Online optimization of fuzzy controller for IMC construction machinery via cycle reservoir with jumps network

Abstract

To improve the adaptability of variable universe fuzzy controllers (VUFC) in independent metering control (IMC) hydraulic systems, this paper proposes an enhanced controller by integrating a Cycle Reservoir with Jumps (CRJ) online neural network. Although conventional adaptive neuro-fuzzy inference system (ANFIS) based controllers enable offline optimization of fuzzy rules, their rule structures and parameters remain fixed after design, which limits their adaptability. In contrast, the proposed CRJ-based fuzzy controller replaces the fuzzy system in the universe adjustment layer with a neural network capable of online learning. By dynamically updating the weights of its nodes during system operation, the CRJ network generates more effective universe adjustment factors, thereby improving control performance. A key advantage of the CRJ fuzzy controller is that it does not require offline training or dataset preparation. Once the basic network structure is configured, it autonomously adapts to changes in system state, such as temperature rise in hydraulic oil or mechanical wear over extended operation. The test results demonstrate that the proposed controller achieves superior performance in both system stability and motion accuracy, validating its effectiveness for adaptive control in IMC hydraulic systems.