For this past Friday’s Jones Seminar, Thayer Professor Reza Olfati-Saber gave a lecture on the possibility of automated transportation technology, particularly self-driving cars, and the obstacles facing such technology.
Olfati-Saber began his lecture with an outline of the system underlying automated transport. Interestingly, he explained that he was inspired to develop cyber-physical transportations systems by natural systems, such as schools of fish and flocks of birds. In broad terms, an ideal cyber-physical system is a network of agents that all employ embedded control, sensory, and communication devices to stay synchronized. The fundamental objective of such a system is autonomy and self-organization with few supervisors or none at all. Olfati-Saber and his colleagues have achieved this by analyzing the collective behavior of the network’s agents and then designing specific rules and protocols to be followed by all agents.
“Intelligent” transportation systems could be extremely useful in reducing the possibility of human error, notably those caused by intoxication, cell phones, and medical emergencies. Olfati-Saber raised the intriguing possibility of a partially-automated car’s analyzing the physiological state of its human driver. Upon detecting an abnormal state, the car’s automated driving system would temporarily assume control to safely remove the impaired driver from the road. Furthermore, with advanced sensor systems, automated driving systems would be more “aware” than drivers of potential obstacles like hidden pedestrians.
Olfati-Saber cited several past examples of autonomous transportation experiments,
including a highly successful Google car, to prove that such cyber-physical systems are indeed feasible. The most successful vehicles were those that utilized built-in sensory and computing devices to complete autonomous tasks, such as lane-changing, braking, and collision-avoidance. Despite the fascinating aspects of single-vehicle autonomous transportation, Olfati-Saber’s current research is focused on “flock” theory. Flock theory describes how a multi-vehicle “fleet” cooperates to avoid collisions and to perform specific tasks like passing.
The term “fleet” refers to any group of vehicles traveling in the same direction who share a general goal of collision avoidance. Olfati-Saber is particularly interested in highway situations because of mechanisms involved in self-organization, which are based mainly on the simple rule that each car will maintain a nearly constant distance between itself and its immediate neighbors. Using this rule, along with more complicated observations, Olfati-Saber has developed a flocking algorithm that holds up very well under computer simulation.
He displayed a few examples of simulated cars successfully performing such tasks as safe one-lane driving, high-speed passing, and even urban driving. The key to successfully completing such tasks is a comprehensive set of logical programmed rules for specific situations. For example, if a car calculates that its front neighbor is driving too slowly and that there is an open lane safely available, the car will “decide” to change lanes.
Olfati-Saber concluded his lecture by mentioning a few of the challenges facing automated cyber-physical systems. For one, the potential use of onboard sensors to analyze and store physiological information about drivers raises serious privacy issues. Secondly, developers must always take into account the threat of hackers disrupting the digital components of any automated system.
Nonetheless, given the past successes of automated transportation and the rapid progress of current research, Olfati-Saber expressed his optimism that such systems would become a reality sooner rather than later.