Design and Development of an Autonomous Delivery Robot (AutoVendi)

Abstract

The Autonomous Vending Robot (Auto Vendi) is an extensive and holistic project undertaken to design, develop, and fabricate a food delivery robot. Autonomous refers to the robot's self-navigation ability, which would be around a predetermined path for the sake of this project. This part can further be broken down into four major phases: perception, object detection, path planning, and consequent actuation of the robot. To complete these phases, Obstacle Detection with Lidar, Inertial Measurement Unit (IMU), and Kinect camera will be implemented to create and update a map. Adaptive Monte Carlo Localization (AMCL) is used for localization with a Dynamic Window Approach (DWA) planner as a local planner and Dijkstra algorithm as a global planner for path planning. Security and anti-theft mechanism is also implemented. Hence, using these navigation algorithms the robot is choosing the least exhaustive path to reach the destination, which is a user. In this case, the user and robot would connect via a product application available on the phone. A vending robot refers to a secured food delivering mechanism inside the robot once it reaches the user. Nvidia Jetson Nano is being used as an embedded platform to solve computationally extensive operations and gather data from sensors like IMU, Lidar, etc. In contrast, ATmega2560 is being used for motor control. We have used Linux environment and Robot Operating System (ROS) melodic distribution for path planning and navigation