Autonomous Vehicle Technology by Luis C. Flores, P.E., ACTAR, CFEI

Autonomous Vehicle Technology: Looking One, Five and Fifteen Years into The Future
An Accident Reconstructionist’s Perspective

I recall a friend of mine asking me about a year ago if I was worried about autonomous vehicles making my job as a vehicle accident investigator essentially obsolete. I retorted: as long as it’s still humans behind the software, there will always be room for, well, human error.

As you can imagine, the proliferation of autonomous vehicle technology relies on more than the safety mechanisms introduced by vehicle manufacturers can offer. More importantly, a coordinated effort amongst traffic engineers and the general public is key to its success. Autonomous vehicle technology operates on the idea that vehicles will be able to: “talk to” or “see” each other; interact with the ever-changing infrastructure they operate on; and account for the non-motorized traffic they encounter (pedestrians, bicycles, wildlife, etc.). Radar and Light Detection and Ranging (LiDAR) technologies sense objects near and far, GPS keeps the vehicle on its route, and computer vision (optics) detect markings and signs along the road.

Photo Credit: Texas Instruments

Photo Credit: Texas Instruments

The Society of Automotive Engineers currently identifies six levels of vehicle automation:

  • Level 0 is no automation.
  • Level 1 is driver assistance (i.e. electronic stability control and adaptive cruise control).
  • Level 2 is partial automation (i.e. driver can deactivate automated accelerating, braking and steering and take over).
  • Level 3 is conditional automation (i.e. autonomous in limited environments such as freeways).
  • Level 4 is high automation (i.e. all environments except severe weather).
  • Level 5 is full automation.

Perhaps the most overlooked item on the autonomous feasibility checklist is cybersecurity. Just last year, WIRED magazine published a story on how a pair of cybersecurity researchers managed to take control of a Jeep Cherokee from its climate control to its audio system. And this year that same pair demonstrated a slew of new attacks perpetrated through the vehicle’s communication network, or CAN bus. It is no coincidence that shortcomings like this one have resulted in a senatorial plan to introduce an automotive security bill in 2015, legislation that would direct the National Highway Traffic Safety Administration (NHTSA) and the Federal Trade Commission (FTC) to establish federal standards to secure personal cars and protect drivers’ privacy[1].

A more obvious and physical deficiency in autonomous vehicle technology is the inability for the distance sensing radar and central computers to operate under perilous weather. A 2014 Hyundai autonomous vehicle competition in South Korea proved just this when on day one of the competition, a Veloster performed a series of tests flawlessly under sunny weather. These varied from following the speed limit to backing into a parking spot. During day two, a rainy and foggy morning, the same vehicle failed more than half of the same tests.

1 YEAR OUTLOOK | The Semi-Autonomous Present
Already 2016 has proven to be a year flooded with TV ads on emergency brake assist, or automatic emergency braking, as Volkswagen calls it. Some manufacturers have gone as far as using the word “autonomous” in these ads. Generally speaking, the average driver has a perception and reaction time of 1.5 seconds. This is known as the window of time in which your brain goes from “Oh no!” to when your foot initially touches the brake pedal. Logically, braking time and distance vary based on initial speed. To put things into perspective, a vehicle traveling at 60 mph covers about 130 feet in those 1.5 seconds, more than a third of a football field. In addition to minimizing that time window, the additional pressure applied to the brake pedal when brake assist is activated is vastly larger than that of the average human foot. Eliminating that human factor in an emergency braking scenario has proved so effective in Europe that the European Commission proposed making brake assist standard on new vehicles being sold after 2009[2]. Differing from the active brake assist is a passive system known as lane departure warning, designed to keep a distracted or tired driver within the lane markings. This system has naturally evolved into lane keep assist, an active system that will help the driver steer back into position.

While autonomy Levels 4 and 5 are not quite a reality, a start-up company by the name of Faraday Future is scheduled to launch its first hyper-connected electric semi-autonomous vehicle through its Advanced Driver Assistance System (ADAS). These autonomy levels have complex challenges, however. Earlier this year, the Tesla Model S made the news when a driver was killed in a crash with a heavy truck while operating in autopilot mode. As the heavy truck was turning left in front of the Model S, the autopilot sensor failed to detect that an object was ahead possibly due to the trailer’s high ground clearance or reflectivity. In other words, it was a matter of perfect timing exposing an imperfect system.

                                                                                                                                                                                                   Photo Credit: Florida Traffic Crash Report

                                                                                                                                                                                                   Photo Credit: Florida Traffic Crash Report

5 YEAR OUTLOOK | Adventures in Ridesharing
It is difficult to dispute the impact that ridesharing services like Lyft and Uber have had on the average commuter. Taxi cab drivers can certainly attest to that. So it wasn’t surprising when, in September of this year, Lyft announced a partnership with General Motors to launch GM’s first autonomous electric vehicle in 2021. Set to arrive on the market in 2017, the Bolt EV and its all-electric range of 238 miles was built with autonomy in mind. Not to be outmatched, Ford also has plans to embrace full autonomy by 2021 by doing away with steering wheels and pedals altogether on ridesharing vehicles. This year alone Ford is testing 30 autonomous Fusion Hybrid Sedans in California, Arizona and Michigan. Also in September of this year, Uber put the world’s first self-driving car on the roads of Pittsburgh, though a driver was on board for safety reasons.[3]

Imagine doing away with the minutiae of waiting in line to pick up a rental car and instead having said rental pick you up where you need it. The phrase “We’ll Pick You Up!”[4] will take on a whole different meaning. Now imagine getting in the backseat and arriving at your destination without having touched a steering wheel. Rental car companies are set to capitalize from autonomous vehicle technology.

Passenger vehicle crashes can be hazardous enough, but topping out at 80,000 lbs., the catastrophe left behind after most heavy truck accidents is multiplied tenfold. Expectedly, doing away with the liability of an inattentive or sleep-deprived truck driver is appealing to the more than 700 billion dollar a year trucking industry. An additional benefit to the driverless truck is the ability for a fleet manager to monitor and control fuel economy through telematics. This year the Freightliner Inspiration Truck, the first autonomous commercial truck, was approved for use in Nevada public roads and in April, a fleet of Daimler Trucks clocked 1,200 autonomous miles across Europe.

In essence, the lesson is that anything that moves people or goods has potential for autonomy; one that will revolutionize every aspect of business and life from the service industry to emergency vehicle transportation (ambulances, firetrucks, etc.).

15 YEAR OUTLOOK | The Great Curve
It took 37 years from the first conception of the airbag in 1952 by John W. Hetrick to its mandatory implementation on new vehicles in 1989. While comparing this safety advancement in vehicle technology to autonomous driving is not a perfect parallel, it can serve as a guideline for what’s to come in the next decade and a half. After all, electronic stability control was introduced in the 1990s.

By the early 2030s, there should be ample studies regarding autonomous vehicle accident statistics. Currently, 92 people die on American roadways each day. That is roughly 30,000 per year. According to NHTSA, 94% of crashes are caused by human error[5]. Arguably, it is still too soon to simply label all autonomous vehicles as more intrinsically safe, but as they slowly begin to take over our roads, statistics are slated to favor them. Still, humans are stubborn by nature and, to some extent, there is an unquantifiable joy in taking a ride on a sunny day, which means that even fifteen years from now, with autonomous vehicles ubiquitously available and, with safety statistics likely favoring autonomy, there will still remain a large percentage of human driven vehicles on the road. It is also likely that, much like the wonder of looking up and seeing an airplane in the sky as a child faded over time, so too will the public perception of autonomous vehicles as gleaming and mysterious.

Odds are that a joint effort by the government and private industry will be necessary to remove all non-autonomous vehicles, which will take time.

Forensic engineering careers like accident reconstruction exist because of liability. So in the event of an accident in the current semi-autonomous stage that we are in, who owns the liability? Is it the vehicle manufacturer’s division in charge of software? And when full autonomy is in place and a collision occurs between two autonomous vehicles, who will own the liability then? It could be akin to the event of a house explosion, where the manufacturers of the water heater, range, furnace and other appliances are put on notice. I predict that the manufacturers of equipment and infrastructure related to object sensing and navigation in the subject autonomous vehicles will similarly be put on notice.

Furthermore, what happens when the decision-making aspect of an autonomous vehicle is tasked with deciding whether to protect a pedestrian or its occupant, essentially playing God? An article by MIT Technology Review from July of 2015 touched on this most philosophical aspect of driverless cars, where the ethical rubber meets the technological road. A scenario was posed in which a self-driving car must decide between hitting a child suddenly dashing into the road or running into an oncoming van. The human answer ultimately lies in altruism, but what happens when we are no longer dealing with conventional human decision-making? Ultimately, four primary strategies to avoiding an accident are to accelerate, swerve, stop or do nothing.

Since its inception, accident reconstruction has evolved from a career demanding the application of the laws of physics to vehicular accidents to one requiring a valid and logical interpretation of Event Data Recorder data. Meaning that a vehicle that used to communicate the events of an accident through its damage can now further corroborate said damage with a crash data report containing pre-crash status information ranging from individual tire pressures to the various accelerations experienced.

As vehicles continue to evolve, so will the approach to investigating the accidents that they are involved in. Accident reconstruction will require a basic understanding of how and why an autonomous vehicle made the decisions that it made during an accident, making a rudimentary understanding in computer language a necessity.

[4] Registered mark of Enterprise