Robot Navigation
Day-to-night navigation using a single experience map
This paper presents a robust monocular visual teach-and-repeat (V-TR) navigation system for long-term operation in outdoor environments. The approach leverages deep-learned descriptors to deal with the high illumination variance of the real world.
3D lidar based object detection with semantic map prior for human-aware navigation
This research presents a multi-sensor perception system to perceive humans, robots and other obstacles. With understanding of the semantics the robot can navigate in the dynamic environment decide to wait or replan the local trajectories according to the types of obstacles.
EPANer participated in the World Robot Challenge 2018 - Partner Robot Challenge (Real Space) in Tokyo in October 2018, and eventually took the fifth place among 14 teams from various countries. Some mature methods and conventional ideas used in robotics were quickly integrated into the competition on-site, such as finite-state machine for task flow control, waypoint-based topological navigation for robot navigation, 3D point cloud processing for object pose estimation, and YOLO algorithm for object recognition. On the other hand, the team also learned some lessons and gained some research inspiration from the competition.
This paper presents a novel 3DOF pedestrian trajectory prediction approach for autonomous mobile service robots. While most previously reported methods are based on learning of 2D positions in monocular camera images, our approach uses range-finder sensors to learn and predict 3DOF pose trajectories (i.e. 2D position plus 1D rotation within the world coordinate system). Our approach, T-Pose-LSTM (Temporal 3DOF-Pose Long-Short-Term Memory), is trained using long-term data from real-world robot deployments and aims to learn context-dependent (environment- and time- specific) human activities. Our approach incorporates long-term temporal information (i.e. date and time) with short-term pose observations as input. A sequence-to-sequence LSTM encoder-decoder is trained, which encodes observations into LSTM and then decodes as predictions. For deployment, it can perform on-the-fly prediction in real-time. Instead of using manually annotated data, we rely on a robust human detection, tracking and SLAM system, providing us with examples in a global coordinate system. We validate the approach using more than 15K pedestrian trajectories recorded in a care home environment over a period of three months. The experiment shows that the proposed T-Pose-LSTM model advances the state-of-the-art 2D-based method for human trajectory prediction in long-term mobile robot deployments. This paper is published in ICRA 2018.