The researchers had previously used the same sensor and software setup in trials of autonomous cars and golf carts, so the new trial completes the demonstration of a comprehensive autonomous mobility system. A user with limited mobility could, in principle, use a scooter to go down the corridor and through the lobby of an apartment building, take a golf cart in the building's parking lot, and pick up an autonomous car on the street.
The new trial establishes that the researchers' control algorithms work both indoors and outdoors. "We were testing them in tighter spaces," says Scott Pendleton, a graduate student in mechanical engineering at the National University of Singapore (NUS) and a researcher on SMART. "One of the spaces we tested was MIT's 'Infinity Corridor,' which is a very difficult localization problem, being a long corridor without many distinguishing features. Anyone can get lost along the corridor. But our algorithms proved to work very well in this new environment."
The researchers' system includes several layers of software: low-level control algorithms that allow a vehicle to respond immediately to changes in its environment, such as a pedestrian crossing its path; route planning algorithms; localization algorithms that the vehicle uses to determine its location on a map; map-building algorithms that it uses to build the map in the first place; a scheduling algorithm that allocates fleet resources; and an online booking system that allows users to schedule rides.

Uniformity:
Using the same control algorithms for all types of vehicles—scooters, golf carts, and city cars—has several advantages. One is that it makes it much easier to perform reliable analyses of the overall system performance.
“If you have a uniform system where all the algorithms are the same, the complexity is much lower than if you have a heterogeneous system where each vehicle does something different,” says Daniela Rus, Andrew and Erna Viterbi Professor of Electrical and Computer Engineering at MIT, and one of the project leaders. “That’s useful for verifying that this multi-layered complexity is correct.”
Furthermore, with software uniformity, information acquired by one vehicle can be easily transferred to another. Before the scooter was sent to MIT, for example, it was tested in Singapore, where it used maps created by the autonomous golf cart.
Similarly, notes Marcelo Ang, an associate professor of mechanical engineering at NUS who co-leads the project with Rus, in their ongoing work the researchers are equipping their vehicles with machine learning systems so that interactions with the environment will improve the performance of their navigation and control algorithms. “Once you have a better driver, you can easily transfer that skill to another vehicle,” Ang says. “That’s the same across different platforms.”
Finally, the uniformity of the software means the scheduling algorithm has more flexibility in its allocation of system resources. If an autonomous golf cart isn’t available to take a user through a public park, a scooter could; if a city car isn’t available for a short trip on back roads, a golf cart can.

Shifting Perceptions:
The scooter trial at MIT also demonstrated how easily researchers could deploy their modular hardware and software system in a new context. “It’s remarkable to me because it’s a project the team completed in about two months,” says Rus. The MIT Open House was in late April 2016, and “the scooter didn’t exist in early February,” says Rus.
The researchers described the scooter system design and trial results in a paper they presented at the IEEE International Conference on Intelligent Transportation Systems. Joining Rus, Pendleton, and Ang on the paper are You Hong Eng, who leads the SMART autonomous vehicle project, and four other researchers from NUS and SMART.
The paper also reports the results of a brief user survey the researchers conducted during the trial. Before riding the scooter, users were asked how safe they considered autonomous vehicles to be, on a scale of one to five; after testing it, they were asked the same question again. The scooter experience resulted in an average safety score between 3.5 and 4.6.
###
Written by Larry Hardesty, MIT News Office