In Response to City Lab’s Proposed Design for Signal-less Intersections
Several prominent articles and studies published recently envision autonomous vehicles(AVs) navigating cities without the aid of traffic signals or other traffic control devices¹. Massachusetts Institute of Technology Senseable City Lab’s vision of the future is typical and shares several commonalities with other schemes. Examining these commonalities we find that they all fail to take into account human nature, mechanical failure, and acts of sabotage making these ideas incompatible with human centered environments like cities. The autonomous vehicle industry and governments seeking to navigate these uncertain waters should not pursue MIT’s design for intersections.
Transportation sabotage could result in the total failure of the type of intersection discussed by researchers from MIT. Their design for signal-less intersections uses a central communication system to provide speed and telemetry for vehicles. These centralized control systems present large and tempting targets for hackers so extreme security measures are essential. In the absence of a human failsafe, the transportation system of entire cities would be at risk. Even a small interruption in communication with a single vehicle could demolish the careful choreography needed to move vehicles through these intersections. A superior system, much less vulnerable to sabotage utilizes decentralized peer to peer communications. A swarm of vehicles managed in a decentralized manner and operating through intersections designed with fewer direct conflict points and which demand slower speeds to navigate has several advantages. Such a system is more resilient to attack since there isn’t one single point of failure, and by design minimizes the frequency and severity of collisions in the event something goes wrong. A centralized system going down via sabotage would need a review of the entire system. This process is likely to cause major disruption to a city’s transportation network. In a decentralized system, failure of a single vehicle requires a much less intense investigation and it would not be necessary to shut down other intersections while examining what went wrong.
Mechanical failure will be part of all systems regardless of how advanced technology becomes. Systems that cannot adapt to those failures will risk lives, create uncertainty, and decrease efficiency. Points of failure in the centralized systems discussed prior need not be disrupted due to sabotage, as simple mechanical failures could also have similar effects. A failure of a single communication or relaying improper information to a single vehicle could result in the chaotic unraveling of an intersection and create conditions ripe for horrific head-on collisions. Another mechanical fail point is with the cars themselves. Vehicles can fail in many ways and can owe their misfortune to inherent design flaws, manufacturing defects or poor maintenance. A vehicle malfunctioning would be unlikely to cause the same level of chaos that a failure of the central control server would cause but the frequency of failures would result in a similar level of disruption.
Implementing safer roadway designs today and then optimizing them for AVs as they become more prevalent is the obvious choice. A road made safe through design will become safer in a driverless future and will help insulate driverless cars from collisions resulting from both mechanical failure and sabotage. Roadways and intersections that promote consistent speeds, and avoid putting cars in direct opposition to each other will always be safer than traditional perpendicular intersections. Modern roundabouts according to the Federal Highway Administration typically achieve a 90% reduction in fatal collisions and a 37% reduction in collisions overall.² These dramatic reductions are attributed to better management of speed and from vehicles traveling in a single direction. Designing roadways in our urban cores that are safer and provide a more consistent experience will cut risks and allow for more human utilization of the scarce urban landscape. Signal-less traditional intersections would have the opposite effect and create cities that entrap its residents in a frenetic grid of fast moving automatons.
The third aspect that shows the shortcomings of signal-less but otherwise traditional intersections is their reliance on humans behaving like machines. Human drivers will not disappear overnight and pedestrians and cyclists will always be part of the urban life. These human players are not compatible with chaotic intersections controlled by a central system and even MIT recognizes that special considerations must take into account the human factor. Its proposals brush off the consequences of normal human behavior and fail to acknowledge how the human element would ruin the efficiency of the systems they propose. Humans are at times irrational, panicked, angry, bored, tired or otherwise in states that make behavioral prediction impossible. They may also suffer from incompetence or may behave in a manner that seeks to take advantage of AVs to hasten their travel time. Roads that by design incentivize competition by giving a time advantage to drivers that constantly change lanes and cut off other users do not support the public’s best interest and will have a devastating effect on efficiency in a world transitioning to autonomous vehicle technology.
While we have yet to write the definitive guide on human and AI(Artificial Intelligence) interact-actions, it is a strong hypothesis that human drivers will treat autonomous vehicles with a more cavalier attitude and will have less regret from taking advantage of a robot versus another human. What we know is that an autonomous vehicle will try to avoid collisions and will detect irregular driving around them, giving way to aggressive human competitors. We need to move towards roadway design that discourages selfish anti-social driving and create safer cities now. If done right these designs will lead to more effective cities in the future as autonomous vehicles become commonplace.
¹http://bgr.com/2016/03/18/mit-smart-intersection-traffic-lights/ http://www.citylab.com/tech/2012/03/what-intersections-would-look-world-driverless-cars/1377/ http://www.cs.utexas.edu/~aim/
