This article also appears in
Subscribe now »

Safely steering around a detected obstacle can be demonstrated now, but has a complex road to production.


Ford demonstrates collision avoidance with automatic evasive steering

The Ford Focus was moving down the right lane at almost 40 mph (64 km/h) headed straight for a stopped SUV. The driver never took his foot off the gas pedal, and unless there was a dramatic intervention a serious accident seemingly was sure to occur. At first a warning had appeared on the dashboard’s digital display, then a chime (both unheeded). The brakes automatically applied, instantly followed by the electric power steering automatically swerving the car safely into the left lane, around the stopped vehicle.

This journalist was the passenger, a Ford test driver was behind the wheel, and the Focus was the system-development vehicle. The driver maintained readiness to ensure that what the sensors detected would be translated by the hardware into an evasive maneuver. This all occurred under controlled conditions at the Ford Dearborn test track (left lane unoccupied), so we were never in danger.

It was an effective demonstration of the potential for advanced accident-avoidance technologies. Ford estimates that the world population of motor vehicles will be in the 2-4 billion range by 2050. That’s a prescription for global gridlock, Randy Visintainer, Director, Research and Advanced Engineering at Ford, said at last week's North American International Auto Show (NAIAS) in Detroit.

The evasive maneuver also was one aspect of Ford’s approaches to the time, perhaps in the 2020s, when automated driving—under some limited conditions, anyway—will become more commonplace. One is a “no-holds-barred self-driving vehicle” with a very powerful computer system.  “That’s more about what is possible in terms of hardware and software, with the timing unknown,” explained Greg Stevens, Global Manager of Driver Assistance and Active Safety Research and Advanced Engineering.

Building-block approach

However, what the customer has seen, and will continue to see, is the building-block approach, Stevens continued. It started with assist-type and warning features, such as active park assistance, blind-spot monitoring, lane-drift detection, cross-traffic warnings, and brake assistance. These initially are being followed by lane-drift correction, first by pulsing the antilock brake hydraulic module, and then by moving (as is being done on some cars today) to electric power steering (initially a fuel-saving addition) to correct for lane drift, and to automatic braking for collision mitigation.

If these all sound like “baby steps” it’s because they are; accepted limits needed for applied steering torque have not been set by any regulatory agency, such as NHTSA (U.S. National Highway Traffic Safety Administration) or even within technical associations. (Although there is an SAE International working group on the subject.) And further, Ford needs much more information on overall road conditions in the immediate area to do more engineering integration.

However, Stevens said, “If we create enough data, we can go with corporate standards.”

The building-block approach actually uses three “development paths,” he explained. One is the parallel self-parking vehicle, a low-speed steering maneuver. Today’s Ford's parking assist system—in which the driver maintains control of shift lever position (forward or reverse), throttle, and brake pedals—has 12 ultrasonic sensors. There are six in front, six in back, with one at each side, front, and rear “looking” to the sides to “measure” the parking space. If the sensors can detect the curb, they use it; otherwise, the system aligns the car with the vehicles in front and back. Nothing startling there.

The system soon also will do perpendicular parking. The sensors and controls in the next generation will park the car with the push of a button, operating the steering, throttle, brake, and transmission. That will be followed by remote-control parking in which the driver exits the car and uses a remote (such as the key fob) to initiate the parking maneuver—useful in cases such as a garage too cluttered for the driver to exit.

Highway commuter assistance

A second development path is superhighway commuter assistance. Adaptive cruise control and lane-keeping assistance using a torque input to the electric power steering to maintain lane alignment are available developments, employing radar sensors and a camera. So if they are integrated, we would have a car that can drive for you in the same lane, controlling throttle, brakes, and steering to a modest extent, Stevens noted.

This highway feature usually is speed-range-limited, so improvements in sensors, response of the hardware, and robustness of the software are necessary to move to higher speeds, he observed, “because at higher speeds things happen faster.” However, the current level of the technology is a form of “traffic jam assistance,” working well at 25 mph (40 km/h) and below.

The collision avoidance path is in an early stage, even if it seems like it should be possible very soon. Collision-impending warning systems are in wide use on new cars, and some cars can apply the brakes.

However, the automatic brake-application systems come with limits.  On the new Hyundai Genesis, the warning system only calls for full brake application at 50 mph (80 km/h) and below. And although Hyundai engineering believes this will be enough to avoid most collisions for vehicles in this speed range, it cannot positively assure a stop before collision in every case.  Ford’s Stevens told Automotive Engineering that his company today is using radar-fed warning systems with pre-charging of the brakes for faster response, but presently leaving the responsibility for actual brake application to the driver.

Way to evasive steering

Adding evasive steering might seem like a logical way to cover the “gray" detection areas inherent with automatic braking, although getting to that stage is far from simple, Stevens said. In the Focus, a forward-looking radar sensor and camera detect a potential collision, in this case up to over 200 m (656 ft) ahead. Rear side-mounted blind-spot-detection radar sensors "look" to the side and toward the rear to "see" if there are open spaces in adjacent lanes. But better ones are needed, and coming.  An improved wide-angle camera not only could identify objects ahead, but report on the type and condition of road shoulders (are they paved, at the same level as the road, or unpaved and otherwise possibly unsuitable?).

When steer-around development has advanced, Stevens continued, Ford will pick the most promising vehicle applications and start the implementation, which means a more rigorous engineering process and more detailed data, followed by the inevitable business case.

Finally, Ford would move the feature to a production team, working with component suppliers, advanced calibrations, and eventual release for production, with later migration to other platforms. The feature likely would not go to a “super-module,” Stevens said, but into individual modules already in use, such as ones for power steering control, automatic parking control, engine and transmission, dashboard display, and body functions.  Partitioning the existing modules, he noted, would enable the feature to more easily be incorporated in the existing electronic architecture.

The latest generation of Ford’s automated driving research vehicle, a Fusion hybrid, was featured at NAIAS. It features high-definition “LIDAR” (laser-radar) sensors, roof-mounted and spinning rapidly to create a real-time map of the existing area, which is compared with stored data on the geographical area. The first generation of these sensors weighs almost 29 lb (13.2 kg). A new generation, mounted on the research Fusion, weighs about 2 kg (4.4 lb). A prototype next generation unit is less than 1 kg (2.2 lb).  So at some point, Ford believes, such sensors will shrink for installation effectively out of sight, but delivering quality data.

The advantages of traffic management and collision avoidance from vehicle to vehicle communication are apparent, but raise still further issues. However, a well-integrated system on each car will be a giant step (from the many small steps) forward.

Continue reading »
X