Task Travel Time Prediction Method Based on IMA-SURBF for Task Dispatching of Heterogeneous AGV System

Area of Science:

• Operations Research
• Artificial Intelligence
• Robotics

Background:

• Heterogeneous automatic guided vehicle (AGV) systems offer flexibility but require accurate task travel time prediction (T3P) for efficiency.
• Existing T3P methods face challenges due to task correlations and dynamic input/output dimensions.

Purpose of the Study:

• To develop a novel, biomimetics-inspired framework for accurate T3P in heterogeneous AGV systems.
• To address the dynamic nature of input/output dimensions and task correlations in T3P.

Main Methods:

• A radial basis function (RBF) neural network is employed for T3P, with dynamically determined input/output dimensions.
• An improved mayfly algorithm (IMA) optimizes RBF initial parameters.
• A selective update strategy refines parameter updates.

Main Results:

• The proposed IMA-SURBF framework demonstrates high accuracy and efficiency in T3P.
• Simulation experiments in a complex assembly line scenario validate the model's performance.
• The method outperforms existing deep learning-based models.

Conclusions:

• The IMA-SURBF framework provides an effective solution for T3P in complex AGV environments.
• Biomimetic approaches offer a promising direction for optimizing AGV system operations.
• Accurate T3P is essential for maximizing the agility and efficiency of heterogeneous AGV systems.