On Autopilot: Assessing the Safety of Self-Driving Vehicles

Application of innovation and technology have transformed human life in various spheres, however, Self-driving advancement of cars is one of the key human endeavors expected to save people’s lives on the roads. The number of road fatalities goes beyond 1.3 million globally and annually. The deaths are blamed on the human frailties and flaws which innovators and researchers envisage that robots can eliminate through their enhanced precision. Robots do not have human weaknesses like getting distracted, feeling drowsy, or even driving under alcoholic influence. Despite investments of billions of dollars by companies like Tesla, Google, and Uber, disturbing questions regarding the technology are yet to be answered (Shladover 2017, 589). First challenge is the definition of safety for the machines; secondly, the duration of self-driving test considered satisfactory to ascertain safety assurance of the autonomous cars. The research addresses the safety and test worries of the autonomous vehicle.

            To paraphrase a segment of the third paragraph, an autonomous vehicle is a pioneering technology where parameters for measuring safety is still unestablished (Furda and Vlacic 2015, p.2421). Determination of safety and expertise of human drivers depend on measurable qualities which traffic instructors can effortlessly measure and confirm the suitability of a candidate driver or vehicle. Such human qualities are not present on autonomous machine drivers. Measurement of safety for drivers relies on the driver’s ability, but such qualities are not easily measurable for machine drivers (Shladover 2017, 584). The safety challenge emanates from the reasons that government as the regulator of the road use should generate a new framework for ranking safety level for self-driving cars. Safety for the driverless vehicles can be measured based on machine flawlessness or ability to break comparatively few laws as contrasted to human beings.

            The process of ensuring safety for the driverless vehicles may involve intensely different approach as contrasted to human drivers (Furda and Vlacic 2015, p.2421). To avoid what has been regarded as the probable creation of immature teenage machines, road safety can be appraised through a search for perfection in driving discourse. Driverless vehicles can earn their chance on the roads if they ensure minimum possible accidents and maximize on adherence to the traffic rules. It is only through a continuous and regular measurement process through which autonomous vehicles can be vetted and rated as roadworthy or not. Similar applications have been undertaken before by insurance firms which install a gadget called a black box to gauge human drivers’ ability and action while on the road (Black box insurance 2018).

            The black box is an example of a safety test which collects and transmits data to an insurance firm highlighting a comprehensive driving report which includes driving period, acceleration, braking performance, Position, cornering, and speed. In the presence of the variables, the regulator can derive a vehicle’s score especially using an algorithm. While the insurance scores determine the premiums payable, the generated ranking would determine whether to allow or disallow autonomous driver. Comparison of black box score to human driver score should confirm the theoretical idea that machine drivers are better than the human drivers (Black box insurance 2018).

            The driverless vehicle is not the only autonomous automobile tried by mankind. For instance, the Docklands light railway established in East-London has been operational over time and worked as an automated light metro framework. The system communicates using electronic loops which are located 25m apart from the tracks (Fainstein 2017, p.768). When each train passes the loops, the signal is sent to the system database. Such details enable calculation of the trains speed and location and when the connection goes off then trains to stop.

            The London’s Docklands light railway system does not directly reflect a driverless car’s operating environment. While vehicles travel on varying road types, for instance, tarmacked, gravel, and can even be parked on grass, trains only move on railway gauge (Fainstein 2017, p.765). While testing the safety of the driverless vehicle, scores should be generated from different road models and other variants such as at day-time or night-time. Several critics have often pointed out multiple limitations posted by the driverless machines and argue that human beings will still be far better than machines. It will be nearly impossible to the pre-program machine to handle all scenarios in advance for the machine, for instance, a machine’s automated vision system may easily mistake a bus shelter full of people as an uninhabited maize field thus causing massive damage. While human beings have certain typical weaknesses, they have certain typical human traits which empower them to handle every emergent issue in their course of work.

            Steven Shladover states that no single assured test that can determine safety of the self-driving vehicles. Steven who works as a research engineer and manager in the transportation field for the University of California believes that inclusion of multiple tests, the involvement of U.S. regulators and industry members can provide safety for the driverless machine (Shladover 2017, 584). The manufacturers of the driverless cars have to prove that their machines are safe enough to be granted privileges to drive on roads with human beings. Road safety tests will consume time, massive resources and will be a continuous process to allow them on the roads. Production and use of such machines should always be regulated until safety assurance can be ascertained through established testing mechanisms to a avoid flooding the roads with perilous adolescent robot drivers.

            To paraphrase how long the self-driving vehicles must be tested before they are considered safe. The vehicles will need to drive for several miles, for instance, hundreds of millions or even hundreds of billions of miles of distance and gather substantial data to show their safety. According to the RAND Corp 2016 report, safety will be demonstrated through cause of limited number or no deaths or injuries. The report further argues that tests on the current fleets of self-driving vehicles may take tens or hundreds of years to drive the required number of distance vital for conducting a statistically relevant safety comparison.

            Duration of machine testing for the self-driving vehicles is highly vital especially during the teenage testing time as per currently. As opposed to human driving tests test which vets the human ability and skills, driverless cars are tested on several functions. Testing of the software’s algorithms, hardware specifications, and mechanical framework of the cars. Currently, the regulator specifies cars and drivers as separate entities but driverless cars combine both driver and the car as one unit. It is more straightforward and simple to test cars based on the evaluation of key tangible features like steering functionality, braking feature s and several others (Benenson, Petti, Fraichard and Parent 2017, p.23).

            Government as the regulator should demand raw test results from the car showing safety levels. Evaluation of judgment and decision-making capacity of the vehicles is vital and can only be provided by the manufacturers especially on annual basis. For instance, how many times a human driver had to intervene when car algorithms or sensors did not work as expected (Furda and Vlacic 2015, p.2421).

            Critical emphasis must be undertaken on the evaluation of the functioning algorithms that directs the car on what to do. For instance, movement of the vehicle on freeways having trucks, in regions having animals, kids, walkers, cyclists and other roadside features. Testing process should assess how the algorithms behave in case of compromised inputs from environmental factors like rain, snow or mist other weather changes. Testing of the vehicles under more varying scenarios will be vital in exposing limitations or incapacities of the machine. For instance, testing them on road having temporary or unfinished construction area, four-way intersecting ways, wrong-way passage, and even police officers providing guidelines which contradict traffic lights and many other complex circumstances (Margulis and Goulding 2017, p.500).

            Unlike human driving tests which entail an assessment of a driver’s judgment and decision-making which emanates from factors like, instincts, reflex, or self-preservation, self-driving cars tests are expected to be comparably more rigorous. Every action which the autonomous take comes from a list of choices and the public should be able to confirm that such decisions are safe. The self-driving cars work on the calculation of probabilities which are algorithms quickly processes. For instance, it can calculate if a particular shape is a figure of a person and what does the collected sensor data reveal about the motion or direction of the person. If the person is walking towards the road, the possibility of him or her stepping on to the street or being hit by the car. It virtually mimics the working of the human brain (Self- Driving Ubers 2018, p.7).

            Self-driving cars being electronic devices will depend on frequent software updates from the manufacturer. Such updates should pass under comprehensive tests before being rolled out to influence the performance of the vehicles. Comparison of decision-making conduct between human beings and driverless vehicles may be imperative especially for varying scenarios. Any self-driving car that produces a better outcome than human drivers should be enlisted for certification as road worthy. Such certification process is akin to drug testing process under which a new medication’s performance is contrasted against prevailing therapies (Benenson, Petti, Fraichard and Parent 2017, p.9).

            Determination of the safety of the driverless is similar to the insurance premiums calculation whereby several factors are considered. Driving insurance premiums for the teenage inexperienced driver will be higher as compared to premiums charged on the mature experienced driver. Likewise, the safety rating for driverless cars will be graded low for a long period until they exhibit a high level of guaranteed safety ranking. Statistics is imperative in illustrating the reliability and safety of vehicles. For example, statistic gathered over a long period of time show that young men drivers get involved in more accidents than young women. Just like insurance companies are able to calculate risk probability of individuals and come up with appropriate payable premiums, the road usage regulator should rely heavily on generating a consistent framework for classifying risk and rating risk likelihood of driverless vehicles (Anon 2018, p.23).

            Expectation regarding the involvement of autonomous machines on the roads is that accidents will be reduced greatly if not eliminated since it is thought that 94 % of vehicles accidents happen due to human causes. If the machines are envisioned to eliminate the human accident causes, then target for driverless vehicles should be perfection. It is arguably right that safety tests for the driverless vehicles should be flawless. It is unimaginable what a slight level of neglect for the driverless machine can do the road users. Continuously, the world has embraced various levels of autonomy for cars, but full autonomy as demonstrated in level 5 is yet to be fully accepted due to multiple underlying factors and weaknesses elicited by both hardware and software of the vehicle system (Benenson, Petti, Fraichard and Parent 2017, p.4).

Are driverless cars a game changer for the future?

            The driverless cars have shown the great potential of solving extreme human weaknesses as drivers. They portray great possibility for adoption in future and will be vital in lowering driving costs. Investors in the technology should avoid immature attempt to try them on public roads before comprehensive tests on safety. Acknowledging that billions have been invested in the technology and probably firms are overstretching their budgets to support the technology with hopes of quick returns, haste may not yield substantial outcomes as vehicle accidents are great societal setbacks.

            A key example is a case where a self-driving Uber car hit a woman in Arizona (Self- Driving Ubers 2018, p.7). While the investigation into the true cause of the accident went on for a long time, the reliable outcome was yet to be known. Human beings expect machines and their parts to function as perfect. However, that is not always the case. Several issues may emerge. Sensors may fail and go undetected by the system. Several vehicle parts may malfunction triggering system collapse. For instance, it is common for optical bar reader to fail in supermarkets. Under such cases, the supermarket attendants either resort to manual process or even tries several times for the system to detect (Anon 2018, p.23).

            Allowing a fully autonomous vehicle on the road presently portends potential risks, but it may be achievable in future days. Grave questions people should answer is that what happens when crucial parts of a car fail. Self-driving vehicle technology is a great leap in human technological development and Waymo understands that better. Waymo which is formerly Google's self-driving car venture was initiated in December 2016 and became a subsidiary of Alphabet, Google's mother company. The company’s mission is to make travel easy for every person globally and creating a safer driver for every person (Waymo Safety Report 2018, p.8).

            Waymo believes that the self-driving machines have the capacity to eliminate road carnage and it recognizes that the technology is still in the teenage stage. The vehicle’s acceptance will be achieved after sufficient data is gathered especially regarding an assortment of fleets of self-driving vehicles and rated to be more efficient than human drivers. The technology will gradually become stable and accepted. Stakeholders have to contribute by supporting the technology and through designing supportive roads and backing public trust. The public should divert their attention from their fears of the technology and trust in the car makers to manufacture safe machines (Waymo Safety Report 2018, p.3).

Conclusion

            In brief, huge investments in the research of the driverless vehicles is a great leap in changing how people travel and an attempt to bring safety to roads. Initial tests have confirmed that the machine will improve human safety on the road but the technology is not yet mature to deliver the expected returns on public roads. Safety definition for driverless vehicles is different from the human-based drivers who are rated based on human abilities. A comprehensive model for appraising driverless vehicles should depend on past data gathered, both software and hardware functionality. Safety ranking for machines should be maintained at flawless driving requirement as any slight allowance of errors for self-driving machines may cause a tragic menace to the society. Self-driving vehicles will only earn the opportunity to be driven on public roads after demonstrating the effectiveness and proficiency on the roads. Irrespective of millions and hundreds of miles of testing, if the vehicles cannot show safety then they should not be allowed on the roads. Human safety is too crucial to be left to the hands of teenage machine drivers.

References

Anon. (2018). Black box insurance. Available: https://www.confused.com/car-insurance/black-         box. Last accessed 21/03/2018.

Anon. (2018). Self-Driving Ubers. Available: https://www.uber.com/cities/pittsburgh/self- driving-ubers/#. Last accessed 25/03/2018.

Anon. (2018). Waymo Saftey Report. Available: https://storage.googleapis.com/sdc-            prod/v1/safety-report/waymo-safety-report-2017.pdf. Last accessed 25/03/2018.

Benenson, R., Petti, S., Fraichard, T. and Parent, M., 2017. Towards urban driverless             vehicles. International Journal of Vehicle Autonomous Systems, 6(1-2), pp.4-23.

Fainstein, S.S., 2017. Mega‐projects in New York, London and Amsterdam. International      Journal of Urban and Regional Research, 32(4), pp.768-785.

Furda, A. and Vlacic, L., 2015, October. Towards increased road safety: Real-time decision           making for driverless city vehicles. In Systems, Man and Cybernetics, 2009. SMC 2009.        IEEE International     Conference on (pp. 2421-2426). IEEE.

Margulis, C. and Goulding, C., 2017. Waymo vs. Uber May Be the Next Edison vs.             Westinghouse. J. Pat. &Trademark Off. Soc'y, 99, p.500.

Shladover, S.E., 2017. PATH at 20—History and major milestones. IEEE Transactions on        intelligent transportation systems, 8(4), pp.584-592.