SmartRoad: A Crowd-Sourced Traffic Regulator Detection and Identification System

Hu, Shaohan; Su, Lu; Liu, Hengchang; Wang, Hongyan; Abdelzaher, Tarek F.

Permalink

https://hdl.handle.net/2142/43856 Copy

Description

Title

SmartRoad: A Crowd-Sourced Traffic Regulator Detection and Identification System

Author(s)

• Hu, Shaohan
• Su, Lu
• Liu, Hengchang
• Wang, Hongyan
• Abdelzaher, Tarek F.

Issue Date

2013-04-18

Keyword(s)

• crowd source
• road sensing
• traffic regulator

Abstract

In this paper we present SmartRoad, a crowd-sourced sensing system that detects and identifies traffic regulators, traffic lights and stop signs in particular. As an alternative to expensive road surveys, SmartRoad works on participatory sensing data collected from GPS sensors from invehicle smartphones. The resulting traffic regulator information can be used for many assisted-driving or navigation systems. In order to achieve accurate detection and identification, SmartRoad addresses various challenges in participatory sensing scenarios, including data unreliability/sparsity, energy constraints, and the general lack of ground truth information. SmartRoad automatically adapts to different application requirements by intelligently choosing the most appropriate information representation and transmission schemes; it also dynamically evolves its core detection and identification engines to effectively take advantage of any external ground truth information or opportunity. With these two characteristics, SmartRoad consistently delivers outstanding performance for its road sensing tasks. We implement SmartRoad on a vehicular smartphone testbed, and deploy on 35 external volunteer users’ vehicles for two months. Experiment results show that SmartRoad can robustly, effectively and efficiently carry out its detection and identification tasks without consuming excessive communication energy/bandwidth or requiring too much ground truth information.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/43856

Owning Collections