ITS World Congress 2013: Autonomous driving technology beyond reality?
Industry experts speculate on challenges to be overcome
ITS World Congress Tokyo 2013 shared the current development status of autonomous driving technologies through the panel discussions by leading global automakers. Autonomous driving technology addressing safety has been recognized widely as a common goal on a long term basis. MarkLines visited the exhibition held concurrently with the congress and confirmed the challenges for autonomous driving to be realized. The challenges include finding a middle ground between cost and safety functions, higher accuracy of object detection through improved image processing, promotion of studies on technologies for recognizing objects other than vehicles and pedestrians. In addition, some findings of future challenges are expressed from literature review, which are different from those of the ITS congress.
ITS: improving transport systems for the future mobility society (May 2013)
ITS World Congress 2013: Autonomous driving technology beyond reality? (November 2013)
Current development status of autonomous driving technologies
"No collision vehicles" and "Google's self-driving car" are only a few of the popular topics regarding ITS. Coincidently, the ITS World Congress was held in Japan again after nine years. MarkLines participated in the congress and summarized the latest technologies in autonomous driving as well as its challenges. The concept of ITS (Intelligent Transport Systems), especially the idea of autonomous driving of automobiles, dates back to the 1970s when urban traffic issues began drawing public attention. Driverless operation of controlled vehicles started to be considered or realized quite some time ago in such fields as urban commuters and construction machinery. It must be noted, however, autonomous driving technology for vehicles that run on public roads is entirely different because artificial intelligence is essential for such vehicles. The artificial intelligence should be able to quickly determine the vehicles’ course of action amid the uncontrolled, chaotic traffic conditions that keep changing constantly. The autonomous driving has been regarded as a possible measure to reduce the driver’s burdens and accidents, and to improve energy efficiency. Therefore, large-scale and on-going research and development efforts for autonomous driving has been conducted since the 1980s. Additionally, automobile collision safety technology has become more related to the ITS domain as it adopted the idea of assisting the driver and actively avoiding accidents. Various demonstration projects were conducted successfully by the end of the 20th century and created confidence among those involved to say “all basic autonomous driving technologies are in our hand.”
In this century, larger scale demonstrations continued using large-sized vehicle groups as autonomous driving can significantly reduce the accident damage and increased fuel efficiency to large-sized vehicles. This type of vehicle-to-vehicle experiments slowed down for passenger cars and was replaced by partial adoption of onboard technologies on individual cars. These technologies have expanded from adaptive cruise control to deceleration and stop functions. However, sometimes these technologies are causing undesired effects as they are not adopted equally on all cars. For instance, a car equipped with an automatic decelerating device slowed down by mistake because of a defect, and was hit from behind by another car that did not feature the same system. Though the accident was caused by a faulty device, the rear-end collision would have been avoided if the automatic deceleration device was also mounted on the vehicle behind. The accident resulted from the lack of unified availability..
The accelerating increase of "No collision vehicles" in the market and the news about Google's self-driving car are rapidly increasing momentum toward the realization of autonomous driving. Various approaches are taken by different automakers toward the common goal as written below:
Autonomous driving attempts among automakers
More details about Japanese automakers are found in a separate report No.1218. (coming soon)
|Corporate name||Aim||Autonomous driving technologies||Timing of market release|
|Toyota||Driver assistance||Adaptive cruise control, lane-keeping control, automatic stop, etc., on production models. "The driver has the leading role. Systems are only there to assist the driver's skills."||Cir. 2015|
|Nissan||Fully-autonomous driving||Autonomous driving to the destination on expressways and ordinary roads. "Fully-autonomous driving with the help of artificial intelligence."||Market launch by 2020|
|Honda||Driver assistance||Developing autonomous driving at low speeds such as automatic stop, autonomous driving on narrow paths and self-parking (Note 1). Focus on collaboration with infrastructure.||To be determined|
|FHI||Driver assistance||Pre-crash braking of higher performance. Next stage of development will focus on lane-keeping control and adaptive cruise control functions.||2014|
|GM||Driver assistance||Vehicles with autonomous driving capability on expressways will reach markets by 2017 as announced in 2012. Semi-autonomous driving technology, Super Cruise, was announced in 2013. The technology has lane-keeping control and auto cruising capabilities.||During 2010s (SuperCruise to be mounted on Cadillac models)|
|Ford||Driver assistance||Traffic Jam Assist technology announced in 2012 is capable of lane-keeping control on expressways and adaptive distancing and stopping functions in congested traffic. The technology is slated for commercial application in 2017. Ford demonstrated self-parking vehicles and announced autonomous steering technology called Obstacle Avoidance in 2013.||2017|
|BMW||Driver assistance||Traffic Jam Assistance on production cars has adaptive distancing and lane-keeping control function in congested traffic. Self-parking (Note 1) is slated for the next commercial application.||2013 (TrafficJam Assistance)|
|Volvo Trucks||Platooning||Autonomous group driving consisting of cars and trucks was demonstrated in a joint project (SARTRE).||2012 (Demonstration test only)|
|Volvo Cars||Driver assistance||Adaptive distancing in congested traffic based on ACC and lane-keeping control technologies was introduced in 2012. It is slated for commercial application.||2014|
|Mercedes Benz||Driver assistance||Adaptive cruise control and lane-keeping control functions are adopted in Stop&Go Pilot＋Distronic Plus with Steering Assist on production cars. The S-class INTELLIGENT DRIVE unveiled at 2013 Frankfurt Motor Show is slated for future application (autonomous driving on streets and urban roads has been successful on production cars by enhanced sensor fusion and 3D digital mapping).||2020 (Launch of autonomous-driving vehicles)|
|VW||Driver assistance||The Vision Zero Project announced in 2013 included plans to adopt advanced information assistance for the driver, autonomous stopping and self-parking (Note 1) on production cars. "Zero" refers to "zero accident."||After a few years|
(Note): Self-parking is an autopilot technology for a car to park itself in a parking space without the intervention of the driver.
ITS World Congress: Special Features
Topic 1: Automated and Connected Vehicle Technologies
Leaders from global auto industry discussed the future of ITS technologies at the panel discussions in the atrium venue.
|- Speakers -|
|GM||: Mr. John Capp Director, Electrical and Controls Systems Research Lab.|
|Ford||: Mr. Peter Hardigan Director of Sustainability, Environment & Safety Engineering|
|Volvo Trucks||: Mr. Jan Hellaker Vice President Transport Solutions & Service Advanced Technology & Research|
|BMW||: Mr. Klaus Kompass Vice President, Vehicle Safety|
|Nissan||: Mr. Kunio Nakaguro Corporate Vice President|
|Toyota||: Mr. Moritaka Yoshida Managing Officer, Chief Safety Technology Officer|
|University of Michigan (DOT)||: Dr. Peter Sweatman Director, Transportation Research Institute (representing DOT opinion)|
Each speaker was given five minutes to introduce their companies' original technologies in Part 1 of the panel discussion. In Part 2, the speakers presented their visions and approaches regarding ITS and autonomous driving as asked by the moderator. Each part lasted for thirty minutes. Excerpts from the hour-long talk session are introduced below:
Part 1 Introduction of technologies
All speakers proudly unveiled their technologies, but virtually all of them, including DOT, felt that it was too early to entrust autonomous driving alone for safety.
General Motors: The ultimate form of Super Cruise is an omnivorous approach of autonomous driving that relies on sensor fusion and infrastructure information to cover all conditions including bad weathers. The system will be introduced in phases starting in the late 2010s.
Ford: Citing Henry Ford's famous "people would have said they wanted faster horses" quote, the speaker stressed that automobiles developed for the freedom of mobility rather than for autonomous driving. The company is also developing autonomous driving technologies.
Volvo Trucks: Introduced the SARTRE and other projects which are under demonstrations toward market launch. The speaker referred to the automated construction vehicles (focusing on higher safety and shorter downtime) and stressed that overall success relies much on cooperation among all concerned.
BMW: Track Trainer is a system that judges the ideal path, acceleration and deceleration once the course information is provided. The system has been proven on some race circuits. Traffic Jam Assistant is an adaptive distancing and lane-keeping control system in congested traffic available on production cars starting this year. Remote-controlled parking system will be developed next.. Safety function will not be available until much after 2020 as it will require accurate ambient information and appropriate response technologies.
Nissan: Safety is the most critical feature of all ITS functions. The speaker presented non-technical aspects of requirements including regulations and social acceptability of ITS. He also stressed the importance of joint researches with universities and other non-industrial partners (Nissan is the only automaker that has declared to sell autonomous vehicles in 2020).
Toyota: Unlike other modes of land transport, an automobile provides the freedom of mobility and fun of driving. However, traffic accidents, aging society and road congestions are calling for technical innovations including autonomous driving and collaboration with infrastructure. Toyota sees autonomous technology as co-driver. The V2I is designed to supplement information available on-board.
DOT: The driver should remain responsible at all times regardless of any assistance or information that may be available. Road forms vary from highways to urban streets to roads in residential areas. Further research, investigation and evaluations are highly recommended. Autonomous driving technologies should be introduced in phases to be determined.
Part 2 Q&A regarding autonomous driving
Speakers' honest views were exposed during the question and answer session.
Q.: People are talking about Google technology with much enthusiasm. What is its impact to automakers?
GM: The public interest is quite understandable. It is what drives the progress of automobiles. As far as vehicle control is concerned, Google technology is an application of old technology that has been around for 20 years. The time is brought to a head today as a result of development of peripheral devices.
Nissan: Google technology has been around for quite some time. We think it is getting public attention as it has become realistic today because of the maturity of the introduction scenario and the collaboration with information technology.
Q.: What technologies do you think will be realized by 2020?
BMW: It's easy to have several demonstration cars running on public roads. Many challenges should be addressed on the part of both the automakers and infrastructure, however, to establish communication among more cars and infrastructure.
Ford: The first step is to expand the scale of the currently available V2I (Vehicle to Infrastructure) communication and verify its benefits. We must also verify the fusion between direct V2V communication by DSRC and wide-area V2I communication in junction with its benefits.
Q.: What is the social acceptability of autonomous driving?
Volvo Trucks: We don't think the success of SARTRE will lead directly to social acceptability. The real road environments are far more chaotic. Autonomous driving will start in a restricted area.
Toyota: It's good to have conventions like this. On the other hand, we must establish social consensus regarding the definition of autonomous driving, scope of its application and its purpose. Autonomous driving alone will not live up to social expectations. We need fusion with the entire transport system including V2V and V2I communication. Social acceptability will increase as we present opportunities that will deepen social understanding.
Q.: Who would be the targeted users of autonomous vehicles?
Ford: Among various targets, for instance, we are thinking of enabling handicapped persons to use such vehicles.
GM: We believe everybody wants automation as in the case of power windows and automatic transmission.
BMW: Our slogan is the fun of driving but we are also aware of environments that spoil it, such as road congestions and parking. Autonomous technologies must start in this type of situations where there are no risks of fatalities or injuries before it is introduced in risky environments such as highways and urban streets.
Q.: Is collaboration with universities or non-automotive industries (software, computer, etc.) necessary for autonomous driving?
Volvo Trucks: It is important to maintain the momentum towards automation upon the cooperation of all concerned.
Nissan: We must collaborate with anyone including those in non-automotive industries who are, for instance, capable of developing artificial intelligence for onboard systems.
Nearly all speakers expressed the following views for autonomous driving:
1. Google technology is a kind of previously known one to all automakers.
2. Many challenges should still be addressed before applying it to a massive number of automobiles.
Contrary to the public enthusiasm, the automakers did not indicate definite timeline of market introduction. It had a modest tone that gave an impression that "it will not be implemented in medium term."
ITS World Congress: Exhibits and technical sessions
MarkLines visited the exhibition and technical sessions after the panel discussion. Suppliers related to driver assistance had grand-scale booths aggressively seeking business opportunities. The information below may indicate the present status of autonomous driving and safety functions:
Costs vs. function of driver assistance systems
Since the technology is still under development, both function-oriented high-end products and cost-oriented products have been introduced. Some suppliers including Denso have both lines of products that are proposed to automakers:
Three grades of safety assistance systems proposed by Denso
|Entry package||Standard package||Premium package|
|* Lane departure prevention * Night vision assist * Road sign recognition assistor * Low-speed crash avoidance assist (vehicles ahead)||* Lane departure prevention * Night vision assist * Road sign recognition assist * All-speed ACC * Crash avoidance assist (vehicles ahead, pedestrians)||Standard package + * Road departure prevention * Crash avoidance assist (crash at meeting point)|
Novelty products are apt to cause differences among parts suppliers because of the varying level of technical expertise. However, Denso is unique as it presented both the best safety and the lower-cost safety plans at the same time. The parts suppliers seem to have burst into cost competition in crash-preventive systems in vehicles. Various grades of safety functions are produced from the trade-off between safety and cost. Unless the vehicle users have correct understanding of the systems available on their vehicles, an accident may occur due to over-confidence in their cheap systems.
Limits of person and object sensing
The person and object sensors, the eye of a vision system, are developed based on various devices including radar, cameras and sonar. These devices have advantages and disadvantages and sensor fusion is often used to complement each other. Another trend for sensing is reducing the number of sensors to reduce the overall system cost.
In either approach, the accuracy of detection and recognition is the key to the system's safety performance. Attempts are being made to improve the limits of sensing. Reported below is one of the attempts in bicycle recognition by IMRA Europe S.A.S. which is owned by Aisin Seiki:
Development of bicycle recognition technology by IMRA Europe S.A.S
|Title||Real-time Bicycle Recognition for Intelligent Rear Collision Warning System|
|Description of development||Recognition of a bicycle from a backing-up car using stereo cameras with ultra wide-angle fish-eye lens|
|Outcome||Obstacles can be detected at high recognition rates over 99% but the best recognition rate for bicycles is about 95%. The company continues to improve the accuracy.|
|Processing of camera information Constant image processing presents a large load to the system. For this reason, constant monitoring using plural CPUs is limited to an object simplified by graphical processing. The following sequence of recognition is used: * The ground is drawn as a simple, flat surface in the system based on the camera information * An object projecting above the simplified surface is searched constantly whether it is stationary or moving (detection) * An object of risk in the vehicle's path is selected * The selected object is analyzed in details to determine its speed and "identity" (recognition) The sequence of recognition is illustrated below. It represents a flow of high calculation load including multiple loops. The image data processing capacity needs to be improved for further improvement of the recognition accuracy. The system is also prone to generate false recognition and missed judgment despite its performance.|
Source: Lectures at ITS World Congress 2013 Technical/Scientific Sessions TS045-2093
As shown in the above example, object sensing contains cases that cannot be detected or recognized at certain rates. Once collision prevention devices are adopted in many vehicles, the pedestrians may start thinking they are being "detected" by the driver and the drivers may start thinking the device is "detecting" the pedestrians. As a result, both the pedestrians and the drivers may become less attentive to each other. Enlightenment in all segments of transport is necessary to inform the limits of detection and recognition.
Motorcycles being left out of recognition: higher risk of accidents than the status quo
The main focus of driving safety assistance is shifting from the recognition of vehicles to that of pedestrians. When the technology is mounted on vehicles, the drivers may grow less attentive (Note 1 below). Motorcycles and other objects that are left out of detection are not only left unrecognized by the system but may be overlooked by the driver as well because of over-reliance on the system. Honda manufacturers and sells both cars and motorcycles, and is testing how motorcycles may be recognized. The current status of its initiative was revealed in technical sessions:
* MAI (Motorcycle Approaching Indication) system in development
The V2V device is installed on both vehicles and a meeting at intersection is detected by means of intercommunication. Although this system works with a motorcycle crossing or heading-on, further improvement is still on-going to detect motorcycles approaching in sharp angle or emerging from blind zone by vehicles, with safety margin.
(ITS World Congress TOKYO 2013 Technical/Scientific sessions TS001-3257、TS012-2125)(Note 1) Please see the second example in the next clause <Future of Autonomous Driving> about the relationship between autonomous driving and accident rates.
Future of Autonomous Driving - Challenges
From the viewpoint of railway technologies - Has a fully-autonomous system been implemented?
Upon returning from the enthusiastic atmosphere of the congress, the reporter reviewed autonomous-driving literatures to look at ITS from different perspectives from past experience of automotive or other industries:
Among different modes of land transportation, autonomous driving has been implemented to certain extent with railway vehicles that only require longitudinal control (acceleration, deceleration). Operations of express or other high-priority trains have yet to be automated. Autonomous driving is not possible either in critical railroad conditions including delay recoveries. Autonomous operations are implemented only when the system can simply stop its operation in the event of malfunction. In other cases, a driver operates manually. These facts indicate following findings for autonomous driving of railway vehicles:
|*||The potential risk of failure continues after technical implementation of autonomous systems (Think about failure frequency of PCs and mobile phones around us. Reliability of these devices must be validated when used onboard to determine if we may risk our lives.)|
|*||Railway control is considered relatively simple, yet it is far from perfection as described above. Road transportation is even chaotic and the risk is much higher. Implementation of autonomous driving with automobiles is a difficult task.|
From the viewpoint of human behavioral science - Will reducing the driver's burden contribute to lower accident rates?
It is a common notion that accidents can be prevented by reducing the driver's burden whether by fully autonomous driving or not. But is that a fact? The current status of traffic accidents was examined with the automatic transmission (AT) that is a leading automation feature on vehicles. Unfortunately, study results tend to suggest the contrary indication:
* Vehicles: Washino et al.: 10 Years of ITS. Jidousha Gijutsu (Automotive Technology) Vol. 64, No. 1, 2010.
* Motorcycles: Chousa Kenkyu News (Research News). Japan Safe Driving Center. No. 18, April 2009.
Both studies concluded that the minor accident rate of vehicles with the AT was twice as high as that of the MT vehicles. This fact suggests that maintenance of the driver tense reduces his or her inattentiveness. When an incomplete safety device is mounted on the vehicle, it is highly probable that more accidents will occur due to the inattention on the human side.
From the viewpoint of health hazard - Allowable millimeter-wave radiation
The car-mounted radar sends a beam of electromagnetic wave to the targeted object and receives the reflected beam. We are exposed to millimeter waves more frequently than ever as a result of progress of communications technology. The exposure level on roads will increase when radar is installed on more vehicles. According to the guidelines regarding "protection from radio wave environment" issued by the Telecommunications Technology Council of the Ministry of Internal Affairs and Communications, the specific absorption rate (SAR) on the human body (whole-body exposure) is set to 0.4W per 1kg of body weight over the average time of 6 minutes. This corresponds to 4mW/cm2 on a standard sized body. For instance, the electromagnetic wave radiation from a mobile phone to the human head is several times lower than the guideline. Since output of the millimeter-wave radar is smaller (2 to 3mW, or hundredth of that of a mobile phone), the whole-body exposure may not be an issue. On the other hand, some point out that further study is necessary with regard to topical exposure, especially the effect of concentrated radiation to the eye. Since motorcyclists are exposed directly to the radiation for rather long time, degree of protection is necessary to be investigated.
Autonomous driving still remains challenges to be addressed before introducing it into the society. Despite the public interest (expressed as 'enthusiasm' in Special Features during the ITS World Congress), all but one automakers declined to say definite timing to implement fully-autonomous driving. This moderate attitude is understandable given the current circumstances.
<Automotive Industry Portal MarkLines>