A Hierarchical Stochastic Decision-Making Framework for Task Scheduling and Safety-Aware Human-Robot Collaboration

In the factories of the future, known as Industry 5.0, the goal is to bring humans and machines back together in the same workspace. For decades, automation meant replacing people with robots to handle dangerous or repetitive work. Now, the vision is different: humans and robots working side by side, combining human adaptability and judgment with robotic precision and endurance. However, sharing a workspace introduces a complex problem. A robot needs to know not just where a part is, but how the human worker is feeling. Is the worker tired? Are they moving quickly or slowly? Is the task urgent, or can it wait? If a robot cannot read these subtle signs, it might move too fast and cause an accident, or move too slowly and waste time. The challenge for engineers is to create a system that makes split-second decisions about who does what, when, and how, while keeping everyone safe and efficient.

Researchers at King Fahd University of Petroleum and Minerals have developed a new way to solve this problem. They created a computer framework that acts like a two-level brain for human-robot teams. Instead of trying to make one giant, complicated decision that covers everything at once, they split the job into two distinct parts that talk to each other. The first part is a high-level scheduler. Think of this as a manager who looks at a list of jobs, checks who is available, and decides which worker—human or robot—should take on the next task. This manager pays close attention to the "fatigue" of the workers. If a human has been working hard, the system knows to give them a break before assigning more work. If a robot has been running hot, it is told to cool down. The manager also prioritizes urgent jobs and figures out which tasks can be done at the same time without causing a bottleneck.

The second part of the system is a safety officer that operates at a lower level. While the manager decides what to do, this safety officer decides how to do it. Because a robot cannot perfectly know a human's internal state—like whether they are feeling agitated or fully engaged—it has to make educated guesses based on what it can see. The researchers used a model that watches the human's body movements, specifically the positions of their shoulders, elbows, and hands. Based on these movements, the system maintains a "belief" about the human's condition. If the movements suggest the human is restless or moving too fast, the robot automatically increases the distance between them to stay safe. If the human is calm and the task requires them to work close together, the robot moves in closer to collaborate. This safety layer is constantly updating its guess as it watches the human move, allowing the robot to react instantly to changes in behavior.

What makes this new framework unique is how these two parts connect. In many previous systems, the planning of tasks and the safety monitoring happened separately, often without sharing information. Here, the two layers are linked by a coordination module that acts as a bridge. When the manager assigns a task, it tells the safety officer the details of that job. If the job is a solo task for the robot, the safety officer keeps a wide distance. If the job requires both to work together, the safety officer allows them to get close. Crucially, the safety officer also sends information back to the manager. If a human gets too tired during a task, the safety officer signals this back to the manager, who then decides to stop assigning work to that person and schedule a rest period. This creates a closed loop where the safety of the worker directly influences the schedule of the factory.

The researchers tested this system using computer simulations to see how it would handle a stream of twelve different tasks. They created scenarios where the workload changed from mixed tasks to those requiring mostly human effort, then mostly robot effort, and finally tasks requiring close collaboration. The system successfully managed all these changes without needing to be reprogrammed for each new situation. It correctly prioritized urgent jobs, assigned tasks to the least tired worker, and ensured that both the human and the robot took breaks when their fatigue levels got too high. The simulation showed that the system could handle these decisions incredibly fast. The time it took for the system to look up the correct decision for a given situation was measured at just 3.082 microseconds. This speed means the system can react almost instantly to changes in the factory, far faster than any human manager could calculate a new plan.

One of the most important findings was that the system could maintain safety without being overly cautious. In the past, robots often had to stop or move very slowly whenever a human was nearby, just to be safe. This new approach allows the robot to be more flexible. When the human is working alone on a task, the robot stays far away. But when the human and robot are working together on a joint task, the system allows them to get close, provided the robot's "belief" about the human's state indicates it is safe to do so. The simulation confirmed that the robot never entered a dangerous collision state, yet it was able to perform collaborative tasks efficiently. The system also demonstrated that it could carry over the fatigue levels of the workers from one set of tasks to the next, ensuring that a worker who was tired at the end of one shift would not be overworked in the next.

The researchers were careful to note that their results come from computer simulations, not from physical robots working with real people in a factory yet. The "fatigue" and "mood" of the human in the simulation were based on mathematical models of how a person might react, rather than measurements from actual human participants. The system used a simplified view of human movement to make its decisions. While the results are promising, the authors state that the next step is to test this framework with real humans and real robots to see if the computer models accurately reflect how people actually behave. They plan to use sensors to measure heart rates and body movements in real time to refine the system's understanding of human states.

This work represents a significant step toward making human-robot collaboration practical and safe. By separating the complex problem of scheduling from the problem of real-time safety, and then linking them together with a fast, two-way communication system, the researchers have shown that it is possible to manage a factory floor where humans and robots work as a team. The system does not just follow a rigid set of rules; it adapts to the condition of the workers and the urgency of the jobs. It proves that with the right decision-making structure, robots can be not only strong and precise but also considerate and responsive to the people they work with. The ability to make these decisions in microseconds suggests that such systems could soon be running the factories of tomorrow, keeping workers safe while maximizing productivity.