ICRA 2019 Papers

This study investigates the high speed 3D navigation of rotating millimeter-scale swimmers. The swimmers have a spiral-shaped surface to ensure propulsion. The rotational movement is used for propulsion and, in future work, could provide the power needed to remove blood clots. For instance, an abrasive tip could be used to progressively grind a blood clot. An algorithm to perform 3D control of rotating millimeter-scale swimmers was implemented and tested experimentally. The swimmers can follow a trajectory and can navigate without touching the walls inside a tube having a diameter of 15 mm. This diameter is smaller than the average diameter of the distal descending aorta, which is the smallest section of the aorta. Several swimmers designs were built and tested. The maximum velocity recorded for our best swimmer was 103.6 mm/s with a rotational speed of 477.5 rotations per second.

Large-scale point cloud generated from 3D sensors is more accurate than its image-based counterpart. However, it is seldom used in visual pose estimation due to the difficulty in obtaining 2D-3D image to point cloud correspondences. In this paper, we propose the 2D3D-MatchNet - an end-to-end deep network architecture to jointly learn the descriptors for 2D and 3D keypoint from image and point cloud, respectively. As a result, we are able to directly match and establish 2D-3D correspondences from the query image and 3D point cloud reference map for visual pose estimation. We create our Oxford 2D-3D Patches dataset from the Oxford Robotcar dataset with the ground truth camera poses and 2D-3D image to point cloud correspondences for training and testing the deep network. Experimental results verify the feasibility of our approach.

It is a significant problem to predict the 2D LiDAR map at next moment for robotics navigation and path-planning. To tackle this problem, we resort to the motion flow between adjacent maps, as motion flow is a powerful tool to process and analyze the dynamic data, which is named optical flow in video processing. However, unlike video, which contains abundant visual features in each frame, a 2D LiDAR map lacks distinctive local features. To alleviate this challenge, we propose to estimate the motion flow based on deep neural networks inspired by its powerful representation learning ability in estimating the optical flow of the video. To this end, we design a recurrent neural network based on gated recurrent unit, which is named LiDAR-FlowNet. As a recurrent neural network can encode the temporal dynamic information, our LiDAR-FlowNet can estimate motion flow between the current map and the unknown next map only from the current frame and previous frames. A self-supervised strategy is further designed to train the LiDAR-FlowNet model effectively, while no training data need to be manually annotated. With the estimated motion flow, it is straightforward to predict the 2D LiDAR map at the next moment. Experimental results verify the effectiveness of our LiDAR-FlowNet as well as the proposed training strategy. The results of the predicted LiDAR map also show the advantages of our motion flow based method.

This paper presents a nonprehensile manipulation based on the vibration of a plate, in which three degrees of freedom (DoF) of a planar part are controlled using only one actuator. First, the model of a manipulator with a flat plate end effector is proposed. The manipulator employs an underactuated mechanism including an active joint and multiple passive viscoelastic joints, in which the joint axes are arranged nonparallel to each other. Based on the model, the orbit of the plate for a sinusoidal displacement input to the active joint is theoretically derived. It is revealed that not only the orbital shape but also the orbital direction can be varied according to the input frequency. Based on the switching frequency of the orbital direction, a design index for the mechanical parameters is shown. Subsequently, the contribution of the switching of the orbital direction to the three-DoF manipulation of a part is explored via simulation. Eight primitives utilizing the plate orbital motions in both counter-clockwise and clockwise directions are provided. Finally, the proposed method is demonstrated by experiments.