Real-Time Road Quality Assessment Using Smartphones and Cloud Lambda Architecture

Abstract

In this paper, the authors are proposing a computer system built in the cloud-computing fashion for the collection of data from smartphones to achieve a crowdsensing-based quality assessment of roads, as well as detection, identification, and assessment of road artefacts (potholes, speed bumps, etc). The proposed information processing methods were based on Lambda architecture and the integration of different types of crowdsourced data to finally calculate the singular value of the human-readable road quality. Finally, the prototype implementation of the system for gathering, processing, integration, and processing the road quality data is also presented.

Crowdsourced Driving Comfort Monitoring

“…As for the detection of road artefacts, which is used in the SRQI calculation, the F-THRESH algorithm was used [16]. F-THRESH has an accuracy of over 94% and a low (less than 2%) rate of false positives.…”

“…The second problem is how to calculate aggregated data from the new drives over the same road fragment -in the previous approach [16], the usage of weighted average was postulated, where "fresh" data was more important than older data, as the road may deteriorate slowly over time before the new user of the proposed computing system will appear, and new data should be always treated as more important than old data.…”

“…As for the road quality scoring, usually the IRI (International Roughness Index) is used [13], or its electronic counterpart, eIRI [14,15]. The IRI is not perfectly understandable for the average road user, so alternative quality indicators have been also proposed [16]. In comparison to the already provided research, this paper concentrates on human-understandable road quality indicators calculation and datapoint aggregation by the usage of polygons and distance windows.…”

Usage of Iot Edge Approach for Road Quality Analysis

“…The concept of online processing system is problematic in one situation, namely when considering the amount of data being sent over the network to the processing system. For such systems, the frequency of data acquisition ranges from 10 Hz to 100 Hz, where each data packet includes information on acceleration, orientation, time and location (Badurowicz & Cieplak, 2019), so that the final data packet can reach 50 -100 bytes. In combination with high-frequency systems it creates a data stream of at least 1 KB per second, where the overhead of a protocol such as MQTT is not so high (Pérez et al, 2021).…”

“…However, everything changes when low-power devices are included. The authors (Badurowicz & Cieplak, 2019) previously discussed the concept of building a specialized low-power, low-cost device that could be attached to cars for crowdsourcing data analysis, in the form of devices built on ARM architecture. Such a concept has also been discussed by other authors (Loprencipe et al, 2021).…”