Problem Solving for Self Driving Cars: a Model

By Herb Wiggins, M.D.; Clinical Neurosciences; Discoverer/Creator of the Comparison Process/CP Theory/Model; 14 Mar. 2014

.

As predicted the interactions between self driving cars and their environments of fog, snow and ice, intense rain, and other weather conditions (tornadoes and hail), plus the road construction site complexities created by loss of landmarks and other stable signs marking safe driving lanes are clear. But ever so much more clear are the problems being caused by humans who are complex systems and NOT rational drivers in the usual sense. This has created at least one commonly seen pattern, the rear ending of human driven vehicles by self driving cars. And the teenage propensity to “play” in traffic, approaching self driving cars and forcing them to stop unexpectedly by stimulating their auto-braking to kick in using various methods. Those persons will find, many other ways of confounding auto-driving cars.
.
The abilities of self driving cars to deal with these complex system inputs is a real problem, because humans are NOT driving according to traffic laws and good driving rules, nor least energy rules, necessarily. & this means that such human/self-driving car interactions are going to be a real problem, in addition to the environmentally complex systems which such car decision making systems will have to face, too.
.
Simply stated, humans will stop for the “heckuvit”, too often, and this results in self driving cars crashing into them from the rear. Now how does the usual driver deal with this very real problem? And how do professional drivers do so? & what are the causes of traffic chain reaction pile ups in fog and snowy weather, which are likely to show us the way out of those complex system problems?
Taking a case by case series of rules, the usual driver will leave about 10-20 feet between a car and the one in front, according to a general rule of 10-20′ and traffic conditions per 10 miles/hour speeds. Thus driving at 40 mph, there would have to be about 40-100′ between the car and the one in front.
.
But, this is a probability confounded by the weights of the vehicles. Large trucks need a longer stopping distance in those being driven in front of others, and those trucks behind other trucks, than shorter and lighter cars. And in addition, lighter cars can often safely stop at shorter distances, as well. There are many other factors, such as if the front driver stops unexpectedly due to a mistake, or other factors. The human driver will often plow into him, esp. in heavy traffic conditions. In addition, if a driver strikes a heavy pole, abutment, or tree near the roadside, the car might well angle out into the lane, and block that lane requiring a fast, complete stop by a following vehicle, esp. in oncoming heavy traffic. That sudden stop is NOT covered by the speed/distance to stop rule, and will often result in a rear ending.
.
Also, if an oncoming car, bike, motorcycle or pedestrian suddenly pulls in front by whatever reasons or conditions, the front car will slam on the brakes & the car behind is likely to hit it. The complexities of these issues are very, very great, and depend largely upon traffic conditions, heavy/rush hour, or lighter, or due to lighting condition, esp. at night, where visual information is lessened, often substantially. The medical conditions and mind altering chemicals also are complexities. So, how to solve these complex system problems?
.
We go back to a rule. Professional drivers are usually better than the average, amateur drivers, just as professionals in most fields are much better than amateurs. They have encountered many, many types of the same situations in which most drivers have not. Thus they have seen the higher, more complex patterns of stabilities repeating. because they have developed more methods to drive safer, & are far, far more able to adjust quickly and avoid accidents. But what are the features of professional drivers? Those are a series of skills & driving styles, and processing of the complex events which are encountered daily. Driving in heavy traffic, they become more alert, brain works faster, and they tend to slow down, according to conditions. Faster driving shortens reaction times, in other words. That principle needs to be built into driverless cars, one way or the other. In other word, the rule is when conditions get critical, recognize those many types and Slow down!!
.
And this is the point, studying professional drivers and how they operate in such conditions will provide the data, esp. with 10-12 or more of those experienced drivers, about they they respond, in real, repeating, efficient ways to such conditions. Those are the stabilities of methods which will show up, and that information can be extracted by comparison processing to be found, and created. & it will be far more efficient as well and result in faster progress than other methods.
.
First of all, what kind of drivers will be needed, who drive professionally? Not taxis, or delivery for cars. But professional drivers who work out of their cars, and related auto industry persons who operate vehicles, such as salespersons. They have to get to places quickly, safely and effeciively. And they are practiced at doing so.
.
Regarding trucks of all kinds, the mostly personal delivery systems would be best, although not UPS, etc, but smaller companies who drive.  In this way the specifics skills sets of such drivers can be delimited and then compared. Each driver will have his own sets of skills, that is style of driving. and mostly by comparing those many kinds of methods will give us the best series of skills needed to drive most safely and avoid most accidents. Clearly, those with the highest level of no accidents/incidents, will be the gold standard of whom to study. For obvious reasons those without drivers’ licences are out. But the standard needs to be an established, professional driving history plus least accidents ratings. Such studies can also be used by insurance companies to predict who is most likely to have and keep a safe driving records, because, clearly, they haven’t had accidents. And that can be easily correlated and studied to figure out HOW, that is the structure/function relationships of good driving.
.
Oddly enough the personality traits of good drivers are also on analysis here, and those traits of sociopaths are not likely to result in good driving records. But those driving records of pros who drive with a good history of safety, not only avoiding accidents themselves, but also in Avoiding the bad drivers, and those signs such give off, when the professionals are driving about. It’s complex system, but doable. It will take work, but the inevitable creation of data about what safe driving skills are the most effective will be a valuable product, as well, from such studies, and will be of benefit to both insurance companies and even traffic pattern studies. & can be incorporated into Drivers education classes found in most all secondary schools statewide in the US. & the legal classes and such of those driver re-training after having accidents, law breaking, and so forth, too.
Thus these studies will be a of widespread benefits, not only for creating safer self driving car strategies, but widely, even internally, as well. What’s the value of making driving safer in a nation with 100K+ injured in vehicular accidents each year and 10K’s of deaths?  Incalculable due to the likely reductions of such problems which these studies & their applications will create.
.
And setting up sudden unexpected conditions for them to drive in, will also show how they deal with those, as well. But in a real time world, NOT on an isolated track somewhere, wherein the reality of the situation does not correspond well, and introduces too much bias into the conditions for really good, practical outcomes research.
.
Thus we use professional drivers, and their methods and skills, which can be observed, and delimited quite well, to guide by human supervision, HOW to avoid those problems of weather, construction areas, and heavy traffic, and those problems so often seen in driving which result most likely in accidents. & then compares those to the average driver, for significance. Because most persons have a set of multiple skills they use, which are often unique to them, a number of professionals will need to studied to see as many as possible of the skills used to more safely, efficiently and effectively drive.
That will take some work, but will very likely result, if properly and creatively approached, in methods which will improve human driving as well as that of self driving vehicles. Human supervision of AI is well established and provides, by long experience, far higher performance in AI as has been shown many times. Thus professional drivers will provide an incentive to improve not only self driving cars, but human drivers, as well. Because it can be used practically and will create a virtual, newly created science of driving methods and strategies. It will make humans better drivers as well, and who could object to that? Get that dopamine boost operating in humans to make the investigations grow faster, and get more financial support for such essential, critical studies of driving, as well.
.
In addition, each of the methods used by professional drivers which work,can be detailed and then tested for which are the best, most efficient ways of avoiding accidents in the higher, more likely, observed stabilities of accident events. Alos which driving skills result in getting most efficiently to the place needed to be visited. Least energy and travelling salesman problem solutions can be addressed as well, including more efficient methods to get to places in series, or alone.
.
Rear ending, avoiding potential collisions from approaching cars, sudden turns into traffic from the right and left; those drivers who cut off other drivers, and suddenly veer in front of professional drivers. The data are already there with those good drivers and there is no need to re-invent the wheel. Plus the fact that as the methods are found, they can be then experimentally improved upon without limit, too, in each and every case. Each new method can also be combined with detection devices and other AI to improve personal, human driving as well. This will exponentiate growth of both human and driverless cars’ driving strategies and methods, and devices, very likely, if done properly.
This will create rules for driving, which will not only improve the strategies of self driving cars, but human drivers as well. Given the success of AI and expert systems in applying those rules already seen, this will substantially shorten & speed up the rate at which self driving cars can develop more quickly and better deal with unexpected events while driving. This will solve thye problems of self driving cars and speed applications to this new capability. And further, will also improve most driving overall, when those principles are elucidated and worked on.
.
This is how it’s done, naturally, anyway. Learn from those who know best in a scientific way, and then go from there, by applying good driving principles to humans and self driving cars. And “knowing where to go”, gets us there a LOT faster, by cutting out the trial and error, slow learning meanderings to far, far fewer, does it not?
.
& further teens can be asked about how they have been able to interfere with self driving cars, and thus effectively use the more creative of them to find & deal with potential problems, if they are repeatable enough, to avoid more problems. We can recall that many chess programs formerly developed which had peculiar quirks which humans did not have, and those could be exploited to beat them most every time. Be assured that self driving computer programs and devices will have those same quirks that Sargon had in chess. And the more complicated they are, the more likely those will be created by the programs, too.
.
This is application of the comparison and least energy processes plus complex systems in a workable, realistic, practical way to provide an incentive for humans, as well as self driving car makers, to solve some of the real problems of driving, across the spectrum of the complexities of driving more safely and better.
Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s