.png){kind=icon}
.png){kind=icon}
Return
Distributed memetic algorithm with heterogeneous co-evolution for deadlock-free scheduling of flexible assembly systems
.png){kind=icon}
PRE
AI
M
L
G
L
DOI：10.1016/j.eswa.2026.133879.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper studies the scheduling problem of flexible assembly systems (FAS) with blocking and deadlock constraints. A place-timed Petri net (PTPN) model is developed to describe system dynamics and constraints, and the problem is formulated as minimizing the makespan under blocking and deadlock constraints. To ensure feasibility, a deadlock repair strategy based on earliest firing time is proposed, which detects and repairs infeasible transitions in candidate sequences according to deadlock avoidance policy. To address this problem, an efficient solution approach called distributed memetic algorithm (DMA) with heterogeneous co-evolutionary mechanism is further proposed based on a hierarchically coordinated distributed evolutionary architecture (HCDEA). The framework consists of a coordination layer and an evolutionary layer. The coordination layer contains a balance population that performs individual role assignment and manages the distribution and collection of individuals, dynamically assigning superior, inferior, and stagnated individuals to different subpopulations and merging updated solutions after evolution. The evolutionary layer includes three subpopulations: an exploitation population for greedy local search with adaptive operator selection, an exploration population for longest-common-subsequence-guided probabilistic offspring generation, and a restart population for diversity restoration using random reinitialization and Pareto-based selection. The deadlock repair strategy is embedded into both initialization and evolution stages to ensure feasibility of all solutions. Experimental results show that the proposed DMA achieves a good balance between exploration and exploitation, maintains feasibility under complex constraints, and outperforms existing methods for the FAS under consideration.
Keywords:
Flexible assembly systems
Deadlock-free scheduling
Distributed memetic algorithm
Co-evolutionary algorithm
Petri net
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
Organization
Cited Papers