基于两阶段调度算法的倍捻车间机器人调度OA

To address the problem of low efficiency of robots in twisting workshops when they cope with high-frequency concurrent tasks,this paper aims to optimize the dynamic scheduling process of robot clusters and propose a complete modeling and optimization scheme.Considering the specific layout of the workshop and the distribution of obstacles,an improved Manhattan distance formula was innovatively designed.This addresses the inaccuracy of traditional distance calculations in workshop environments,thereby enhancing the accuracy of path cost estimation.Furthermore,a multi-objective dynamic scheduling model was constructed,integrating constraints such as robot battery power,the number of available robots,and dynamic task requests.The model's goal is to minimize travel mileage cost(TMC)and time window violation cost(TWVC).A time-segmented re-optimization framework was adopted to enable an efficient response to dynamic tasks,providing technical support for the intelligent scheduling of twisting workshops. A two-stage scheduling algorithm was designed to solve the proposed model.In the static scheduling stage,an adaptive non-dominated sorting genetic algorithm Ⅱ(ANSGA-Ⅱ)based on a cultural algorithm framework was introduced.This algorithm uses a regional division strategy for task decoupling and incorporates the belief space of the cultural algorithm to dynamically guide population evolution.A hybrid large neighborhood search(HLNS)was also integrated as a local search operator to help escape local optima and improve solution quality.For the dynamic scheduling stage,a hybrid dynamic scheduling strategy(HDSS)combining instant response and periodic restructuring was designed.This strategy ensures millisecond-level response speeds for new tasks while allowing for dynamic route optimization,balancing both real-time scheduling needs and route rationality. Experimental validation based on real data from a twisting workshop in Shaoxing,China,showed that the proposed ANSGA-Ⅱ algorithm significantly outperformed comparison algorithms in terms of TMC,TWVC,and Pareto solution set quality.The HDSS strategy maintained zero time window violations,with computation times controlled at the millisecond level,performing better than purely reactive insertion or complete re-optimization strategies.These results verify the effectiveness and practicality of the proposed modeling and optimization approach in complex,dynamic environments,offering key technological support for the intelligent upgrading of twisting workshops.In the future,we could introduce machine learning to predict spindle abnormalities,develop adaptive parameter tuning mechanisms for the algorithm,extend the framework to other manufacturing scenarios,or incorporate more complex constraints to further enhance its generalization capability.