A Low-cost IoT Architecture to support Urban Mobility for Visually Impaired People

Abstract: People with visual impairments struggle with urban mobility and independent travel, opening up opportunities for technological advances to improve their quality of life. The Internet of Things (IoT) plays an essential role in bringing improvements and accessibility for visually impaired people. Although alternatives aimed to use IoT in urban mobility, those solutions are still in the initial stages and do not supports urban mobility for people with visual impairment. This paper proposed and evaluated a low-cost IoT architecture that uses Single-Border Computers (SBCs) to support urban mobility. A performance evaluation showcased that our low-cost architecture handles bus trace workload and is suitable for supporting impaired people to get information concerning bus location on Smart Cities scenarios.

• The paper presents a low-cost IoT architecture using SBCs and bus sensors for real-time tracking to enhance urban mobility for visually impaired people.
• Performance evaluation showed message size impacts latency and message count impacts CPU use; QoS Level 2 significantly increases gateway CPU consumption.
• The low-cost architecture is viable but QoS adds significant overhead, suggesting future research into alternative data transmission methods, gateways, and AI for predictions.

The paper introduces a low-cost IoT architecture designed to enhance urban mobility for visually impaired individuals. The proposed system leverages single-board computers (SBCs) to provide edge intelligence, facilitating real-time bus tracking and notification capabilities within smart city environments. The architecture aims to address the challenges faced by visually impaired people in navigating urban spaces, where a significant portion of the population relies on public transportation.

The architecture consists of three primary components:

• A GPS (Global Positioning System) NEO-6M location sensor,
• ESP32 (ESP-WROOM-32) LoRa module,
• Power supply unit installed on public transport buses.

These components transmit location data to an IoT gateway, which is based on a Raspberry Pi3 SBC. The gateway preprocesses this location information and publishes it to a cloud-based IoT Urban Mobility application. This application then notifies users of the proximity of public transport vehicles through audible and mechanical signals via a mobile application.

The paper evaluates the performance of the proposed architecture in handling bus trace workloads. The experimental setup involves a testbed topology where entities are directly connected via Ethernet. The IoT gateway, running Mosquitto on a Raspberry Pi3 with Ubuntu 20.04 LTS, is subjected to a workload using the MQTT tool Bench to measure its request-response capability for message publications.

The evaluation focuses on the impact of message size and the number of messages on latency and CPU utilization, considering both with and without QoS (Quality of Service) levels 0 and 2. The research questions addressed include:

1. The relationship between the latency of MQTT messages and the message size sent by the buses.
2. The correlation between the number of messages sent by the bus and the latency and CPU utilization of low-cost IoT devices.
3. The processing overhead imposed by the addition of QoS on low-cost devices.

The results indicate a strong correlation ($0.999$) between message size and latency, suggesting that latency increases with the size of messages sent by buses. A strong correlation ($0.729$) was also found between the number of messages and the IoT gateway's CPU consumption. However, the correlation between the number of messages and latency was weak and negative ($-0.017$). The increased CPU consumption is attributed to the number of interruptions caused by message arrival.

The evaluation of processing overhead with QoS revealed that the average CPU consumption for processing messages from 1,000 buses without QoS was $10.47\%$, while imposing QoS Level 2 increased CPU consumption to $23.94\%$. This represents an average increase of approximately $2.2$ times with QoS. In contrast, with a smaller number of buses (100), the overhead of applying QoS was not statistically significant.

$$Correlation = \frac{\sum_{i=1}^{n}(x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum_{i=1}^{n}(x_i - \bar{x})^2}\sqrt{\sum_{i=1}^{n}(y_i - \bar{y})^2}}$$

where:

• $x_i$ is the value of the x-variable in the sample i
• $\bar{x}$ is the mean of the values of the x-variable
• $y_i$ is the value of the y-variable in the sample i
• $\bar{y}$ is the mean of the values of the y-variable

The study concludes that the proposed low-cost IoT architecture is capable of handling workloads imposed by embedded sensors on buses, with IoT gateways sufficiently capable of managing incoming messages. However, as the number of buses increases, transmitting messages with QoS leads to higher CPU consumption. Future work involves evaluating alternative data transmission mechanisms such as mobile networks, assessing different IoT gateways, and exploring the application of AI to predict delays and arrival times.

The paper references several works related to IoT applications for visually impaired individuals and urban mobility. These include navigation systems based on BLE (Bluetooth Low Energy), mobile applications for bus tracking using LoRa and GPS (Global Positioning System), and bus detection systems using decision tree and K-NN classifiers. The proposed architecture distinguishes itself by considering GPS (Global Positioning System)-based location for accuracy and enabling route tracking for both disabled and non-disabled users. The authors also acknowledge financial support from the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES).