A Model to Predict Children’s Reaction Time at Signalized Intersections

: Traffic accident statistics in urban areas, both locally in Croatia and at the European level, identify children as a group of vulnerable road users. The analysis of the parameters that influence the interaction of child pedestrians and other road users requires special attention. This paper presents the results of research about the reaction time of children, measured both in laboratory conditions, via a computer reaction time test, and in actual traffic conditions. The results of the reaction time in the situation of expected stimuli (computer test) of children aged 6 to 10 years were compared with the results of the reaction time of adult traffic participants, drivers, who also took part in the computer test. In actual traffic conditions, the reaction time of children aged 4 to 16 years at the signalized intersection was measured. The model for predicting the reaction time of children in real traffic conditions was created using a neural network. Parameters influencing children’s reaction time in real traffic conditions have been identified by applying both statistical analysis and the developed neural network model. The case study was conducted at selected signalized intersections in the city of Osijek, Croatia.


Summary
In this research the reaction time of children was measured under controlled conditions using computer software and in actual traffic conditions at a signalized intersection. The aim of this research is to increase the safety of specific segments of urban traffic network in close vicinity of schools and kindergartens. The case study was made for a selected intersection with traffic lights in the city of Osijek.
The methodology followed to develop the whole research is based on two steps: first of all, a laboratory experiment was carried out with a computer test, where both a target (children) and a control (adult) group had to react to an external displayed stimulus; the second step is the measure the reaction time of the same participants in real traffic conditions. The first step allowed to identify some parameters influencing the reaction time under controlled conditions, and put the basis for the analysis and selection of the most influential parameters to be set as inputs for the neural network model.
Data collection under laboratory conditions was made within the framework of the project of the Croatian Science Foundation Problems in the Behavior of School-aged Children: The Role of Executive Functions, Individual, Family and Genetic Factors-ECLAT, project leader Asoc.prof. Silvije Ručević, HRZZ-IP-2016-06-3917. This research did not engage other project resources except for the availability of the target group for the research.
In the working phase of data collection, an article with descriptive statistics was published: Ištoka Otković, Irena; Ručević, Silvija; Borovac, Tijana; Marschhauser, Max; Jeremić, Kristina. Analysis of the results of traffic participants' time of reaction research for prevention of traffic insecurity // 6th International Scientific Symposium Economy of Eastern Croatia -Vision and Growth / Mašek Tonković, Anka (ur.). Osijek: Studio HS internet d.o.o., Osijek, 2017. pp.826-835 http://www.efos.unios.hr/red/en/proceedings/ The final database expanded by 20% of reaction time measurements was used in this article as well as statistical analyzes not made in the research work phase.

Methods
The methodology followed to develop the whole research is based on two steps: first of all, a laboratory experiment was carried out with a computer test, where both a target (children) and a control (adult) group had to react to an external displayed stimulus; the second step is the measure the reaction time of the same participants in real traffic conditions.
The target group consists of kindergarten and elementary school children, and the control group consists of drivers from the city of Osijek. The database contains measurement results of 448 target group respondents (four groups of 112 measurements each) and 112 control group examinees. The reaction time was measured in controlled conditions, using Human benchmark Reaction time test (http://www.humanbenchmark.com/tests/reactiontime) (computer on-line test) on all participants, and additional data were gathered through a survey filled in both by the target and the control group.
The target group was divided into four subgroups-preschoolers (children from 6 to 7 years old), first class (children from 7 to 8 years old), second class (children from 8 to 9 years old) and third class (children from 9 to 10 years old) children. In this phase of the research, the working hypothesis was that the reaction times of each target group was longer than the reaction times of the control group, even in the conditions of expected stimuli. To each respondent it was explained what would be measured, how they should respond (press the mouse button when a red color appears on the screen), and each of them had the opportunity to try the test first, without recording the results. The reaction time was measured in milliseconds and the mean reaction time of 5 consecutive measurements for each respondent was registered.
Regarding ethical questions of research involving children, individual permissions for interviewing and measuring different indicators of executive behavior of each child were sign by the parents and only those children whose parents signed the consent were examined. Figure 1 is the approval of the Ethics Committee of the Josip Juraj Strossmayer University of Osijek, Faculty of Humanities and Social Sciences, to collect data and measure different indicators of children's executive behavior (reference number: 1/4/2017i). The primary objective of the second phase of research, the in-situ research, was to collect data about the behavior and reaction time of children in actual traffic conditions, when children are influenced by the usual distractors in a familiar environment. A database of measured reaction time of children aged 5 to 16 years at a selected signalized intersection set on urban arterial road was created. The observed intersection is located near two primary schools and kindergarten. The signalized intersection is traditionally considered as a pedestrian-friendly traffic solution because it does not require a detailed assessment of the traffic situation and was selected for the first phase of the study of traffic behavior of children. No license was required for video recording of traffic without personally identifying traffic users.

Database1
A database of measurement results of the reaction time of the target group (child pedestrians) and a control group (adult drivers) under controlled conditions was created. Table 1 shows the reaction time measured by a computer test, as described in the previous section.

Database2
Based on previous experience and technical judgement of the situation, seven independent variables affecting the reaction time of children in traffic were selected: • age group • gender • children with special needs (motoric disabilities, low vision and blindness, wheelchair mobility, etc.) • movement in a group-the number of children in a group • supervision by adults • mobile-text-messages/internet • mobile-talk, listening music The age group input parameter was formed in such a way that each respondent with a measured reaction time was classified into one of the seven categories, as follows: 1→≤ 5 years old children; 2→6 to 7 years; 3→8 to 9 years; 4→10 to 11 years; 5→12 to 13 years; 6→14 to 15 years; 7→>15-over 15 years old).
Other  Table 2 shows the database measured in actual traffic conditions.  The database of 192 measured reaction times in actual traffic conditions was used to define and train the neural network. Table 3 shows the results of neural network prediction Independent validation of the model was made on a new database (which has not been previously presented to the NN, neither in the training set nor in the test set) consisting of 45 measured data (15 measurements at the first intersection, 30 measurements at a second location) of the reaction time of children at two different signalized intersections. Table 4 shows the results of model validation.