AUTHOR=Lee HyeokSoo , Jeong Jongpil 

TITLE=Velocity range-based reward shaping technique for effective map-less navigation with LiDAR sensor and deep reinforcement learning

JOURNAL=Frontiers in Neurorobotics

VOLUME=Volume 17 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2023.1210442

DOI=10.3389/fnbot.2023.1210442

ISSN=1662-5218

ABSTRACT=Robotics has long represented an interesting subject in science fiction movies and cartoons, and people have generally imagined robots as having some resemblance to humans. In the 1960s, robots began to be used in actual industrial sites, but the associated software technology was lacking at this time, so their use was limited to the level of simply driving and utilizing industrial robots with a focus on machinery and hardware. However, in recent years, in the field of high-tech electronic components, sensor components similar to human sensory functions have been developed and used There are also substantial developments in software technology. In particular, the use of artificial intelligence technology has enabled cognitive abilities and decision-making, such as prediction, analysis, and judgment. Recently, robot products have begun rapidly applying these hardware and software technologies, thus achieving a level of performance and completeness that was previously unimaginable. Robots that are in a humanoid form, as have been previously imagined, are expected to be commercialized in the near future. This paper studies the path planning of autonomous mobile robots, which are widely used in recent logistics and manufacturing sites, and optimal path planning through learning using LiDAR sensors as well as deep reinforcement learning without map and grid coordinate information in the workplace. We establish and discuss the factors that should be considered for further performance improvement.