Application of Evolutionary Algorithms for Humanoid Robot Motion Planning

This paper proposed a new method for humanoid robot gait generation based on several objective functions. The proposed method is based on multiobjective evolutionary algorithm. In our work, we considered two competing objective functions: MCE and MTC. Based on simulation and experimental results, we conclude: Multiobjective evolution is efficient because optimal humanoid robot gaits with completely different characteristics can be found in one simulation run. The nondominated solutions in the obtained Pareto-optimal set are well distributed and have satisfactory diversity characteristics. The optimal gaits generated by simulation gave good performance when they were tested in the real hardware of "Bonten-Maru" humanoid robot. The optimal gait reduces the energy consumption and increases the stability during the robot motion. In the future, it will be interesting to investigate if the robot can learn in real time to switch between different gaits based on the environment conditions. In uneven terrains MTC gaits will be more

When a human needs to generate a large force, the human will try to apply an impulsive force cooperating whole body. However, exerting impulsive force is one of the most difficult tasks for humanoid robots, because a reaction force caused by the applied impulsive force may bring the humanoid robot down. This paper discusses a humanoid robot motion to generate a large force utilizing an impact. Breaking wooden boards by Karate-chop is taken as a case study because it is a typical example of tasks that utilize impulsive force. A humanoid robot motion for the Karate-chop is carefully designed to maximize the speed of a hand keeping the stability. In order to validate the designed motion, experiments are carried out using a small humanoid robot Fujitsu HOAP-2. In the experiments, three kinds of wooden boards are tested to break by the Karate-chop. The strength of wooden boards is estimated striking the boards with a free-falling plummet. Stabilizing control is also applied to the robot in order to stabilize the body after the impact. The experimentation results are reported in the paper

This paper contributes to the problem of humanoid robot gait generation in unknown environments. The intention of the proposed method is to create an autonomous humanoid robot, able to take decisions and generate the appropriate optimal gait based on the information received by the eye system. Up to now, we have created two modules: walking and going upstairs. In order to create an autonomous humanoid robot, we plan to consider other tasks like going downstairs, creeping, obstacle overcoming, etc. In this paper, we present the simulation and experimental results for real time humanoid robot gait generation realized with the “Bonten‐Maru I” humanoid robot. The results showed that the Neural Network modules generate in a very short time a stable humanoid robot motion.

In the near future, humanoid robots will act as the partners of human beings in daily life. Among numerous human-like competencies, motion of humanoid robots is critical for providing humans with richer interactions with such robots. Motion plays an essential role in complementing spoken communication. Moreover, the motions of humanoid robots generate nonverbal communication in various contexts. Through this nonverbal communication, humans can interact with robots not only directly but also indirectly or even unconsciously, as if the robot were simply part of the environment. Before the developments of humanoid robots, embodied conversational agents (ECAs) were introduced as virtual embodied representations of humans that communicated multi-modally with humans and there has been a great deal of research on ECA behavior. ECAs and humanoid robots share many features in terms of how they communicate with humans. Nevertheless, simply adapting knowledge gathered from current ECA studies to a humanoid robot study is insufficient for the following reasons: 1. ECA studies lack knowledge focused on nonverbal communication, which has become more important in the physical world; 2. ECA studies have focused on developing agent-centered intelligence rather than a user- centered experience; 3. ECA studies have developed logics to generate motions automatically rather than to provide designers with the practical knowledge necessary to design desirable motions. Motivated by these three arguments, we seek to pioneer a new field of motion design between robot engineering and design discipline. To bring this motion study into design discipline, we focused on human-centered experience through nonverbal communication with a humanoid robot. This paper aims to outline sharable user experience in order to help designer create desirable motions for humanoid robots in various speechless contexts.

Learning-based algorithms remain a challenge in the field of humanoid robot motion control. Due to the diverse range of humanoid robot motions and the complexity of control, existing motion control methods often require the development of distinct control frameworks or using different network architectures. This paper proposes a keyframe-driven whole-body motion control framework for humanoid robots that allows the robot to perform a variety of motions through manually designed actions or video demonstrations. We introduce a humanoid robot motion classification method to enable the robot to learn various types of action. Moreover, we design a Dual Actor Single Critic Network(DASCN) network for whole-body control of the humanoid robot. Experimental results demonstrate that our method enables the humanoid robot H1 to execute a range of movements based solely on pre-designed motions or demonstrated videos.

Humanoid robots are similar in dimension and degrees of freedom to humans. Compared with other types of robots, humanoid robots are more adaptable to complex living environment and various tools of humans. The combination of wheeled robot and humanoid robot is an effective way to increase the moving speed, range of motion and environmental adaptability of humanoid robots, which can accelerate the application of humanoid robots in human living environments. In this way, the use of humanoid robots in industry and services can be effectively improved. This article introduces a novel electrically-driven wheel-legged humanoid robot BHR-W with 14 degrees of freedom. It can use wheels to achieve rapid motion on structured pavement. In addition, BHR-W has two degrees of freedom of hip joint, giving it excellent ability to perform complex actions such as walking, jumping and roller skating to cross unstructured environments. BHR-W is an innovative design which enables humanoid robots to have capabilities of bipedal walking and wheel rolling. This article introduces the mechanical and system design of the robot, and designs a controller for balance control. Finally, the stability and reliability of the system and the feasibility of the balance control strategies are verified by several practical experiments.

Whole-Body Control has been extensively used to achieve humanoid robot force and motion tasks simultaneously during recent years. However, most existing results have not incorporated low-level actuator dynamics and time delays yet. In this study, we propose a novel time-delayed Whole-Body Operational Space control (WBOSC) with series elastic actuator (SEA) dynamics. This type of controller generalizes our previously proposed distributed control structure to multi-input and multi-output free floating humanoid robotic systems. Namely, Cartesian stiffness control is adopted to design the WBOSC at the centralized level while motor damping control is implemented at the embedded level to remedy the stability deterioration caused by time delays. Additionally, embedded-level torque feedback control is formulated and physically interpreted as a shaping of the motor inertia. To ensure passivity, we separate the overall system into two subsystems interconnected in a feedback configuration. By the Lyapunov-Krasovskii functional technique, we propose a delay-dependent passivity criterion of the closed-loop system in the form of linear matrix inequalities (LMIs), and solve for the allowable maximum time delays via the passivity criterion. Numerical simulations of a dynamic locomotion process are used to validate the proposed passivity criterion and the WBOSC framework.

The present study evaluates the cognitive representation of a kicking movement performed by a human and a humanoid robot, and how they are represented in experts and novices of soccer and robotics, respectively. To learn about the expertise-dependent development of memory structures, we compared the representation structures of soccer experts and robot experts concerning a human and humanoid robot kicking movement. We found different cognitive representation structures for both expertise groups under two different motor performance conditions (human vs. humanoid robot). In general, the expertise relies on the perceptual-motor knowledge of the human motor system. Thus, the soccer experts' cognitive representation of the humanoid robot movement is dominated by their representation of the corresponding human movement. Additionally, our results suggest that robot experts, in contrast to soccer experts, access functional features of the technical system of the humanoid robot in addition to their perceptual-motor knowledge about the human motor system. Thus, their perceptual-motor and neuro-functional machine representation are integrated into a cognitive representation of the humanoid robot movement.

A humanoid robot’s development requires an incredible combination of interdisciplinary work from engineering to mathematics, software, and machine learning. NAO is a humanoid bipedal robot designed to participate in football competitions against humans by 2050, and speed is crucial for football sports. Therefore, the focus of the paper is on improving NAO speed. This paper is aimed at testing the hypothesis of whether the humanoid NAO walking speed can be improved without changing its physical configuration. The applied research method compares three classification techniques: artificial neural network (ANN), Naïve Bayes, and decision tree to measure and predict NAO’s best walking speed, then select the best method, and enhance it to find the optimal average velocity speed. According to Aldebaran documentation, the real NAO’s robot default walking speed is 9.52 cm/s. The proposed work was initiated by studying NAO hardware platform limitations and selecting Nao’s gait 12 parameters to measure the accuracy metrics implemented in the three classification models design. Five experiments were designed to model and trace the changes for the 12 parameters. The preliminary NAO’s walking datasets open‐source available at GitHub, the NAL, and RoboCup datasheets are implemented. All generated gaits’ parameters for both legs and feet in the experiments were recorded using the Choregraphe software. This dataset was divided into 30% for training and 70% for testing each model. The recorded gaits’ parameters were then fed to the three classification models to measure and predict NAO’s walking best speed. After 500 training cycles for the Naïve Bayes, the decision tree, and ANN, the RapidMiner scored 48.20%, 49.87%, and 55.12%, walking metric speed rate, respectively. Next, the emphasis was on enhancing the ANN model to reach the optimal average velocity walking speed for the real NAO. With 12 attributes, the maximum accuracy metric rate of 65.31% was reached with only four hidden layers in 500 training cycles with a 0.5 learning rate for the best walking learning process, and the ANN model predicted the optimal average velocity speed of 51.08% without stiffness: V1 = 22.62 cm/s, V2 = 40 cm/s, and V = 30 cm/s. Thus, the tested hypothesis holds with the ANN model scoring the highest accuracy rate for predicting NAO’s robot walking state speed by taking both legs to gauge joint 12 parameter values.

Robot walking on two feet is a complex motion. Researchers are trying to improve biped robots walking in case of walking speed. The analysis of biped walking patterns is used to obtain more detailed information in this field. Several researches have been done to achieve this purpose and the equation of walking trajectory is one of them. This article will introduce a new algorithm in which an evolutionary computing, based on learning automata along with a continual action on control signals of humanoid robot's motion, showing the success of the proposed method as the result to be used for optimizing the parameters of Truncated Fourier Series (TFS) after being compared with the results of Genetic Algorithm (GA) implementation. It is notable that the conditions of the experiment for these two algorithms are considered to be identical.

In these two decades, the technology of bipedal and humanoid robots has been made great progress. Up so far, many bipedal and humanoid robots are successfully developed (Hirai et al., 1998; Takanishi et al., 1985; Ogura et al., 2004; Kaneko et al., 2004; Nakasaka et al., 2004; Loffler et al., 2003; Lohmeier et al., 2004). In these robots, the gait planning and control for bipedal walking are based on ZMP concept (Vukobratovich et al., 1990; Goswami, 1999; Kakami, 2000). There are generally two methods for implementation of the dynamic and stable walking. The first one is based on a multiple rigid body robot model (Takanishi et al., 1985). This model is comparatively precise, but it requires a huge computation cost. The second one is simply based on the assumption that the whole robot mass is concentrated to robot’s CoG (Center of Gravity). The typical approach is using a 3D inverted pendulum model (Kajita & Kobayashi, 1987; Kajita et al., 2002; 2003; Zhu et al., 2003, 2004), in which, the robot locomotion in the sagittal and lateral planes are supposed to be independent. However, up to now, the bipedal walking nature and limitations of walking speed, stride, stride motion time, and so on, have not been completely understood. Partially this is because (1) few attentions are paid on the motion in double support phase and its effect to whole bipedal walking; (2) the investigation is not done intensively even for single support phase; for example, motion in sagittal plane are simply dealt with independent of the motion in lateral plane in most literatures on biped and humanoid robots. As revealed latter, in fact, such an approach is improper since the motions in the two planes are strongly coupled together. (3) ZMP is fixed to a point in robot sole for most of bipedal or humanoid robots. This paper mainly discusses the above first two problems based on ZMP concept and an inverted pendulum model with the assumption that ZMP is fixed at the center of the robot sole in single support phase. To do this, the relation between the motions in the sagittal and lateral planes is investigated first. By dividing a whole bipedal walking cycle into a double support, a deceleration, and an acceleration phases, and synchronizing the motions in the sagittal and lateral planes, we point out that the motions in these two planes are strongly coupled together, in other words, the motion in the lateral plane greatly restricts the motion in the sagittal plane, vice versa. Then, the role of the double support phase in biped walking is discussed. By changing the start and finish points of the double support phases in the lateral plane, the walking parameters such as walking speed, walking period, phase stride

This paper presents a component-based framework of humanoid robot task coordination using Predicate Transition Nets (PrT Nets). Humanoid robots have a large number of degrees of freedom (DOF) and they are expected to generate human-like stable behaviors to finish missions. Their motions have to satisfy a set of constraints (balance, coordinated motions, collision free movements), which increases the complexity of the system - each movement may consist of several types of movements and tasks simultaneously. Task is defined as a sequence of movements and/or actions towards to completing a given mission. It is key to represent the task coordination precisely and correctly to program on humanoid robots' motions. To solve the challenge issue of multiple task coordination for a humanoid robot, we presented a framework that integrates Component Based Software Development (CBSD) with Predicate Transition Nets to analyze the correctness of task co ordinations. Each component's behavior can be represented by a Predicate Transition Net, and the constraints of the behaviors are described by a set of temporal logic formulae. To ensure the task co ordinations are met, a rewriting logic-based model checker is applied to verify the system against the constraints. This formal framework is general and can be used for other type humanoid robots. It allows the representation of motion plans and provides the flexibility of semantic analysis on the humanoid robotics systems.

Planning collision-free, dynamically-balanced movements for humanoid robots is a challenging problem. An effective approach consists of first planning a motion satisfying geometric and kinematic constraints (such as collision avoidance, joint angle limits, velocity limits, etc.) and, in a second stage, modifying this motion so that it respects dynamic balance criteria, such as those relative to the Zero Moment Point (ZMP). However, this approach currently suffers from the issue that the modified motion may give rise to new collisions with respect to the original motion, which can be very costly to deal with, especially for systems with many degrees of freedom and cluttered environments. Here we present an algorithm to modify the motions of humanoid robots under ZMP constraints without changing the original motion path, making thereby new collision checks unnecessary. We do so by adapting the minimum-time path parameterization under torque constraints algorithm of Bobrow et al. to the case of ZMP constraints. In contrast with a previous approach based on finite differences and iterative optimization to find the optimal path parameterization under ZMP constraints, our Bobrow-based algorithm finds this optimal parameterization in a single pass. We demonstrate the efficiency of this algorithm by simulations.

In the last decade we have witnessed a rapid growth of Humanoid Robotics, which has already constituted an autonomous research field. Humanoid robots (or simply humanoids) are expected in all situations of humans’ everyday life, “living” and cooperating with us. They will work in services, in homes, and hospitals, and they are even expected to get involved in sports. Hence, they will have to be capable of doing diverse kinds of tasks. This forces the researchers to develop an appropriate mathematical model to support simulation, design, and control of these systems. Another important fact is that today’s, and especially tomorrow’s, humanoid robots will be more and more humanlike in their shape and behavior. A dynamic model developed for an advanced humanoid robot may become a very useful tool for the dynamic analysis of human motion in different tasks (walking, running and jumping, manipulation, various sports, etc.). So, we derive a general model and talk about a human-and-humanoid simulation system. The basic idea is to start from a human/humanoid considered as a free spatial system (“flier”). Particular problems (walking, jumping, etc.) are then considered as different contact tasks – interaction between the flier and various objects (being either single bodies or separate dynamic systems).

This paper presents a method to optimize and filter trajectories generated from recorded human motion for a humanoid robot with physical limits. The objective function is responsible for mimicking human trainers, enhancing the possibility for fast convergence, while constraints are used to transform motion within the limit of the capabilities of the humanoid robot. Those constraints for angle, velocity, and dynamic force are represented as B-spline coefficients. An iterative soft-constraint paradigm is proposed to enhance the quality of velocity and force constraints. Collision avoidance is also considered as a constraint. For precision refinement, in regions of high-frequency motion not adequately modeled by an initial splining, B-spline is extensible into a hierarchy so that optimization that meets global criteria can be performed locally. Furthermore, all of the constraints can be used solely as filtering. To use these filters, an effective method to directly decompose a trajectory to a B-spline is also presented