Decentralized and Intelligent Systems to Enhance Workplace Safety

Occupational injuries and illnesses continue to affect workers worldwide, resulting in severe consequences for the affected individuals, companies, and society in general. Despite initiatives to improve workplace safety and legislative efforts, the number of accidents still remains very high, highlighting the fact that relying entirely on manual control by safety managers and on the common sense of workers is not enough. Although several efforts have been made to try to solve specific problems in this area through cutting-edge technologies, they offer limited solutions that often struggle in real-world scenarios due to technological, scalability, or privacy-related limitations. In this context, this thesis addresses these state-of-the-art challenges, particularly those arising in the fields of the Internet of Things (IoT), Artificial Intelligence (AI), and Distributed Ledger Technologies (DLTs), in order to implement strategies that improve how companies manage safety in their workplaces. Specifically, we propose an overall framework for monitoring the workplace's safety, certifying compliance with standards, and providing assistance. This thesis presents two main contributions focusing on monitoring potential hazardous situations and ensuring compliance with safety or sustainability policies. The first contribution consists of a system to improve monitoring of Personal Protective Equipment (PPE) use, while the second is a system that certifies sensor data from a workplace to demonstrate the company's engagement in following safety and sustainable procedures. The personal protective equipment monitoring system is based on an Operator Area Network (OAN) consisting of two different types of IoT devices we defined and developed using ESP32-S3 devices. We detect situations where a PPE has not been used correctly using Machine Learning algorithms that analyze the Received Signal Strength Indicator (RSSI) followed by a post-processing algorithm to enhance performance. This work improves the state of the art by offering better portability between different workers and work environments: it is non-intrusive, inexpensive, and requires no wiring to be used in the field. The system for certifying data, on the other hand, is based on using DLTs, specifically a Hyperledger Fabric blockchain, to certify data from sensors. We have developed a complete certification system empowered by smart contracts, allowing users to create groups, define custom certification policies, choose the best one for the group, and monitor compliance with the adopted policy. By empowering user groups to define and vote on certification policies, our system promotes transparency, accountability, and trust among stakeholders. Experimental results show that our system allows users to identify the most valuable subset of IoT data for certification, reducing the amount of data stored in the blockchain and thus improving efficiency while meeting stakeholders' diverse needs. This thesis addresses open problems in several areas of computer science and takes advantage of them to improve occupational safety. It enhances worker protection measures and promotes a more comprehensive utilization of cutting-edge technologies in order to support industrial safety.