Learning and planning for optimal synergistic human-robot coordination in manufacturing contexts

TL;DR

This work tackles efficient, safe collaboration between humans and robots in manufacturing by learning synergy effects between task pairs and integrating them into a MINLP-based task allocation and scheduling framework. It introduces two variants, STP and R-STP, that couple robot task durations to human actions via synergy coefficients learned probabilistically with Bayesian estimation and MCMC (NUTS). Empirical results in simulations and an e-waste disassembly case study show up to 18% reductions in makespan and safer human-robot proximity, validating both the learning component and the planning formulation. The approach is offline-planning oriented, with future work aimed at risk-aware penalties and fast replanning to handle unexpected human behavior.

Abstract

Collaborative robotics cells leverage heterogeneous agents to provide agile production solutions. Effective coordination is essential to prevent inefficiencies and risks for human operators working alongside robots. This paper proposes a human-aware task allocation and scheduling model based on Mixed Integer Nonlinear Programming to optimize efficiency and safety starting from task planning stages. The approach exploits synergies that encode the coupling effects between pairs of tasks executed in parallel by the agents, arising from the safety constraints imposed on robot 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. 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 that the proposed method produces improved human-aware task plans, reducing unuseful interference between agents, increasing human-robot distance, and achieving up to an 18\% reduction in process execution time.

Figures (8)

• Figure 1: Visual description of the proposed method. The boxes on the left show how task statistics and task synergies of the various agents are estimated using process experience, highlighting good or bad couplings between them. In the center, it is shown how these synergies are integrated to improve the MINLP algorithm for task allocation and scheduling. The boxes on the right show two industrial application scenarios that can benefit from the proposed method. Supplementary material can be found: https://jrl-cari-cnr-unibs.github.io/synergistic_hrtp/.
• Figure 2: Visualization of the involved variables during the overlapping between human and robot tasks. The measured durations ($\tilde{d}^H_k$ for the human task and $\tilde{d}^r_i$ for a task performed by a generic robot) together with the nominal duration $\hat{d}^r_i$ of the task $\tau_i$ and the nominal $\mathrm{OV}_{i,k}$ and actual $\widetilde{\mathrm{OV}}_{i,k}$ overlapping are reported.
• Figure 3: Framework architecture: a three-tiered structure that manages the pipeline from planning to execution.
• Figure 5: Results of Scenario 1: static safety zones
• Figure 6: Results of Scenario 2: continuous safety velocity scaling