Modern hybrid robot cells leverage heterogeneous agents to provide agile production solutions. Effective agent coordination is crucial to avoid inefficiencies and potential hazards for human operators working among robots. This paper proposes a new human-aware proactive task allocation and scheduling model based on Mixed Integer Non-Linear Programming (MINLP) to optimize efficiency and safety during task planning, scheduling, and allocation. The approach introduces a synergy index that encodes the coupling effects between pairs of tasks executed in parallel by the agents. These terms are learned from previous executions using a Bayesian estimation; the inference of the posterior probability distribution of the synergy coefficients is performed using the Markov Chain Monte Carlo method with the No-U-Turn Sampler (NUTS) algorithm. The synergy enhances task planning by adapting the nominal duration of the plan according to the effect of the operator's presence. Simulations and experimental results demonstrate the effectiveness of the proposed method in obtaining a proactive human-aware solution starting from the task planning level. The proposed model reduces process execution time (up to 18 %) and achieves solutions with less interference between agents, and larger human-robot distance, thus making them safer for agents.
A mosaic composition problem (simulating pick-place operations in an industrial setting) was tested in simulation. In this scenario, each agent is tasked with creating its own mosaic using 4 proprietary objects from its workspace and 2 objects from a collaborative workspace, to evaluate the coupling effects between agents. The problem involves a total of 24 tasks with precedence constraints governing the pick & place operations. The problem under consideration requires all the constraints in the proposed model. Two different safety scenarios are tested: "velocity scaling" (modulated robot speed based on h-r distance) and "safety areas" (safety zones around the robot that determine slowing down and stopping) both conforming the ISO/TS 15066.
Robot Synergies "Safety Areas" simulation
Robot Synergies "Velocity Scaling" simulation
The figure below illustrates the plans obtained using different task planners (these results are inherent to the 'safety areas' scenario are given as an example). Notably, the proposed methods (Synergistic TP and Relaxed S. TP) demonstrate solutions where the interactions between humans and robots are beneficial. These methods avoid causing slowdowns and risky situations for the operator.
The graphs below show the results obtained for both scenarios (safety areas and velocity scaling). The first comparison is relative to the execution durations (plots are interactive), while the second comparison is relative to the likelihood of the human-robot distance.
Duration comparison for "Safety Areas" simulation
Duration comparison for "Velocity Scaling" simulation
(Minumum) Human-Robot Distance comparison for "Safety Areas" simulation
(Minumum) Human-Robot Distance comparison for "Velocity Scaling" simulation
| Logo | Description |
|---|---|
|
In the reported links you can find the data (h-r distances, tasks durations, tasks properties, tasks synergies, task statistics) of simulations and the real-world case study presented in the paper. In addition, you can find the code used to reproduce the charts reported in the paper results. |
|
|
Shortcut to safety areas simulations data. |
|
|
Shortcut to velocity scaling simulations data. |
|
|
Shortcut to e-waste experimental case study data. |
|
Shortcut to statystical analysis (Mann Whitney Wilcoxon Test) on questionnaire responses |
|
|
Code with task allocation and scheduling planning algorithms, along with task statistics and synergies estimation. The code is experimental and has a lot of dependencies. It is currently being refactored, organized and modularized to create a library of MINLP-based planners with minimal dependencies. |
@article{synergystic_hrtp_sandrini,
title = {Learning and planning for optimal synergistic human–robot coordination in manufacturing contexts},
journal = {Robotics and Computer-Integrated Manufacturing},
volume = {95},
pages = {103006},
year = {2025},
issn = {0736-5845},
doi = {https://doi.org/10.1016/j.rcim.2025.103006},
url = {https://www.sciencedirect.com/science/article/pii/S0736584525000602},
author = {Samuele Sandrini and Marco Faroni and Nicola Pedrocchi},
}