Development of an Autonomous EV Navigation System Using Instance Segmentation for Medical Supply Delivery

This paper describes an approach for using instance segmentation techniques to create an autonomous navigation system for driverless electric vehicles (EVs). To replace conventional transportation methods that mostly depend on human drivers, the main goal is to move medical supplies between buildings in the Suranaree University of Technology (SUT) Hospital region. Five primary components comprise the development of the EV-based medical supply transport system: i) the steering and drive systems; ii) autonomous navigation and obstacle detection systems; iii) instance segmentation techniques; iv) RP lidar for obstacle detection; and v) the EV working algorithm. The test route is designed to cover 525 m from a starting point to an endpoint, with a speed range of 0 to 1.5 m/s for the EV performed in this study. The EV will come to a turn left at 251 and 335 m along this path and a turn right at 293 m. In the instance segmentation autonomous navigation test, the EV maintained an average speed of 0.99 m/s and an average power usage of 0.22 kW while driving in a straight line. However, the EV's speed and power consumption decreased to zero to enable directional corrections when it encountered turns, including a turn right at 293 m and turns left at 251 m and 335 m. As a result, the EV was able to complete the required turns and proceed straight ahead until it reached the destination.