In the midst of the Covid-19 pandemic, the world has been attacked at this time, food delivery robots are one of the solutions to reduce connectivity between people in restaurants. Droplet is the transmission of Covid through coughing, sneezing, talking and also breathing out of the mouth and nose. This research has developed a prototype of a waiter robot that can run autonomously without markers to replace the role of humans in delivering food. A waiter robot can pass a previously mapped path. Track mapping is formed by moving the robot manually throughout the work area. The robot can generate path nodes and robot coordinates (x, y) from rotary sensors both encoder and odometry. After all trajectories of the work area have been mapped, the robot can determine the shortest path using Dijkstra's algorithm. The optimization of the movement of the autonomous waiter robot is built using three omni-directional. The results obtained through a series of tests with Dijkstra are that it is able to work well in finding the destination node from the origin node even though there are changes in the work area and the smooth movement generated by the three omni-directional trajectories.