My current research is part of the California PATH program, and is being conducted under Professors Benson Tongue and Andy Packard. This program seeks to increase the throughput of the existing highway system. Many highways, especially in urban areas, have already grown to their limit. In areas such as the I-80 stretch from Richmond to Emeryville, a major traffic corridor has reach its growth potential but is still often clogged. With the San Francisco Bay on one side and overcrowded cities on the other, the only way to increase the size of the road is to build an upper deck. In earthquake prone areas such as the Bay, this is not an attractive alternative.
To deal with such problems, the options seem to be to encourage more public transportation usage, or to somehow increase the throughput of the existing roads. As a whole, Americans are very fond of their automobiles, and only rely on public transportation when driving themselves is difficult. To allow the public to keep their personal vehicles and the freedom that goes along with them the PATH program seeks to develop automotive control systems that would allow more cars to be puton the road, without increasing travel time.
To sufficiently increase the throughput of the existing highway system, the PATH system must not only pack more cars onto already crowded highways, but move them faster and safer than they do today. The current goal of the program is to be able to space cars approximately 1 meter (about 3 feet) apart while travelling at around 70 MPH. To do this a computer will take over complete control of the car while travelling in special PATH lanes. The computer will be responsible for controlling the gas, brake, and steering of the car. Each car will be specially equiped with sensors to detect the cars around it, as well as a radio link with other cars and the PATH command stations located along the side of the road. The general strategy as of now is to move the cars in strings of 5 to 10 vehicles called "platoons". It is within the platoons that the cars will be spaced closely.
So what is my part in all of this? At the UCB Dynamics and Vibrations labratory we are developing a platoon controller rating system. The purpose of our controller rating software is to allow developers of different control systems to simulate the performance of their controllers, then run it through a program to tell them exactly how well it performed in a number of different areas. Before the rating program can be run, series of simulated runs of a platoon of cars must be run. My research has centered on the development of the computer simulation of a platoon.
While I have done much of the general planning of the simulation model, my specific reseach has been to develop a simplified model of the dynamics of a collision between two cars. This model is responsible for calculating the forces generated on each of the cars involved in the collision, as well a keeping track of just how much each car crushes. The model has been fitted actual crash test data from the National Highway Traffic Safety Administration so that the forces and deflections are realistic.
Currently, I am working on modifications to a numerical integration
algorithm so that my collision model can be included in the overall
simulation package. The characterist times associate with the
dynamics of a collision are several orders of magnitude smaller than
those of a platoon operating in a normal manner. In order to efficiently
include the collision model in the overall package, a modification to
the adaptation algorithm of the Runge-Kutta numerical ODE solver is
required. My modifications center on the fact that if certain states
of a system are known to be numerically stiff, that this information
should be included in the integration step size algorithm so that
unecissary calculations are not performed.