Fast Pace of IV R&D Evident at Intelligent Vehicles 2002 (06/02)

Cutting-edge researchers from all over the world converged on Versailles, France, from June 18-20 to report their latest progress in implementing intelligent vehicle technologies at the Intelligent Vehicles 2002 conference. The meeting was sponsored by the Institute of Electrical and Electronics Engineers (IEEE) and the French National Institute for Research in Information and Automation (INRIA). General Chair Dr. Michel Parent of INRIA and Program Chair Dr. Uwe Franke of DaimlerChrysler worked with experts worldwide to put together a fascinating presentation of technical progress.

Held at the circular Palais de Congress near the Chateau of Versailles, the conference attracted over 250 registrants. Strongest representation was of course from Europe -- France, Germany, and Italy in particular. Engineers from the Asia-Pacific region were also numerous, with about 35 from Japan, 3 from China, and 2 from Australia. About 25 persons attended from the USA. From the commercial sector, 20 or so key vehicle industry companies were represented (see below), and dozens of university researchers added to the mix. The size of the group and the extent of vehicle industry indicates that the IV R&D community continues to grow steadily.

A sampling of companies attending IV2002
PSA-Peugeot-Citroen Honda Nissan DaimlerChrysler Denso TRW VALEO Siemens VDO Bosch Volkswagen	HRL Laboratories Sarnoff IBEO Automobile Sensor GmbH Renault Toyota Volvo Technological Development BMW Hitachi Centro Ricerche Fiat

Since IV2002 is primarily a technical conference, representation from governments was limited, but Australian, Dutch, French, and European Commission officials were there to present their programs and consider the impact of IV technologies on public policy and the operation of the road transportation network.

The conference opened on June 18^th with a European perspective provided by Eric Ponthieu of the European Commission Program for Environment and Sustainable Development. Representing a mix of technical presentations and poster exhibits, sessions during the day focused on vulnerable road users (pedestrians, etc.), vehicle navigation, hardware/software, lane recognition, machine vision, human-machine interaction, and driver monitoring.

Of note was the significant recent progress made in detection of pedestrians and other obstacles in complex urban environments -- a capability seen as a key enabling step towards implementing low speed Adaptive Cruise Control on cars. Results presented from multiple researchers showed what appeared to be very robust and reliable pedestrian detection, primarily using machine vision. DaimlerChrysler showed a technique which can detect obstacles in the time of one image frame, between 40 and 80 msec. Other work in this area is being done by the University of Palma in Italy, the University of Maryland in the USA, the University of Ulm in Germany, and Honda R&D. An evening reception was sponsored by M. Frank Borotra, President of the Yvelines District, in which Versailles is located.

A sampling of research institutions attending IV2002

TNO (Netherlands)
- Traffic and Transport
- Automotive
- Physics and Electronics Laboratory
- Conekt
Ecole des Mines des Paris (France)
INRIA (France)
LIVIC (France)

INRETS (France)
CNRS (France)
LCPC (France)
AHSRA (Japan)
TRG Southampton (UK)
Fraunhoffer Institute (Germany)

The keynote address opening the sessions on June 19^th was provided by the renowned Dr. Ernst Dickmanns, a long-time pioneer in the use of machine vision for intelligent vehicle systems. Dr. Dickmanns reviewed the history of this field, and noted that gaze-controlled machine vision holds great promise for future developments. Following sessions focused on communication and control, obstacle detection and collision avoidance, machine vision, and future transportation systems. Current research in the machine vision field is strongly tilting towards exploiting the capabilities of stereo vision, which mimics the human two-eye approach to provide depth information of the scene.

Presentations within the Future Transportation session included a review of the new AWAKE European project to develop a next-generation driver alertness monitoring system; the guided snowplow project in Minnesota which is part of the US Intelligent Vehicle Initiative, and CarTalk 2000, a European project which is developing inter-vehicle communications techniques. The day ended with a reception at the Eiffel Tower.

Thursday, June 20^th opened with a second keynote address, this one provided by Dr. Sadayuki Tsugawa of the Japanese National Institute of Advanced Industrial Science and Technology. Dr. Tsugawa has also been a pioneer in the IV field, beginning in the 1970’s. His talk focused on work they have accomplished in vehicle platooning using inter-vehicle communication and satellite positioning.

Following sessions focused on communication & control and lane recognition, before the group departed en masse in buses to travel to the site for Demo 2002 and the final sessions. Demo 2002 was organized separately and overlapped with IV2002. After hearing about such great technology for the last two days, IV2002 registrants finally got to see some of the systems in action.

Four dramatic live demonstrations, beamed via video into an on-site conference room, showed capabilities such as lane recognition, obstacle detection and avoidance, side collision awareness, and lateral and longitudinal control for urban trucks. A variety of scenarios were played out on the test track to emulate real-world situations. A full report on Demo 2002 is provided in an accompanying IVsource article.

The conference closed with a reception sponsored by M. Etienne Pinte, the mayor of Versailles.

It is interesting to examine the topics and players of IV2002 to gain some insight into industry trends. From a look at the participants in the conference, it appears that there is relatively very little academic or commercial IV R&D work in the US compared to Europe and Japan. This can be attributed to more fragmentary government R&D funding in the US, and the reality that the major automakers selling cars in the US are either based overseas (DaimlerChrysler, Honda, Toyota, etc.) or have major subsidiaries overseas which spearhead R&D (GM’s share in Fiat, Ford’s ownership of Jaguar and Volvo, etc.). Europe and Japan are seen simply as more fertile environments for this type of research. European work will enter a new phase next year, as projects are awarded under the new Sixth Framework R&D program of the European Commission.

Another observation: while the idea of automated driving is tantalizing to engineers and robotics researchers, the overwhelming focus of the technical work presented at the conference is focused on improving vehicle safety through crash avoidance. For example, DaimlerChrysler has stated their ultimate goal as “accident-free driving.”

Also, the bulk of testing and R&D is primarily focused on cars (as opposed to buses and trucks), although much of the technology is applicable to any vehicle type. Notable exceptions are the Dutch government testing of Lane Departure Warning Systems for heavy trucks, and the presentation of the Phileas automated bus system which is scheduled to begin operation in Eindhoven, Netherlands later this year. Another exception is the new focus on specialized automated people-movers for non-highway environments – for example, the CyberCar concept (www.cybercar.org).

Within the safety domain, work focused on the highway situation continues -- basic capability achieved a few years ago for lane detection and obstacle avoidance is becoming more robust. For example, new work from DaimlerChrysler addresses lane detection in poor weather and lighting conditions (such as direct sun). Similarly, Hughes Research Labs presented work done under ACAS, the largest US government sponsored project in this area, to detect the travel lane at long range to assist forward collision warning radar systems in tracking the correct targets. Others are working on handling a “cut-in” from a preceding vehicle and on active steering to avoid collisions.

Probably the hottest research topic is complex urban environments: city streets with pedestrians emerging from behind parked cars, kids darting into streets, and copious 'non-target' clutter. How does an IV system sort through everything happening in the scene to perceive critical events? Researchers from places such as DaimlerChrysler, Honda R&D America, the University of Parma, INRIA, and the French Ecole de Mines, have waded deep into this quagmire and they appear to be making significant progress. Laser scanners are seen as particularly useful for these situations; these high-tech (and costly) units are the new darling of the sensor world, as they offer perception abilities beyond radar and vision.

On the sensor scene, machine vision has clearly taken center stage in R&D activity, as first generation automotive radar systems have become relatively mature in recent years. In particular, it seems that everyone is working on stereo vision – active players include DaimlerChrysler, the University of Rouen, University of Parma, and automation pioneers Carnegie Mellon University. Multisensor fusion is also a key focus – that is, bringing together the data from multiple radars, cameras, and laser scanners to develop an accurate perception of the road and traffic environment.

Browsing the talks, demos, and exhibits, two organizations stood out – the Dutch lab TNO had a strong presence from at least four of their subsidiary labs, reflecting a new business focus on Intelligent Vehicles, according to TNO’s Dr. Bart van Arem. And, at least in terms of published work, DaimlerChrysler Research clearly dominates the R&D scene.

IV2003 will be held in Columbus, Ohio during June 9-11 and is already in the works under the leadership of Dr. Umit Ozguner of Ohio State University.