Home 🎓 NavAware: Data-centric design of a user-aware navigation agent for blind mobility
Post
Cancel

🎓 NavAware: Data-centric design of a user-aware navigation agent for blind mobility

With the rise of digital technologies including artificial intelligence, machine learning, and sensor technologies, all the products we use every day are increasingly becoming ‘intelligent’. Machines, objects, and tools are slowly evolving into robots, objects-with-intent (Rozendaal et al., 2019), and agents. At the same time, their capabilities are also increasing. These capabilities come from their ability to process large amounts of data. In this digital era, how can designers create user-centric solutions by leveraging the strengths of these digitally-enabled products at an early stage in the design process?

This thesis explores this topic in the context of pedestrian mobility. More specifically, outdoor mobility for people with vision impairments. When people with vision impairments travel, they must process an extensive amount of spatial information without relying on their vision. How can an agent partner with these users to make this process less demanding and thus, make it easier for people with vision impairments to travel independently?

Combining these two visions, this thesis focuses on uncovering the problem areas, information needs, and desires of PVIs traveling outdoors by leveraging behavioral data. Two main user studies, both involving different types of behavioral data have been explored.

Based on the exploratory research, in which physiological data was collected, two behavioral states, ‘following’ and ‘reorienting’, were identified. These two states also reflected the PVI’s mental state, moments of ease, and moments of uncertainty. This inspired the initial idea for a user-aware navigation agent, which could announce different types of data depending on whether the user was ‘following’ a route or ‘reorienting’ on the route.

In the evaluative research, the agent’s capability to detect the two states was tested by training a machine learning algorithm. In addition, a second evaluative user study was conducted to test the hypothesis of the information needs in the ‘following’ and ‘reorienting‘ states. The study setup was designed to generate a new type of behavioral data, which consists of the location users requested additional information. Thus, the user needs were quantified.

This kind of behavioral data; one that embeds rich information about the user needs, becomes the building block of future prototypes of the user-aware navigation agent. Those iterations will then produce more data that embed further insights. This positive feedback cycle will be key to keeping the PVIs constantly in the product development loop and hopefully, result in a navigation agent that allows PVIs to easily travel safely and independently.

Master Thesis: TU Delft repository

This post is licensed under CC BY 4.0 by the author.