Optimization of Laboratory Equipment Sharing Scheduling Strategy Based on Reinforcement Learning

Abstract

There are many problems with the ideal scheduling of equipment resource in shared laboratories, such as unstable and irregular demand, constant fluctuation of inventory, and conflicting goals. The problem with the previous planning method is that it cannot handle a large number of states and changes in the environment; therefore, this paper proposes a way to improve adaptive planning based on deep reinforcement learning. Based on the above method, a dynamic scheduling model can be constructed to integrate digital representations of multi-source heterogeneous equipment, time-window-constrained random demand sequences, and real-time state-aware inventory maps, and thus obtain an organised state input for intelligent decision-making. Then, the state space of inventory features and a multi-objective reward-shaping function are constructed, and finally, the proximal policy optimisation algorithm is used to train a scheduling network that can learn independently and produce an adaptive scheduling strategy. Based on the results of the time series forecast, a demand response module will be added to change the schedule dynamically and be connected with the collaborative pricing mechanism for the remaining inventory. Based on the simulation experiments, the above method has improved the utilisation of resources, the rate of request satisfaction and system stability compared with the baseline method; therefore, it can increase the operating efficiency and supply-demand matching accuracy of the laboratory equipment sharing system.