Recent research reveals that human supervisors possess the ability to mitigate obstacles when it comes to the implementation of autonomous vehicles.
Reliable human drivers log a hundred million miles for each fatal car crash, outperforming autonomous vehicles in decision-making and complex perception. Embracing autonomous vehicles offers benefits like clearing congestion, enhancing mobility, and improving driving efficiency to combat climate change.
Researchers at the Massachusetts Institute of Technology (MIT) envision overcoming the current shortcomings of autonomous vehicles to ensure their safe deployment. Recognizing safety and reliability as the primary barriers, the researchers address the bottlenecks. A potential approach is a hybrid system where autonomous vehicles handle easy scenarios independently, such as highway cruising, while complex maneuvers are transferred to remote human operators.
Humans and cars: finding a balance that’s just right
The researchers addressed autonomous vehicles’ struggle with merging. They focused on on-ramp merging, where vehicles must accelerate or slow down to avoid collisions. To ensure a safe merge, remote human supervisors could temporarily take control. Evaluating system efficiency and safety, the team set a maximum time for each supervisor to handle a merge. They explored the feasibility of a few supervisors managing a larger fleet of autonomous vehicles while maintaining safety coverage. As autonomous vehicles increase, coordinating with each other reduces the need for remote supervisors. Coordinated vehicles can adjust speed, avoiding risky merging situations. Using queuing theory, the team analyzed supervisor requirements for handling merge conflicts and derived a theorem on the impact of cooperative autonomous vehicles in enhancing reliability during merging. In a scenario with 30% cooperative autonomous vehicles, a human-to-vehicle ratio of 1:47 covered 99.9999% of merges. However, coverage fell below 99% without vehicle cooperation, which is deemed unacceptable.
Cruising toward the future
The team chose merging as the focus due to its difficulty for autonomous vehicles and its association with highway driving. Highways account for a significant portion of the total miles traveled in the US, and automating highway driving can reclaim a significant portion of driving time for individuals. Handling the start and end of highway driving and managing passenger inattentiveness or sleep would be crucial. Remote human supervisors guiding vehicles at critical moments could eliminate the need for passenger intervention. Lane changes and overtaking slower vehicles are additional highway challenges.
