Supporting Mobility Planning in Small Cities and Communities with Low-power, Machine Learning Based Sensing
Cities and communities around the country have very limited resources to invest in projects that go beyond supporting their traditional operational needs. Unfortunately, it is not uncommon to encounter cases where these limited resources are employed in large "smart city" projects that have little or no impact on the social good of the community. On the other hand, the work in this article considers specific needs and financial limitations of suburban or rural communities in the mobility sector. This article presents the work in the design, test and performance evaluation of a low-cost, machine learning based, solution for distributed monitoring of the use of public transit in small communities. The proposed solution utilizes a low-resolution infrared array, that provides non-personally identifiable information coupled with a low-power deep learning module. We also compare the machine-based solution to a traditional algorithmic approach. We believe the solution allows small communities, with limited resources, to collect near real-time data that can be employed for evidence-based decisions of coverage and frequency in public transit. The successful collection of this kind of mobility data is vital for small communities as this is needed to support grant applications that are increasingly commonly used to improve public transit or other mobility modes.
Aráuz, Julio, "Supporting Mobility Planning in Small Cities and Communities with Low-power, Machine Learning Based Sensing" (2021). J. Warren McClure School of Emerging Communication Technologies Open Access Publications. 1.