Design and Implementation of an Autonomous Driving Delivery Robot

We have developed a self-driving robot as a fully autonomous last-mile delivery solution for a variety of outdoor scenarios. The robot features a completely original design of the modern appearance, compactly and intelligently arranged sensor layout as well as a hardware platform with an extremely flexible four independent steering function. Our software framework has been established initially based on Autoware, a software platform for self-driving. To enable entirely autonomous, stable, and reliable navigation in the case of our delivery robot and taking specific characteristics of its operating scenarios into account, we propose a few crucial algorithmic enhancements and improvements. Incorporating a lidar (light detection and ranging) sensor, an IMU (Inertial Measurement Unit), a wheel odometer, and GPS (Global Positioning System) contributes to accurate mapping and localization in various critical environments such as high-tech parks with high-rise buildings or open areas. The fusion of sensing results of a 16-line lidar, a semi-solid-state lidar as well as an ultrasonic sensor allows a precise obstacle detection. With effective path planning, ground filtering, and object tracking algorithms employed, our robot is able to accurately perceive its driving environment in real time and navigates from the starting point to the target smoothly, safely, and reliably even in the presence of complicated and unexpected driving conditions. Experimental results show that the hardware platform and software algorithms of our robot meet the requirements of self-driving and delivery, significantly saving manpower and improving the distribution efficiency.