Custom global planner for robots in retail environment

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu | Master's thesis
Control, Robotics and Autonomous Systems
Degree programme
AEE - Master’s Programme in Automation and Electrical Engineering (TS2013)
As robots, especially the wheeled mobile ones are becoming more and more ubiquitous in many sectors of industry and influencing many aspect of our daily life, the environments that they operate are also becoming more and more diverse and posing more unique challenges that requires tailor made solutions in order to achieve the desired requirements. The scope of this thesis work is about global path planning for mobile robots that operate in retail environment to gather real-time inventory data. The specific environment and operation require robot point to point navigation to have good margin from shelves and obstacles as well as high determinism where the path is preferred to be through larger aisles and walk way which can be influenced by the robot operator before runtime. Of-the-shelf solutions were found to fail to meet these requirements, therefore, a custom global path planner was designed based on a popular well-proven sampling-based path planning algorithm called RRT Connect with states sampling performed on a system of highway that carries all user preferences. Previous work on the topic of user preference path planning either requires past robot experiences that could be graded and then reused when new similar planning situation arises or uses cost map filter to design inverse keep-out zones that double as lanes that robots is only allowed to navigate on. The former limits the user ability to infuse their preference into the planning while the latter limits the robot itself to only certain lanes hindering any possibility to go off-trail when needed. The proposed solution, as shown in multiple experiments, has met all the requirements and therefore, has been used on two different mobile robot platforms that operate in the target retail and an extended hospital environment. The custom global planner is also extended with a flexible goal replanning logic that allows the robot to reach the goal as closely as possible, thus, satisfies the camera orientation constraints for the inventory scanning missions.
Visala, Arto
Thesis advisor
Ahapainen, Atte
robot navigation, path planning, user preference, ROS, highway, retail environment