Semantic trajectory insights for worker safety in dynamic environments

Abstract

Existing literature reveals that unsafe worker movement behaviors are one of the major reasons of construction site fatalities resulting in serious collisions with site objects and machinery. For capturing worker movements in dynamic construction environments which involves moving and changing objects, a solution based on semantic trajectories and Hidden Markov Model (HMM) is presented in the form of four subsystems. First, a real-time data collection and trajectory pre-processing subsystem is constructed for extracting multifaceted trajectory characteristics and stay locations of the workers using spatio-temporal data that will help in recognizing the important regions in the building for categorizing the worker movements. Second, to enable the desired semantic insights for better understanding the underlying meaningful worker movements using the contextual data related to the building environment, an ontologybased STriDE (Semantic Trajectories in Dynamic Environments) model is applied which has an ability to track information about the evolution of moving and changing building objects, and outputs semantic trajectories. The third subsystem uses the Hidden Markov Model (HMM) which is the most preferred probabilistic approach in the literature for describing the object behavior in time. An entire set of trajectories belonging to a stay location (a semantic region) is analyzed by categorizing the worker movements into four states using the HMMs along with the Viterbi algorithm. In the end, the output of the Viterbi algorithm is visualized using a BIM model for identifying the most probable high-risk locations involving sharp worker movements and rotations. The developed system will help safety managers in monitoring and controlling building activities remotely in dynamic environments by better understanding the worker behaviors for an improved safety management in day-to-day building operations as well as by preventing other workforce from accessing such hazardous locations which involve risky movements.

Spatio-temporal dataset of building occupants

“…The described dataset presented above has been utilized in developing applications for monitoring worker safety by analyzing their movements in real-time [3,4]. As the collected spatio-temporal data consist of ordered sequences of discrete-time triples in the form of 〈latitude, longitude, timestamp〉 include a timestamp (i.e.…”

“…For computing the step lengths and turning angles, a package named ‘moveHMM’ in R studio is used [5]. For more details on the calculation of step length and turning angle see our research article [4].

Fig.

…”

“…The resulted analyses will help us to generate stay, walk and run segments by utilizing the trajectory information. The generated stay segments will provide insights about the stay locations of users in different regions of a building [4]. Whereas, the run segments will indicate the disturbance in a building which should be considered for safety management [4].…”

“…1BLE beacons by Kontakt.io used for data collection.

Fig. 2Mounting BLE beacons on different floors in a building [3,4].

Fig.

…”

Improving the mathematical model of change in the body state of an employee

“…Studies were described in [13] on the use of meta-synthesis and interpretative structural modeling to identify individual cognitive factors that affect the dangerous behavior of Iranian industrial workers. Issues of application of some current technologies for collecting and processing knowledge about the behavior of workers in dynamically changing environments are considered in [14] on an example of a construction company. The problems of using systematic data maps to fill in the gaps in objective data on the state of a person and solving other problems of managing health risks associated with exposure to chemicals were considered in [15].…”