Teleoperation of Semi-autonomous Robots Through Uncertain Environments

Robot navigation tasks are usually decomposed into sequential sub-tasks, thus requiring a hierarchical design of sub-systems governed by multiple decision and control loops. In practice, people are the experts in making high-level decisions based on environmental knowledge. However, obtaining the local, real-time environment information is difficult by ourselves, especially in unreachable or dangerous environments. On the contrary, robots are good at achieving lower-level tasks such as building a cost map and detecting nearby obstacles. Therefore, teleoperation is used to coordinate strategic human brains and faithful robot sensors and actuators, providing an information-sharing protocol that helps to build a human-in-loop system. In this paper, we design a set of interactive Robot Operating System (ROS) based teleoperation windows for the operators to supervise the robot’s navigation through uncertain environments. Human operators will have the ability to modify the global cost map of the robot, which describes where obstacles are present in the robot’s environment. Currently, the cost map is constructed by a static map of the environment and image data from the onboard depth camera and is used to formulate the robot’s global path plan with Djikstra’s algorithm. With our enhancement, users can draw artificial walls around non-traversable areas and have the robot update its global path plan in real-time. The user can also switch to keyboard teleoperation mode to either stop the robot or adjust the robot’s orientation. Finally, the user can specify waypoints and perform waypoint-based navigation, where the robot is given a set of smaller destinations rather than one final destination. This teleoperation stack is verified by a navigation task in a manually set complex Gazebo simulation environment.