9 Challenges for Total Autonomous DrivingDec 03, 2019 | 1083 views
Autonomous driving is one of the most stimulating challenges to our imagination. What will the future look like on the roads and in cities when vehicles are able to take us from one origin to a destination by itself without human intervention?
In a few years autonomous driving has moved from our imagination to real roads and cities and their discussion has long been beyond the automotive industry, as its impacts will be political, economic and social. But are we already close to the day when a car without a steering wheel and no pedals will take us to our destination?
By 2015, Tesla CEO Elon Musk predicted that their cars would be fully autonomous by 2017. General Motors said in 2018 that it would launch a fleet of cars without a wheel or pedals by 2019. Google, viewed by many as the leader of the industry has committed to launch a driverless taxi service in Phoenix, where it has been testing its cars in late 2018, but after a year only part of the city is covered and only certified users can use it.
While Phoenix's extremely quiet streets are among the easiest to drive in the world, Waymo / Google cars have human drivers behind the wheel, “just in case”… and a few months ago Google had already admitted to taking 30 years until we have truly autonomous cars.
It's a matter of time as advances in Machine Learning, the subfield of Artificial Intelligence, will allow cars to learn to drive on their own by taking advantage of the huge amounts of data that is collected from drivers. By September 2019 there were already 135 cities in the world with public driving initiatives in delimited zones.
Below, I review nine challenges that must be overcome before we have a standalone car without a driver behind the wheel, or a car without a wheel or pedals:
1. Construction and maintenance of driving maps
Aiming at Google's example, the driving that it develops is based on the simultaneous combination of the car's perception system, which with its sensors detects real-time obstacles along the way, with detailed and pre-built maps making a manual recognition of the car. intended route. The idea of building a map in advance serves to minimize the processing power required.
As you might imagine, building these detailed maps involves a mammoth effort that increases with constant road changes. However, the more autonomous cars there are, the easier it becomes to update the map they make themselves by updating a central system that will communicate with each other.
2. Complex Social Interactions
The second challenge has to do with the complex social interactions on which driving is based and which are still difficult for a robot / machine to implement. In fact, driving is a social process that often involves interactions with other drivers, cyclists and pedestrians.
In many situations we make decisions based on common sense, which is something that machines still cannot implement. Imagine, for example, that we approach a crosswalk and have a person looking at your phone. A human driver can deduce whether the pedestrian is stationary or distractedly will cross the road / street even without a green light. Or consider all the various situations involving subtle eye contact and communication with a pedestrian or other driver or, for example, a policeman waving and directing traffic in an accident. Easy for us but still hard for a robot.
3. Adverse weather conditions
The third challenge is driving under adverse weather conditions such as heavy rain, snow or fog with poor visibility. Like our eyes, sensors do not work well under those conditions either.
4. Legislation and regulation
The fourth challenge is not technical - it is political - and has to do with legislation and regulation including insurance models. Before VAs (autonomous vehicles) popularize the streets and roads, regulators will have to approve them, and one thing they will ask is: How safe are these things anyway? And the tricky part is, we probably won't know.
In the United States there is a fatal accident for every 100 million miles traveled. Ideally VAs have to be at least at this level. But it is highly unlikely that we will be able to prove it in the near future. Between 2009 and 2017 Google only drove 1.3 million miles. It will take decades to drive hundreds and hundreds of millions of miles to prove the level of safety. The only alternative is to use models and simulations, but it will be a rather long certification process. Therefore, the emergence of legislation and regulatory aspects framing driving and their responsibilities. New insurance models are also needed to insure a trip over the vehicle we use (as with air travel, what is safe is the service and not the object) and who will be imputed, the driver or the manufacturer, if any. accident.
The fifth challenge is cybersecurity. In a vehicle transformed into a computer, its integrity must be guaranteed. And there is a long way to go before we can affirm it.
6. Electrification and charging network
The sixth challenge is related to vehicle electrification and charging network. Why when we talk about VAs we always imagine them electrified? Apart from any environmental concerns, with the idea that VAs will be used in a more sustained mode of sharing as a service rather than owning it, the conventional 12V battery architecture in fossil fuel vehicles is insufficient to provide power to all on-board computer systems with sensors, actuators and communication equipment.
On the other hand, the use in the sharing model will take over 100,000 km / year the distance traveled by each car which further highlights the cost of maintenance, which is expected to be much lower in an electric car. For this reason, the theme of the charging network and also the spaces to be used on the street or in indoor public spaces must also be part of a city's mobility policy.
7. Urban Mobility Planning
The seventh challenge is integrated urban mobility planning. While a positive overall impact on cities is expected in the end by using shared VAs, the need for parking space will be halved, providing an opportunity to redesign cities. But without proper and comprehensive planning, city mobility could have a huge negative impact.
Early estimates in a Boston study show that the level of congestion will increase in the city, mainly because VAs will be chosen as a substitute for short-haul public transport, increasing travel time by 5.5%. In the suburbs, mobility replacing the private car will have a positive impact with a shorter travel time of 12.1%.
Probably the best contribution we can find for a more fluid mobility comes from another quadrant: a change in the organization of work in companies and institutions that allows remotization of functions and, in addition, a digitization and dematerialization of processes in institutions, facilitating access through Internet over physical access will be the largest contributor to minimizing displacement.
The very organizational model of society, with large cities growing and small cities becoming depopulated, will lead to a bottleneck in mobility if nothing is done to counteract this trend. For all these reasons, the only guarantee of success will be the existence of an integrated mobility management platform in cities, which requires collaboration at various levels and with different stakeholders.
Urban mobility should be viewed holistically as an integrated multi-channel system: bicycle, scooter, bus, train, subway, single car, shared car, taxi etc.
8. Communications Network
The eighth challenge is the support network for inter-vehicle and vehicle-infrastructure communications. VAs need to know what is going on around them, determine their exact position on the road or street, and decide what behavior to take in a given situation. That is why they are highly dependent on software to bridge the physical sensors and the mechanical actuation of the vehicle, namely steering wheel, brake and accelerator.
It is impossible to think of a scenario where VA acts only on the information it collects from its sensors. Communication with other vehicles, which cross in the opposite direction and which may alert a certain situation, communication with the road infrastructure, with traffic management systems that warn of certain dangers, such as at a junction where a vehicle is approaching. if without priority but with no chance of stopping on time, these are examples of how critical it is to have a real-time communications system available linking vehicles and road infrastructure with each other, which can be supplemented with external information from other sources.
It is believed that 5G can play this role if it meets its intended objectives with low latency (near real time) and high device density capability.
9. Manage new technology and legacy
The ninth and last but not least challenge is managing double coexistence. The biggest problem in introducing a new technology is always living with the legacy technology. Everything would be simpler if overnight all manual driving vehicles became autonomous vehicles. The problem is that it will be many years before manual driving disappears from cities and roads, making it much more complex.
When we talk about VAs, we first think of automobiles, but we cannot ignore all other motorized mobility vehicles, whether trucks, goods or heavy, as well as mass transit and those using air and air. which will also lead to a transformation of cities and our way of life.
Proof of concept and experimentation will be vital in defining a mobility policy. Each city is a case. Think for a moment about cities like Bombay or Cairo with their chaotic traffic and what should be the strategy of introducing VAs… Also, driving with electric vehicles raises other social questions: what business models are there? Do we still need to get a driver's license? What will happen to high schools? What will happen to electrification and the thousands of machine shops? What will we do inside Autonomous Vehicles over long distances?
There are many questions for which the answers are still scarce or incomplete. For this reason, it is necessary to thoroughly discuss the issue so that, contrary to the 20th century where we cannot predict the future effects of the car on cities and we do not design them for that purpose, we will be able to anticipate the future of cities in Autonomous Driving era.