Research and/or Engineering Questions/Objective With the artificial intelligence as a foundation, this research aims to develop a low-speed autonomous mobility platform with a sensor suite that gives the mobility the eyes for sensing and perception, the brain for computing and planning and the feet for moving around. It can plan an efficient and suitable path with collision avoidance via the A.I. technologies of deep learning and sensor fusion autonomously after receiving commands. Adapting to the shift of the modern world with a severe labour shortage, the low-speed autonomous mobility platform can offer flexible and versatile alternatives and extend to a wide range of applications, including last-mile delivery, inspection, roaming vending machine, pop-up store, battery swapping, autonomous micro-mobilities (scooter, wheelchairs). Methodology Unlike the industrial robots bolted firmly to the floor which blindly perform the same movement, the low-speed autonomous mobility is equipped with a sensor suite comprising 3D LiDAR, cameras, GPS, inertial measurement unit (IMU) and ultrasonic sensors. Through the technologies of deep learning with sensor fusion (a combination of sensor and vision technology), an advanced form of artificial intelligence and dynamic way of computerised decision, the low-speed autonomous mobility is capable of planning an efficient and suitable path with collision avoidance of stationary or moving obstacles in crowded and dynamic environment, not only limited to apartment complexes, parks or warehouses, but also some shared and public space. Results A prototype of low-speed autonomous mobility with a storage cabinet as last-mile delivery mobility is developed to validate the performance of the sensor suite. The prototype has been tested in our R&D building in Hong Kong for various delivery tasks with collision avoidance. Such testing allows us to enhance the system design even under an in-house condition and during the development phase. In 2020, the system will be equipped with a robotic arm to perform more versatile work. Limitations of this study Although there is no infrastructure and investment required for the mobility platform, for the target areas to deploy this autonomous mobility, the human intervention is required in the first time to build a map for the localisation. Besides, the mobility platform cannot climb up stairs due to the selection of wheel type. What does the paper offer that is new in the field including in comparison to other work by the authors? The mobility equipped with a sensor suite and software platform can be extended to various application scenarios, making it more flexible and versatile, rather than serving only one purpose. The idea of extension shortens the development time and resource. Conclusions The capability of planning an efficient and suitable path, traversing to assigned destination autonomously and collision avoidance unlock the possibilities of the low-speed mobility to perform assigned tasks autonomously in a less labour intensive, yet more effective and efficient way.
Ms. Wing Ting Law, Hong Kong Productivity Council, CHINA - HONG KONG Dr. Wai Chung Lee, Hong Kong Productivity Council, CHINA - HONG KONG Mr. Ki Sing Li, Hong Kong Productivity Council, CHINA - HONG KONG Mr. Tiande MO, Hong Kong Productivity Council, CHINA - HONG KONG