sorts of specs, including torque and hp
curves, from which you can derive some of what you need.
Most useful to you might be velocity curves of the 0-x mph
runs.  

*However*, I think you're totally at sea if you expect to
learn anything from this kind of simulation.  How will you
model changing grip due to tread deformation, road surface,
tire pressure, sidewall windup, suspension geometry, tire
temperature, tire compound, road material, slip rate and
angle, etc.?  And this is only the tip of the iceberg when
it comes to analyzing automobile handling.

You could argue that some of these effects on the
trajectory of your modeled vehicle are negligible, but if you
want to know about the extremes of adhesion in turns,
acceleration, and braking, then I won't believe your
arguments.   

If you want to use a behavior-based system (or any other,
really) to control a vehicle with nontrivial dynamics, my
suggestion is to build a (real) vehicle that can travel at
sufficient (scale) speed.

--Jim Jennings
--Cornell Computer Science Robotics Lab

yamauchi@cs.rochester.edu (Brian Yamauchi) (01/18/91)

In article <50852@cornell.UUCP> jennings@svax.cs.cornell.edu (Jim Jennings) writes:
   In article <YAMAUCHI.91Jan15180754@heron.cs.rochester.edu> yamauchi@cs.rochester.edu (Brian Yamauchi) writes:
   >
   >As part of my thesis research in behavior-based robotics, I'm going to
   >be building a simulated robot control system for freeway driving, and
   >I'd like to include reasonable assumptions about automobile dynamics
   >in the simulator.

   How will you
   model changing grip due to tread deformation, road surface,
   tire pressure, sidewall windup, suspension geometry, tire
   temperature, tire compound, road material, slip rate and
   angle, etc.?  And this is only the tip of the iceberg when
   it comes to analyzing automobile handling.

The goal of this research is not to study automobile handling, but to
demonstrate a behavior arbitration algorithm which can deal with large
numbers of possibly conflicting sensorimotor constraints.  I don't
claim this will generate any new insights into the analysis of
automobile performance, although I do hope it will provide insights
into how to build reactive robots which need to resolve conflicts
between behavior agents.

   If you want to use a behavior-based system (or any other,
   really) to control a vehicle with nontrivial dynamics, my
   suggestion is to build a (real) vehicle that can travel at
   sufficient (scale) speed.

This would have been my first choice by far, but, alas, the resources
are unavailable.  However, this vehicle simulation is only part of my
thesis research.  I also plan to demonstrate by behavior arbitration
algorithm by building a (real) visually-guided dextrous manipulation
system (an area in which we do have the robotics/vision hardware).

I have a strong personal preference for working with real robots, and
I agree that building real systems is the best way to demonstrate
robotic algorithms, but sometimes you just don't have an ALV handy...
--
_______________________________________________________________________________

Brian Yamauchi				University of Rochester
yamauchi@cs.rochester.edu		Computer Science Department
_______________________________________________________________________________

nagle@well.sf.ca.us (John Nagle) (01/19/91)

     As mentioned above, some very detailed models of automobile 
handling are available.  Existing models are good enough to predict
vehicle behavior even under extreme conditions.  The first really
good model was developed at the Cornell Aeronautical Laboratory
(now Calspan) back in the 1970s.  As a demonstration of its
validity, it was used to design a set of ramps for a car stunt
in which a car executed a 360 degree roll in the air and landed 
smoothly on the landing ramp.  (A version of this stunt can be
seen in one of the James Bond movies; the ramps look like the ruins
of a bridge, but were very carefully calculated.  A movie credit
appears for the computation.)

     The California Department of Transportation (Caltrans) has been
doing some simulations of automated freeway driving, although their
model is 2D and has rather simplistic dynamics.  Contact the
Caltrans PATH (Programs to Automate the Highway) office in Sacramento.

					John Nagle