We present a highly accurate and efficient, yet simple, two-stage voting scheme for distinguishing inlier 3D correspondences by densely assessing and ranking their local and global geometric consistencies. The strength of the proposed method stems from both the novel idea of post-validated voting set, as well as single-point superimposition transforms, which are computationally cheap and avoid rotational ambiguities. Using a well-known dataset consisting of various 3D models and numerous scenes that include different occlusion rates, the proposed scheme is evaluated against state-of-the-art 3D voting schemes, in terms of both the correspondence PR (precision-recall) AUC (area under curve), and the execution time. A total of 374 experiments were conducted for each method, which involved a combination of four models, 50 scenes, and two down-samplings. The proposed scheme outperforms the state-of-the-art 3D voting schemes in terms of both accuracy and speed. Quantitatively, the proposed scheme scores 97.0%±12.9% on the PR AUC metric, averaged over all of the experiments, while the two state-of-the-art schemes score 74.2%±22.2% and 78.3%±26.4%. Furthermore, the proposed scheme requires only 24.1%±6.0% of the time consumed by the fastest state-of-the-art scheme. The proposed voting scheme also demonstrates high robustness against occlusions and scarce inliers.

If we define a mode as a set of specific configurations that hold the same constraint, and if we investigate their transitions beforehand, we can efficiently probe the configuration space by using a manipulation planner. However, when multiple mobile robots together manipulate an object by using the non-prehensile method, the candidates for the modes and their transitions become enormous because of the numerous contacts among the object, the environment, and the robots. In some cases, the constraints on the object, which include a combination of robot contacts and environmental contacts, are incapable of guaranteeing the object’s stability. Furthermore, some transitions cannot appear because of geometrical and functional restrictions of the robots. Therefore, in this paper, we propose a method to narrow down the possible modes and transitions between modes by excluding the impossible modes and transitions from the viewpoint of statics, kinematics, and geometry. We first generated modes that described an object’s contact set from the robots and the environment while ignoring their exact configurations. Each multi-contact set exerted by the robots and the environment satisfied the condition necessary for the force closure on the object along with gravity. Second, we listed every possible transition between the modes by determining whether or not the given robot could actively change the contacts with geometrical feasibility. Finally, we performed two simulations to validate our method on specific manipulation tasks. Our method can be used in various cases of non-prehensile manipulations by using mobile robots. The mode transition graph generated by our method was used to efficiently sequence the manipulation actions before deciding the detailed configuration planning.

The transportation of large-scale objects in a narrow space is a challenging, but useful application for mobile robots. We have developed a three-mobile-robot system to cooperatively lift, support, and transport a large object on a mobile robot system. To facilitate the manipulation involved with loading an object onto the robots, where the robots must maintain firm contact with the object and anchor at its location, we designed an adaptable mechanism for the object-loading platform and a liftable brake to ensure that the robot remains stationary when necessary. Furthermore, each robot is designed to act as an omnidirectional wheel; therefore, all three robots can work as an omnidirectional block when collectively loading an object. The kinematic constraints of the object–robot system and forward kinematics of the robots’ cooperative motion are proposed, and experiments are conducted to confirm the designed mechanisms and confirm that the robots can load an object and cooperatively transport it along the expected trajectory.

Muscle functional MRI (mfMRI) is an imaging technique that assess muscles’ activity, exploiting a shift in the T2-relaxation time between resting and active state on muscles. It is accompanied by the use of electromyography (EMG) to have a better understanding of the muscle electrophysiology; however, a technique merging MRI and EMG information has not been defined yet. In this paper, we present an anatomical and quantitative evaluation of a method our group recently introduced to quantify its validity in terms of muscle pattern estimation for four subjects during four isometric tasks. Muscle activation pattern are estimated using a resistive network to model the morphology in the MRI. An inverse problem is solved from sEMG data to assess muscle activation. The results have been validated with a comparison with physiological information and with the fitting on the electrodes space. On average, over 90% of the input sEMG information was able to be explained with the estimated muscle patterns. There is a match with anatomical information, even if a strong subjectivity is observed among subjects. With this paper we want to proof the method’s validity showing its potential in diagnostic and rehabilitation fields.

Local image features are invariant to in-plane rotations and robust to minor viewpoint changes. However, the current detectors and descriptors for local image features fail to accommodate out-of-plane rotations larger than 25°–30°. Invariance to such viewpoint changes is essential for numerous applications, including wide baseline matching, 6D pose estimation, and object reconstruction. In this study, we present a general embedding that wraps a detector/descriptor pair in order to increase viewpoint invariance by exploiting input depth maps. The proposed embedding locates smooth surfaces within the input RGB-D images and projects them into a viewpoint invariant representation, enabling the detection and description of more viewpoint invariant features. Our embedding can be utilized with different combinations of descriptor/detector pairs, according to the desired application. Using synthetic and real-world objects, we evaluated the viewpoint invariance of various detectors and descriptors, for both standalone and embedded approaches. While standalone local image features fail to accommodate average viewpoint changes beyond 33.3°, our proposed embedding boosted the viewpoint invariance to different levels, depending on the scene geometry. Objects with distinct surface discontinuities were on average invariant up to 52.8°, and the overall average for all evaluated datasets was 45.4°. Similarly, out of a total of 140 combinations involving 20 local image features and various objects with distinct surface discontinuities, only a single standalone local image feature exceeded the goal of 60° viewpoint difference in just two combinations, as compared with 19 different local image features succeeding in 73 combinations when wrapped in the proposed embedding. Furthermore, the proposed approach operates robustly in the presence of input depth noise, even that of low-cost commodity depth sensors, and well beyond.