On the Road to Freedom

For those with disabilities or injuries that prevent them from walking, a wheelchair can restore their independence and allow them to participate fully in normal daily activities. Wheelchairs come in various forms, from manual wheelchairs, which are propelled by the user, to electric wheelchairs, which can be operated using a controller. However, it is important to note that while some individuals benefit tremendously from a wheelchair, others do not have the physical capacity to operate one, either manually or by using a controller.

An undergraduate at the Karunya Institute of Technology and Sciences in India named Kabilan KB was inspired by his cousin, who has a mobility disorder, to develop a new type of wheelchair that can help anyone — even those that cannot operate the controls of a motorized wheelchair — to regain their independence. KB also noted that where he lives, many people cannot afford an expensive wheelchair, so he was determined to build his solution in the most cost-effective way possible.

The solution proposed by KB is an autonomous wheelchair that uses relatively low-cost components to drive the rider to a predetermined location. It was noted that the destination could be chosen by typing a number on a numeric pad, for example. But in theory, the destination could also be spoken, or supplied in some other form, as the abilities of the individual using the wheelchair allow.

KB started by ordering the most basic motorized wheelchair that he could find, then it was outfitted with an NVIDIA Jetson Nano Developer Kit to provide the processing horsepower for the deep learning workload that would be required for autonomous operation. Data about the surrounding environment was captured by the combination of a SLAMTEC RPlidar laser range scanner, and an Intel RealSense depth camera. An Arduino development board was also interfaced with the Jetson Nano to handle communication with a motor driver, which triggers movements of the wheelchair’s motors. The hardware is all powered by a rechargeable battery pack.

After the user of the wheelchair selects a desired destination, the Jetson Nano captures measurements from both the lidar sensor and the depth camera. The camera data is fed into a deep learning algorithm that detects any nearby objects, while the point cloud obtained from the lidar sensor is used to identify obstacles and generate a 2D map of the surroundings. With a knowledge of nearby objects of interest, as well as obstacles to avoid, the Jetson prepares a collision-free navigation plan, which it then carries out by sending control signals to the motor driver via the Arduino.

By processing the sensor data in real-time, the wheelchair is able to smoothly and safely carry its rider to their destination without the need for any input, aside from a choice of destinations to go to. This is quite an impressive feat considering the low cost of the components involved in the build. A system such as this could really change a lot of people’s lives for the better one day.

While creating an autonomous wheelchair was his focus, KB notes that similar techniques could be leveraged in other autonomous vehicles, surveillance robots, or even drones that map out unknown regions. KB has provided plenty of details in his blog post for anyone else that may be interested in building an inexpensive autonomous vehicle of their own.