To address the issue of decreased autonomous positioning accuracy of hospital ward patrol robots under conditions of complex lighting

frequent occlusions

low texture

and dynamic disturbances

an improved visual-inertial fusion ORB-SLAM3 positioning and navigation method is proposed.Based on the traditional ORB-SLAM3 framework

this method introduces a visual inertial measurement unit(IMU)tightly coupled residual optimization model and enhances robustness and mapping accuracy in low-light and occluded environments through improved pre-integration.Considering the compact layout of hospital wards and the f

requent movement of personnel and equipment

the method combines the Navigation2(Nav2) framework

designs a global path planning strategy based on A

*

and a local obstacle avoidance strategy based on the dynamic window approach (DWA)

and incorporates a pure pursuit algorithm with speed smoothing constraints to ensure the safe and comfortable operation of the robot in medical scenarios.Experimental verification shows that this method effectively improves the stability and accuracy of task execution for ward patrol robots in complex ward environments.