Evaluating bluetooth and Wi-Fi sensors as a tool for collecting bicycle speed at varying gradients

The wide applications of mobile units with communication technology opens up new possibilities in data collection among road users. Sensors detecting Bluetooth and WiFi units have been successfully applied in collecting travel times from motorized vehicles. The same technology could potentially be applied to bicycle transport. As part of a policy for facilitating sustainable transport modes, more knowledge about bicyclists is needed. This paper presents a study answering two research questions: 1. Will the use of Bluetooth and WiFi sensors provide reliable travel time data from bicyclists? 2. How are bicycle section speeds affected by different gradients? To answer the first research question a two step test was set up to validate the equipment. Each equipment set consisted of one Bluetooth sensor and one WiFi sensor. Two sets of equipment were placed along a bicycle path with a distance of 550 meters between them. There was no adjacent road, ensuring only bicyclists and pedestrians would be registered. Firstly, a controlled test was conducted with known Bluetooth and WiFi units brought by test cyclists. The test took place nighttime to avoid disturbance from other travelers. To validate the measured travel times, actual travel times were also registered manually. The test concluded that travel times based on Bluetooth were more accurate than travel times based on WiFi. However, both sensors provided travel times that were not significantly different from the actual ones. Secondly, an open test was conducted along the same cycle path section during rush hours to find the penetration rate (i.e. the number of registered travel times divided to the total number of passing bicyclists), and to manually register all travel times to see whether the registered travel times were representative of the travel times for all passing bicyclists. The test showed that to reduce the variance in the registered data in order to calculate a valid average travel time, a minimum of 20 observations were needed. The test further showed that more than 90% of the registered travel times were detected by the WiFi unit. The second research question was addressed by selecting 7 road sections with average gradients varying from 0.5% to 9.2%, using section lengths between 500 to 1500 meters. A total of approximately 2400 travel times were registered, giving the following main findings: The average speed varied from 22.4 km/h to 8.6 km/h uphill according to increasing gradient. The variance is reduced by increasing gradient. When cycling downhill the average speed varies between 21.3 km/h and 36.5 km/h, registering the highest speed at 5.4% gradient, showing that bicyclists choose lower speeds on steep downhills. The variance seems unaffected by the gradient. Based on this study, sensors using Bluetooth and WiFi technology can be recommended as tools to collect travel times from bicycle traffic. However, some important prerequisites are needed: the sensor locations have to be carefully chosen, as do appropriate filtering of data to avoid inclusion of pedestrians and motorists in the dataset. (C) 2016 Published by Elsevier B.V.