Mobile Robot Navigation in Warehouses by MPC Handling Multiple Travel Strategies Considering Independent Safety LiDAR

摘要

This research deals with navigation, mainly trajectory planning and tracking for mobile robot in warehouses, considering safety and travel strategies. Conventional navigation systems generally integrate the measurement results of onboard sensors into a map, and if there is an obstacle obstructing the travel path, the navigation system commands the robot to avoid or decelerate in response to the obstacle. However, due to the effects of map resolution and measurement errors, the distance to obstacles in real space and on the map may diverge, which is an issue from a functional safety point of view. Therefore, recent industrial robots are being considered to be equipped with safety LiDAR to ensure higher safety functions. The safety LiDAR dynamically changes its detection range according to the robot's velocity and angular velocity measured by an on-board encoder. Since, the robot operates independently of navigation, it is difficult for conventional navigation systems to achieve optimal travel control, taking in account of changes in the detection range. In addition, in actual warehouse operations, operators must change the robot's travel strategy from time to time depending on the onsite conditions, such as work efficiency, energy consumption, and load stability. Therefore, this study proposes a navigation system that can flexibly respond to travel strategies that are important for warehouse operations, considering the safety LiDAR mentioned above. To be Specific, we focused on Model Predictive Control (MPC), which can explicitly handle a wide variety of constraints, and designed objective functions for multiple travel strategies considering the predictions of the safety LiDAR behavior. To reduce the computational load, the system was designed so that safety LiDAR behavior and travel strategies are considered only in the trajectory planning performed by the server PC, while the on-board PC only focuses on following the target trajectory. This allows the robot to only follow the planned trajectory, and to take into account the behavior of safety LiDAR, and to respond to variety of travel strategies in real time. The effectiveness of the proposed method was evaluated under real warehouse conditions.