Controlling smart cars, part 2

Nov. 14, 2016
Developing the controls for autonomous cars will go faster if designers draw on the knowledge and experience of other industries.

This article suggests improvements in controls for self-driving cars based on experience in the process industries. The state of the art, as exemplified by the Tesla accident May 7 in Williston, Fla., is discussed in part 1.

 Over the past century, general industry (power, chemical, oil, nuclear, etc.) faced the same safety concerns that the automobile industry is facing today. One camp argued that manual control is safer because all sensors can fail, computers can freeze, etc., while the other camp argued that automatic control is better because operators can be intoxicated, untrained, tired, distracted, etc. Both camps assumed there's no third option. It took a long time to realize that one must not choose between manual and automatic control, but should benefit from both simultaneously. In other words, one should always consider both and select the safer one for control. This is called Selective Safety Control (SSC).

It would be wise for the transportation industry to learn from the experience accumulated in other industries, and not attempt to "rediscover the wheel."

Where we are today

 In just the U.S., more than 35,000 people died in car accidents last year. According to most estimates, smart cars could eliminate 95% of these accidents. Yet 71% of the American public believes that smart cars are less safe than regular ones. These numbers contradict each other, and that's unfortunate because technological advances require public support, buy that support won't evolve until there's confidence in the safety of such technologcal advances. It's for this reason that I'm writing this series of articles.

[pullquote]Many smart cars are electric, and the acceptance of electric cars is growing. Tesla sold 50,000 such cars in 2015, 82,000 in 2016, and has some 400,000 customers on its waiting list for 2017. In 2017, it hopes to market an electric car (Model 3) that will cost around $35,000 and can travel 240 miles on a charge. Other car manufacturers (GM's Bolt, Nissan's Leaf, BMW's i3, Volkswagen's e-Golf, etc.) place less emphasis on driving distance per charge because they're focusing on commuting.


The smart car designs on the road today use autopilots and require that drivers keep their hands on the steering wheel, so they can take over control when needed. As of today, Tesla has some 70,000 such cars on the road, and others like Mercedes-Benz plan to start selling them in 2017. Tesla is also working on driverless cars—it will have all the sensors needed for autonomous driving in its Model 3, but will not enable the system until more testing is performed. Model 3 has some 300,000 preorders, which Tesla will start filling in 2017. Other manufacturers, like GM and Audi, are also taking an incremental approach by developing cars capable of gradually updating from the autopilot mode of operation to the driverless one. Others are starting out with limited application goals—an example is Google (working with Otto), which plans to market trucks that will be self-driving only on highways. Development of completely driverless cars is still in the testing or field trial stage (Uber in Pittsburgh, Google in California, Ford, etc.) and are generally expected to be available by 2021.

Autopilot vs. driverless

 People asked whether the autopilot was on or off at at the time of the Tesla accident in Florida. This is the wrong question. Just as we don’t need to turn on the seat belt alarm to be reminded if the belt buckle is closed, the autopilot should have been on all the time—in addition to the driver being always ready to take over. The important difference between the design that existed in the Florida eccident and the correct one is that neither the driver nor the autopilot should be able to overrule the other (be the primary source of control). However, whenever they disagree, the control system (SSC) should automatically select the safer one to control braking, accelerating or steering.

[sidebar id =3]

As I noted earlier, smart cars have great lifesaving potential because they can provide safe driving even if the driver is tired, panicked, drunk, under the influence of drugs, inexperienced, distracted or has slow reflexes, decreasing vision and hearing, etc. Yet their lifesaving potential is reduced because all computer systems can fail, and even if they don't fail, their software can be hacked or be insufficiently sophisticated to recognize complex situations.

[sidebar id =1]It's for this reason that the software packages in operation (Figure 1) can safely handle only simpler tasks such as changing lanes, stopping at red lights, parking or keeping safe distances between vehicles, but they can't yet distinguish between, say, a pedestrian trying to hitch a ride or a police officer flagging the car down. Such "fuzzy" conditions haven't yet been effectively enshrined in computer code, while the human driver can usually recognize them.

A great advantage of smart cars is that the software of the whole fleet can be improved over the air whenever new information becomes available. In other words, whenever the causes of an accident are determined and the software is modified to prevent reoccurance of that accident, the revised software package can be immediately transmitted wirelessly to the entire fleet. Hence, the safety of the fleet can be continually improved.

Advocates of driverless cars argue that using the autopilot is less safe than autonomous driving because even if the driver's hands are on the steeringwheel, the driver, being passive, can’be expected to snap back and make split-second decisions when needed. They refer to studies that have found that the time needed to "wake up"the average driver is 17 seconds, and a car moving at 65 mph travels five football fields during that time. Yet as of today, "driver-assisted collision avoidance software" (autopilot) is better developed, and so for some years more, the "hands on the wheel" mode of driving is likely to prevail.

It's also interesting to note that the major process control firms (ABB, Emerson, Honeywell, Schneider Electric, Siemens, Yokogawa, etc.) seem to be doing very little to develop sensors and control software for this new market. Newer companies are starting to fill this gap, such as Nirenberg Nouroscience, Otto or Saips in the fields of machine and computer vision, and Velodyne in the area of miniaturized lidar (laser imaging and ranging), etc.

The next part in this series will discuss the capabilities of today's sensors, potential for developing additional or better ones, and improvements in control software packages needed to improve smart car safety.

[sidebar id =2]

About the Author

Béla Lipták | Columnist and Control Consultant

Béla Lipták is an automation and safety consultant and editor of the Instrument and Automation Engineers’ Handbook (IAEH).

Sponsored Recommendations

Measurement instrumentation for improving hydrogen storage and transport

Hydrogen provides a decarbonization opportunity. Learn more about maximizing the potential of hydrogen.

Get Hands-On Training in Emerson's Interactive Plant Environment

Enhance the training experience and increase retention by training hands-on in Emerson's Interactive Plant Environment. Build skills here so you have them where and when it matters...

Learn About: Micro Motion™ 4700 Config I/O Coriolis Transmitter

An Advanced Transmitter that Expands Connectivity

Learn about: Micro Motion G-Series Coriolis Flow and Density Meters

The Micro Motion G-Series is designed to help you access the benefits of Coriolis technology even when available space is limited.