One of the obstacles to deploying autonomous vehicles is the high cost of some components, but researchers are experimenting with self-driving golf carts that use minimal and relatively cheap gear.
The scientists from MIT and Singaporean universities deployed two modified Yamaha electric golf carts in the Southeast Asian city-state. They envision the self-driving vehicles being used in a shared transportation system, as rental bicycles are used in many cities.
As seen in a YouTube video, the carts transported 500 people along winding paths in public gardens in Singapore while autonomously navigating and watching for obstacles such as pedestrians and animals.
The carts picked up people at 10 stations in the gardens. They traveled at a maximum speed of only 24 kilometers per hour, so that the computers had time to process all the obstacles.
The only problem was when a slow-moving monitor lizard crossed a cart's path, causing it to stop and wait, according to MIT. Nearly all the passengers said they would ride in the golf carts again.
The researchers, part of the Singapore-MIT Alliance for Research and Technology (SMART) collaboration, focused on using less gear than self-driving vehicles while relying on computation-efficient algorithms.
An algorithm known as the Dynamic Virtual Bumper handles the navigation and obstacle avoidance, setting the cart's path. It is a computational "tube zone" with its center line as the path, according to a paper on the research that will be presented at the 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems in Hamburg later this month.
The size of the virtual tube is a function of the cart's speed and position, and when an obstacle is detected, the tube is redrawn to exclude it.
In addition to a webcam, each cart is equipped with four single-beam LIDAR (light detection and ranging) sensors from German maker Sick that have a field of view of about 270 degrees.
"We do not use the 3D scanners that have a very high price point and produce a panoramic image," Daniela Rus, director of MIT's Computer Science and Artificial Intelligence Laboratory and a coauthor of the paper, said via email.
Two of the sensors were mounted in the cart's front and used for determining its position and obstacle detection. The other two were cheaper, shorter-range sensors and were mounted on the back corners of the cart to scan for obstacles behind and on either side of it.
The cost of the sensors was still high -- in the order of US$30,000 -- but that's less than solutions used in more sophisticated robotic vehicles. Google has used $80,000 Velodyne LIDARs on its earlier self-driving cars.
"Prices for sensors keep coming down, and capabilities are increasing," Emilio Frazzoli, a professor of aeronautics and astronautics at MIT who also coauthored the paper, said via email.
The researchers plan to improve the booking system for the carts, and develop a method that would let the vehicles communicate their intentions to nearby pedestrians.
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.