International Journal of Humanoid Robotics

In this paper, a new hybrid kinematic/dynamic control scheme for humanoid robots is proposed. Its basic idea lies in the tracking of several values in both operational and joint spaces. These values include (i) the relative pose of the robot’s feet, (ii) the position of the center of mass, (iii) the body’s orientation, and (iv) the admissible range of variation of the joints. A zero-moment point (ZMP)-based dynamic feedback is included in the proposed scheme to improve the stability of dynamic motions. The proposed stabilizer is based on a spherical projection of a nonlinear PID regulation control law. Through the proposed study, it is shown that these objectives allow us to produce smooth dynamically stable whole-body motions. The effectiveness and robustness of the proposed control scheme are demonstrated through four real-time experimental scenarios conducted on the HOAP-3 humanoid robot.

This paper proposes a nonlinear optimal control approach for mulitple degrees of freedom (DOF) brachiation robots, which are often used in inspection and maintenance tasks of the electric power grid. Because of the nonlinear and multivariable structure of the related state-space model, as well as because of underactuation, the control problem of these robots is nontrivial. The dynamic model of the brachiation robots undergoes first approximate linearization with the use of Taylor series expansion around a temporary operating point which is recomputed at each iteration of the control method. For the approximately linearized model, an H-infinity feedback controller is designed. The linearization procedure relies on the Jacobian matrices of the brachiation robots’ state-space model. The proposed control method stands for the solution of the optimal control problem for the nonlinear and multivariable dynamics of the brachiation robots, under model uncertainties and external perturbations. For the computation of the controller’s feedback gains an algebraic Riccati equation is solved at each time-step of the control method. The global stability properties of the control scheme are proven through Lyapunov analysis. The new nonlinear optimal control approach achieves fast and accurate tracking for all state variables of the brachiation robots, under moderate variations of the control inputs.

This paper focuses on loop-closure detection (LCD) for a visual simultaneous localization and mapping (SLAM) system. We present a strategy that combines a Bayes filter and features from a pre-trained convolution neural network (CNN) to perform LCD. Rather than using features from only one layer, we fuse features from multiple layers based on spatial pyramid pooling. A flexible Bayes model is then formulated to integrate the sequential information and similarities that are computed by features at different scales. The introduction of a penalty factor and bidirectional propagation enables our approach to handle complex trajectories. We present extensive experiments on challenging datasets, and we compare our approach to state-of-the-art methods, to evaluate it. The results show that our approach can ensure remarkable performance under severe condition changes and handle trajectories that have different characteristics. We also show the advantages of Bayes filters over sequence matching in the experiments, and we analyze our feature fusion strategy by visualizing the activations of the CNN.

This paper proposes a novel Point-to-Point (P2P) motion planning method which can ensure consistent external behaviors when applied on a group of anthropomorphic arms. Anthropomorphic arms are special kinds of manipulators which have a Spherical–Revolute–Spherical (S-R-S) structure like a human arm. Assuming they are covered by skin, the arms with equal lengths of upper arm and forearm will be seen as the same, although their internal mechanical structure, i.e., the joint arrangement, might be different. Their base frames are equivalent and related to human body’s anatomical planes, with respect to which task parameters are normally prescribed. Consequently, with the conventional joint space method, the arms with different joint arrangements will behave inconsistently when performing an identical task. This phenomenon is essentially unhuman-like. To conquer this problem, consistent P2P motion planning method is proposed. The proposed method constructs a virtual manipulator serving as motion generator on the arms’ unique corresponding physiological structure. The virtual manipulator motion is planned in the joint space. Simulations on an anthropomorphic arm and its variant are performed to prove consistency of the proposed method and inconsistency of the conventional joint space method.