Indoor Wayfinding for an Electric Wheelchair Based on Wi-Fi Fingerprinting Localization

Date:

Conference proceedings talk at 2020 IEEE/SICE International Symposium on System Integration (SII), Honolulu, Hawaii

This presentation was based on a peer-reviewed paper, which can be found here


Paper Abstract
Currently, society is experiencing a shift in the median age known as population aging. The number of people aged 60 years or over is at its highest, and it’s expected to double by 2050. With age, people frequently experience difficulties in mobility, requiring them to use assistive technology, from wearable devices to mobile devices such as wheelchairs or walkers. Recently, power-assisted versions of these systems have been developed to further enhance the provided assistance, such as electric wheelchairs and power-assisted walkers. In scenarios such as shopping malls or airports, people often need to walk long distances to reach their destination, so these power-assisted systems can help them move without difficulties. However, these systems are still quite expensive, and having one device per user might prove difficult. To overcome this, these facilities could have a reduced number of devices and allow users to share them upon request. However, it’s not desirable to have each user move toward the assistive device due to their limited mobility. Instead, in this research we propose a method to enable the device to navigate toward the user upon request by using onboard sensors and the existing Wi-Fi infrastructure. Specifically, we create a map with the Received Signal Strength Indicator (RSSI) of existing Wi-Fi access points (using a method called Fingerprinting). When a user requests an assistive device, the RSSI values at the user’s position are sent to it, and the device determines the rough position of the user using a KNN algorithm. However, typical Fingerprinting methods are affected by infrastructural changes which alter the profile of RSSI values at each location. Therefore, we propose that the map is constantly updated while the device moves in order to avoid errors due to changes in the infrastructure. Through experiments we confirmed that the device could locate the position where the request originated with an error of 2.612 m, and it was able to navigate towards it.