Multi-legged robots offer enhanced stability to navigate complex terrains with their multiple legs interacting with the environment. However, how to effectively coordinate the multiple legs in a larger action exploration space to generate natural and robust movements is a key issue. In this paper, we introduce a motion prior-based approach, successfully applying deep reinforcement learning algorithms to a real hexapod robot. We generate a dataset of optimized motion priors, and train an adversarial discriminator based on the priors to guide the hexapod robot to learn natural gaits. The learned policy is then successfully transferred to a real hexapod robot, and demonstrate natural gait patterns and remarkable robustness without visual information in complex terrains. This is the first time that a reinforcement learning controller has been used to achieve complex terrain walking on a real hexapod robot.

This work presents a learning-based approximation scheme to improve the computational burden of general distributed model predictive control (DMPC). Under the framework of dual decomposition, an independent neural network approximator with rectified linear unit is designed for each subsystem. The primal and Lagrangian dual analysis indicates that this error-containing approximation is a suboptimal solution of the global DMPC optimization problem. In addition, the distributed conditions designed to guarantee the feasibility and stability of global system, which inspired by an explicit-implicit procedure to approximate an MPC law, are derived from an decoupling process using dual decomposition. In cases with infeasible approximator output or the distributed conditions are violated, an backup controller will used to promote the implementation of approximation. The proposed learning-based DMPC approximator with feasibility and stability guarantees is finally employed to a reactor-separator process, and simulation results demonstrate the efficiency and superior performance of proposed strategy.

Staircase is a typical obstacle for the legged robot to overcome in buildings. This paper studies the stair climbing capability-based dimensional synthesis for a hexapod legged robot, i.e., exploring how to determine the leg length and the longitudinal body length concerning the target staircase in the mechanical design stage. In climbing a staircase, leg-staircase interference is one of the predominant issues. The three possible interference cases are illustrated in detail with a 2-DOF (degree of freedom) leg mechanism and the staircase size, based on the predefined tripod gait sequence. The mathematical relationships between the leg length, longitudinal body length, and the target staircase size are derived. The leg length and the body length are finally determined with the target staircase size. The virtual simulations and prototype experiments verify the effectiveness of the dimensional synthesis for the hexapod robot.

The autonomous robots consisting of an immovable lander and a rover are widely deployed to explore extraterrestrial planets. However, these robots have two main limitations: (1) the separate design for lander and rover respectively results in heavy mass and big volume of the whole system, which increases the launching cost sharply; (2) the rover’s detection area has to be restricted to the vicinity of the immovable lander. To overcome these problems, we designed a novel six-legged mobile repetitive lander called "HexaMRL", which integrates the functions of both lander and rover, including folding, deploying, repetitive soft-landing, and walking. A hybrid compliant mechanism taking advantages of both active and passive compliances was adopted on its leg. An integrated drive unit (IDU) was utilized to imitate the dynamics of a spring and a damper to absorb the landing impact energy, while the structure remains intact. Moreover, a control method based on state machine for soft-landing on the Moon was proposed. HexaMRL achieved repetitive soft-landing on a 5-DoF lunar gravity testing platform (5-DoF-LGTP) with a vertical landing velocity of 1.9 m/s and a payload of 140 kg. The drive torque safety margin is improved by 23.4%p based on the hybrid compliant leg comparing with the standalone active compliant leg.

Summary form only given. The design and control of quadruped robots has become a fascinating research field because they have better mobility on unstructured terrains. Until now, many kinds of quadruped robots were developed, such as JROB-1 [1], BISAM [2], BigDog [3], LittleDog [4], HyQ [5] and Cheetah cub [6]. They have shown significant walking performance. However, most of them use serial mechanism legs and have animal like structure: the thigh and the crus. To swing the crus in swing phase and support the body's weight in stance phase, a linear actuator is attached on the thigh [2, 3, 5, 6], or instead, a rotational actuator is installed on the knee joint [1, 4]. To make the robot more useful in the wild environment, e.g., the detection or manipulation tasks, the payload capability is very important. To carry the sensors or tools, heavy load legged robot is very necessary. Thus the knee actuator should be lightweight, powerful and easy to maintain. However, this can be very costly and hard to satisfy at the same time.

Central pattern generator (CPG) is a kind of neural network which is located in the spinal cord. It has been found to be responsible for many rhythmic biological movements, such as breathing, swimming, flying as well as walking. Many CPG models have been designed and proved to be useful. But the CPG outputs of these models are often sine waves or quasi-sine waves. Also these outputs are directly used as the control signals to control joint trajectories or joint torques on robots. This is obviously not an accurate design in robot walking control especially when sine or quasisine waves are not the best signals to set walking patters because of the complexity of tasks. In this paper, based on the idea of Righetti, Buchli and Ijspeert, a CPG model is designed, which is inspired by Fourier series and can produce outputs with any shape. There are a limited set of sub-components in the proposed model. Each sub-component learns one harmonic of a reference wave. A summation of these sub-components is used to approximate the wave. In this way, the wave will be learned and embedded in the CPG model. In the proposed model, FFT is used to see the harmonics and calculate the frequency. The system is designed in polar coordinates with new Hebbian learning items and Kuramoto model items. Because the whole system is a limit cycle system, it is robust to perturbation. The experiment conducted on an AIBO robot shows the effectiveness of the proposed model.

From basic neuro-physiological evidences, it is now generally accepted that animals' walking control is subject to the combination function of central pattern generator(CPG) located at the spinal cords and reflexes from the peripheral stimulus. Since phase oscillators have the advantage of mathematical tractability, it's convenient to adjust the phase relationship between them. In this paper, coupled phase oscillators were designed to simulate CPG's behavior and establish vestibular reflex with feedbacks from accelerator sensors. Afterward, the synchronization condition of this proposed CPG model was studied. Forward and backward walking, gait transfers between trot and walk were realized as well. With feedbacks, AIBO detected uphill and downhill terrain and changed its posture automatically to fit for the new environment. Simulations were done in Webots to verify this method.

Three-dimensional fuzzy logic controller (3-D FLC) is a novel FLC developed for spatially distributed parameter systems. In this study, we are concerned with data-based 3-D FLC design. A nearest neighborhood clustering algorithm is employed to extract fuzzy rules from input-output data pairs, and then an optimization algorithm based on geometric similarity measure is used to reduce the obtained rule base. The consequent parameters are estimated using linear support vector regression. Finally, a catalytic packed-bed reactor is taken as an application to demonstrate the effectiveness of the 3-D FLC.

Many industrial processes are spatiotemporal dynamic systems. A three-dimensional fuzzy-logic controller (3-D FLC) has been recently developed to process the inherent capability of spatiotemporal dynamic systems. Sensor placement, which is always crucial to the control of spatiotemporal dynamic systems, is also critical to the design of the 3-D FLC. In this paper, a new sensor-placement strategy is developed. Its main feature is to position the sensor by utilizing the main characteristics of spatial distribution. The key technique is to use a spatial-constrained fuzzy c-means algorithm to extract the characteristics of spatial distribution. For an easy implementation, a systematic sensor-placement design scheme in four steps (i.e., data collection, dimension reduction, data clustering, and sensor locating) is developed. Finally, control of a catalytic packed-bed reactor is taken as an application to demonstrate the effectiveness of the proposed sensor-placement scheme.