Caltrans’ New Technology and
Research Program
John Slonaker
Instructor: Kevin
Almeroth
March 21,
2001
Introduction
I found out about the
Media Arts and Technology (MAT) program
through my job as an electrical engineer with the New Technology and Research Program
in Caltrans (The California Department of
Transportation). One of the
Program’s activities is to fund university research of technologies expected to
support Intelligent Transportation Systems
(ITS). My office, within the
program, funds research based at UCSB and other California Universities and has
facilities on campus as well as a couple miles off campus. One of the projects we were funding was
research being done at the Center For
Research in Electronic Art Technology (CREATE). Some of the research staff of CREATE are
now faculty and staff members of the MATP and were looking for students to
enroll in their fledging program. Since MAT offers a course of study in
“Multimedia Engineering,” which includes courses covering many of the
technologies in which my office is interested, I was able to arrange to enroll
in the program on a part time basis.
The ATON Project
CREATE and MATP staff
are currently researching a Caltrans funded project called ATON, which is short
for Autonomous Transportation agents for On-scene Networked incident
Management. The project is a
collaborative effort among CREATE, the Computer Vision and Robotics Research
Laboratory at the University of California, San Diego, and my office, the
Caltrans Test-bed Center For
Interoperability. The main goal
of this project is to find technical solutions for the design of an automated
traffic-incident
detection, monitoring and recovery system. This design includes clusters
of video and acoustic sensors, mobile robotic agents and interactive multimedia
workstations and interfaces, connected using high-speed communication
links. The researchers at UCSD are
working with Omni-Directional Video Sensors (ODVS), which use a focal
plain array camera that points straight up into a parabolic mirror. They then process the resulting image
with software that re-maps the radial coordinates to equivalent plainer
coordinates in order to “flatten” the view. These cameras work in conjunction with
conventional “rectilinear” cameras that provide higher resolution of selected
views. The system is described in
more detail here, and some
results are shown here. A live, unprocessed view from an ODVS
set up on the UCSD campus can be seen here.
Another
technology being investigated for ATON is digital video segmentation, which is
used to extract images of individual vehicles and their respective shadows from
a video feed of a surveillance camera looking at traffic on a state
highway. This will enable computers
to automatically detect incidents or congestion. The system being developed by the ATON
researchers uses a distributed activity recognition database that decomposes
complex activities into spatio-temporally bounded primitive events, stores them,
and provides querying mechanisms to retrieve them. The query mechanism includes rules for
composition of complex activities.
For example, a “collision of multiple cars” is a complex activity that is
defined as a set of patterns of less complex activities. These patterns are
built from primitive events like “car stops suddenly” or “car enters a region,”
where the “region” has been pre-define as a part of the road, as seen from the
camera’s view, not normally traveled by vehicles. The primitive events are detected by
visual and other signal processing layers and can be limited in number. The
complex activities are exponentially greater in number, but it is not necessary
to define them individually. In order to provide the researchers with more data,
I’m working to set up a video server using “Video Logger” software from a
company called Virage and “Real Producer” from Real Networks. I’ve compiled several video tapes of
various types of traffic patterns and incidents by tapping into live video feeds
in the Caltrans District 7 TMC (Traffic Management Center) in Los Angeles. Operators in the TMC can select from,
and simultaneously monitor, several feeds from cameras on freeways and city
streets in the Los Angeles area.
The technology being developed by the ATON project will enable computers
to constantly monitor the camera feeds and alert the operators of detected
activity. I’ve procured and
installed two video capture cards in the computer on which I installed the
aforementioned software. I will use
this system to encode the analog NTSC video signal from a VCR into compressed,
streamable digital files. Once
encoded, I can define and index clips from the files showing specific traffic
behavior and post them as hyperlinks to be retrieved by the researchers. The communications medium for this will
be the CalREN-2 network, which I’ll mention in more detail a little
later.
I’m also working to
render an animated computer simulation of a robotic incident response system as
envisioned by the ATON researchers.
Using a program called Maya, from Alias Wavefront, I have modeled a
system that travels on the concrete median or “K-rail” of a freeway to an
incident site and deploys a CMS (Changeable Message Sign) to alert motorists of
the incident, ODVS and rectilinear cameras to allow remote operators to closely
monitor the incident site, and wireless-controlled vehicles carrying expandable
cones to cordon off the area of the incident. I still intend to animate the model and
add elements such as background, lighting, textures and surface reflectivity to
make it look realistic. In the mean
time, you can try to get an idea of the mechanics of the model in the
low-resolution Figures 1a through 1h.
Figure
1a
Figure 1b
Figure
1c
Figure 1d
Figure
1e
Figure 1f
Figure
1g
Figure 1h
CalREN-2
CalREN-2 is short for California Research and
Education Network, and links various California Universities and research
institutions via high speed OC-12 and OC-48 fiber optic connections. I am currently working to establish an
OC-3 (155 mbps) VC (Virtual circuit) from our facility on the UCSB campus to the
Computer Vision and Robotic Research Laboratory at UCSD. This involves establishing a physical
single mode fiber link from our facility to the main UCSB switch room, which is
about a mile away. This link will
be between the Fore ASX 200BX ATM switch in our facility and a Cisco Lightstream
ATM switch in the switch room. The
Fore switch has interfaces for both single and multi-mode fiber interfaces, but
the Cisco switch only has multi-mode interfaces. Because or this, I procured a media
converter from Black Box that will sit in the switch room between the two ATM
switch ports. (Please see note 1 in
Figure 2.)
Another thing I did in
preparation for this link was work with UCSB Communications services to update
the configuration of the router that connects the LAN in our facility to UCSB’s
main backbone fiber. The CalREN-2
VC will effectively extend a portion of this LAN to the Computer Vision and
Robotic Research Laboratory at UCSD.
In order to prevent anyone there from hooking this portion onto another
subnet and routing other UCSD or Internet traffic across the CalREN-2 VC (or
even running spoofing attacks by "impersonating" a host not on our subnet), we
set up "access lists" that only permit routing (in-bound and out-bound) of
packets whose addresses start with the class C network address of our LAN, which
is a good security precaution in general.
We also did a few other things to “modernize” the configuration. The router had been running RIP, which
we changed to OSPF (open Shortest Path First). This enabled it to communicate with more
other routers at UCSB (also running OSPF) directly, without having to go through
a RIP/OSPF translation server.
Also, it can now choose an alternate route to the Internet if its regular
gateway router goes down. This
gateway router had previously had a static route set up to forward traffic
destined for our subnets to our router, and it advertised this route on our
router’s behalf. Now, our router
advertises its own subnets, and can route outgoing (outside UCSB) traffic
through any router on the UCSB backbone that has a path to the Internet. This didn’t change the load on its CPU
significantly; it is still under 10% in the steady state. We added NTP (Network Time Protocol)
which enables the router to log events by time of day to aid in
troubleshooting. We added SNMP
(Simple Network Management Protocol) so a configuration management server can
get our router’s configuration file and store it for backup purposes. For
security, we specified that SNMP messages can only be exchanged between itself
and the configuration management server.
I’ve submitted the
administrative paper work and, at this point, I’m waiting for the group that
controls access to CalREN-2 to authorize the OC-3 bandwidth allocation.
Network Modifications for
Office Relocation
During the Winter 2001
quarter, we moved our off-campus facility to another location within the Pacific Technology Center. Besides moving equipment, this involved
relocating and re-terminating two data communications media (in addition to
moving the POTS--“Plain Old Telephone Service” lines). One of these is an ISDN (Integrated
Services Digital Network) PRI (Primary Rate -- 1.472 Mbps) line, which we use to
connect our video conferencing system to our multipoint video conferencing
bridge as well as other video conferencing systems at Caltrans and its research
partners. The physical medium for
this line is a “D-screen” cable containing two shielded twisted pairs which were
terminated to pins 1, 2, 4 & 5 (as opposed to 1, 2, 3 & 6 as in
Ehternet) of an RJ-45 jack on a patch panel. The layer one protocol is a T-1 line,
which is leased from Verizon. The
layer two protocol is the ISDN signaling, which is transmitted on channel 24 of
the T-1 line and allows for switching of calls at MCI’s DMS 250 switch (just
like a POTS line). Our terminal
equipment for this line is an Ascend MAX 6000 IMUX (Inverse Multiplexer) which,
in turn, connects to a PictureTel Concorde 4500ZX video Conferencing system
(please see note 2 in Figure 2).
The PictureTel system, as well as other composite NTSC video sources,
connect to the analog inputs of a Jupiter NT850 system which is essentially a
Windows NT computer with special I/O hardware that enables it to display a
composite image on four separate monitors.
These monitors are a 2 x 2 array of 52” LCD projection “cubes” from
Clarity Visual Systems (please see Note 3 in Figure 2). The system just described allows NTSC
composite video (including images of remote video conferencing sites) as well as
SVGA video from the computer to be presented on a 104” “video wall” (Please See note 4 in Figure
2).
The other data
communications medium is a multi-mode fiber optic pair which is used to connect
our terminal equipment, a Fore Systems ASX 1000 ATM, with a Positron Osirus
fiber multiplexer in the main communications room of the Pacific Technology
Center. The other end of this VC
terminates at a similar Positron Fiber MUX in UCSB’s main switch room. Our terminal equipment on campus is Fore
ATM mentioned in the previous section.
(Please see note 5 in Figure 2.)
The data transmitted on this link is IP over LANE (“LAN Emulation”) over
ATM (Asynchronous Transfer Mode) over SONET (Synchronous Optical Network). The link enables host PCs at our Pacific
Technology Center facility to be logically part of the LAN in our UCSB
facility.
Figure 2
