Full-size Humanoid Robot Research Articles

A Bio-Plausible Approach to Realizing Heat-Evoked Nociceptive Withdrawal Reflex on the Upper Limb of a Humanoid Robot

In this letter, we present a method for realizing the heat-evoked nociceptive withdrawal reflex (NWR) in the upper limb of a humanoid robot so that it can avoid the potential damage caused by noxious heat. We use a spiking neural network whose structure, encoding scheme, and form of information transmission mimic the reflex arc in humans to improve bio-plausibility. The proper synaptic strengths between the sensory neurons and the interneurons in the first two layers are learned using the bio-plausible reward-modulated spike timing-dependent plasticity learning algorithm. By monitoring the spikes from the motor neuron in the third layer, a reflex matching the intensity of the stimulation can be evoked. Experimental evaluations show that noxious heat stimulation can be detected online and evoke the NWR. The experiments on a full-size humanoid robot show that the method enables robots to avoid potential damage robustly with proper NWR, depending on the site and intensity of the stimulation. We also verify that the method takes advantage of the intrinsic characteristics of its neuromorphic encoding scheme to reproduce essential features of the NWR, e.g., spatial summation effect and temporal summation effect in humans. The improved bio-plausibility and the capability to reproduce the human-like feature.

A survey on control of humanoid fall over

Humanoid robot operation requires balancing to prevent failures, such as fall over. This is a crucial task in legged robots and thus several researchers are working on this topic. Fall prediction, controlled fall, and fall recovery become important topics in understanding robot control and allow legged robots to function in challenging real-world environments. This paper aims at setting up methodically the problem definition of humanoid falling and further identifying and surveying working techniques in the literature. The focus is to categorize all methods that were used in the community, identify the solved and open questions, as well as propose directions of research in the field. The paper is based on experimental research that has been done on a full-size humanoid robot.

Preemptive Foot Compliance to Lower Impact During Biped Robot Walking Over Unknown Terrain

In this work, we present a novel method for ankle/foot compliance for biped humanoid robots walking over uneven terrain. Based on distributed plantar proximity sensing, we developed the Preemptive Foot Compliance (PFC) control that generates a Preemptive Ground Reaction Wrench that modifies the foot orientation to maximize the largest contact area to reduce the impact force produced at landing. PFC can be easily included in any walking controller for flat ground and become capable of walking on uneven terrains. The PFC was tested on two full-size humanoid robots running two different walking controllers, originally designed only for flat-ground walking. Both robots increase their capability to walk over uneven terrain and the walking impacts in flat grown were reduced by approximately 80%.

Carved Turn Control with Gate Vision Recognition of a Humanoid Robot for Giant Slalom Skiing on Ski Slopes.

The performance of humanoid robots is improving, owing in part to their participation in robot games such as the DARPA Robotics Challenge. Along with the 2018 Winter Olympics in Pyeongchang, a Skiing Robot Competition was held in which humanoid robots participated autonomously in a giant slalom alpine skiing competition. The robots were required to transit through many red or blue gates on the ski slope to reach the finish line. The course was relatively short at 100 m long and had an intermediate-level rating. A 1.23 m tall humanoid ski robot, ‘DIANA’, was developed for this skiing competition. As a humanoid robot that mimics humans, the goal was to descend the slope as fast as possible, so the robot was developed to perform a carved turn motion. The carved turn was difficult to balance compared to other turn methods. Therefore, ZMP control, which could secure the posture stability of the biped robot, was applied. Since skiing takes place outdoors, it was necessary to ensure recognition of the flags in various weather conditions. This was ensured using deep learning-based vision recognition. Thus, the performance of the humanoid robot DIANA was established using the carved turn in an experiment on an actual ski slope. The ultimate vision for humanoid robots is for them to naturally blend into human society and provide necessary services to people. Previously, there was no way for a full-sized humanoid robot to move on a snowy mountain. In this study, a humanoid robot that transcends this limitation was realized.

Neural Policy Style Transfer

Style Transfer has been proposed in a number of fields: fine arts, natural language processing, and fixed trajectories. We scale this concept up to control policies within a Deep Reinforcement Learning infrastructure. Each network is trained to maximize the expected reward, which typically encodes the goal of an action, and can be described as the content. The expressive power of deep neural networks enables encoding a secondary task, which can be described as the style. The Neural Policy Style Transfer (NPST)11NPST: Neural Policy Style Transfer. algorithm is proposed to transfer the style of one policy to another, while maintaining the content of the latter. Different policies are defined via Deep Q-Network architectures. These models are trained using demonstrations through Inverse Reinforcement Learning. Two different sets of user demonstrations are performed, one for content and other for style. Different styles are encoded as defined by user demonstrations. The generated policy is the result of feeding a content policy and a style policy to the NPST algorithm. Experiments are performed in a catch-ball game inspired by the Deep Reinforcement Learning classical Atari games; and a real-world painting scenario with a full-sized humanoid robot, based on previous works of the authors. The implementation of three different Q-Network architectures (Shallow, Deep and Deep Recurrent Q-Network) to encode the policies within the NPST framework is proposed and the results obtained in the experiments with each of these architectures compared.

Enabling garment-agnostic laundry tasks for a Robot Household Companion

Domestic chores, such as laundry tasks, are dull and repetitive. These tasks consume a significant amount of daily time, and are however unavoidable. Additionally, a great portion of elder and disabled people require help to perform them due to lack of mobility. In this work we present advances towards a Robot Household Companion (RHC), focusing on the performance of two particular laundry tasks: unfolding and ironing garments. Unfolding is required to recognize the garment prior to any later folding operation. For unfolding, we apply an interactive algorithm based on the analysis of a colored 3D reconstruction of the garment. Regions are clustered based on height, and a bumpiness value is computed to determine the most suitable pick and place points to unfold the overlapping region. For ironing, a custom Wrinkleness Local Descriptor (WiLD) descriptor is applied to a 3D reconstruction to find the most significant wrinkles in the garment. These wrinkles are then ironed using an iterative path-following control algorithm that regulates the amount of pressure exerted on the garment. Both algorithms focus on the feasibility of a physical implementation in real unmodified environments. A set of experiments to validate the algorithms have been performed using a full-sized humanoid robot.

Vision-Based 3D Modeling of Unknown Dynamic Environments for Real-Time Humanoid Navigation

In order to achieve real autonomy, robots have to be able to navigate in completely unknown environments. Due to the complexity of computer vision algorithms, almost every approach for robotic navigation is either based on previous knowledge of the environment, such as markers or as resulting from learning methods, or makes strong simplifying assumptions about it (height-map representations, static scenarios). While showing impressive success in certain applications, these approaches limit the potential of legged robots to achieve the amazing flexibility of humans in more complex environments. In this work, we present a strategy for full 3D vision processing that is able to handle changing, dynamic environments. These are modeled using 3D geometries that are processed in real-time by the motion planner of our biped robot Lola for avoiding moving obstacles and walking over platforms. In order to allow for a more intuitive development of such systems in the future, we present tools for visualization including two mixed reality applications using both an external camera and Microsoft’s HoloLens. We validate our system in simulations and experiments with our full-size humanoid robot Lola and publish our framework open source for the benefit of the community.

Intelligent control of humanoid manipulation with uncertain weight objects

This article addresses a new intelligent full-body balancing control framework on a full-sized humanoid robot for stably tracking a given reference while grasping onto objects with unknown weight information. Dealing with external uncertainty from the inherent stability of humanoid is an interesting challenge for developing stable manipulation performance during their operation. To this end, intelligent fuzzy-proportional–integral control scheme where parameters are adaptively adjusted to respond to those uncertainties is proposed in this study. Application of momentum information at the torso as additional information is to make the improvement of the manipulation capability on a humanoid over an unknown weight object. Results from both experiment and simulation over a range of unknown weight objects on humanoid hand verified that the proposed scheme is more representative of the intelligent control on humanoid operation with external uncertainties.

Effects of rotation and systematic occlusion on fiducial marker recognition

Fiducial marker systems consist of patterns that are placed in environment for miscellaneous applications and are further automatically detected with cameras. A variety of applications determines the criteria, which characterize qualitative properties of a marker and include such evaluation benchmarks as resilience to occlusion, distance to a marker, false positive and false negative rates, sensitivity to illumination, and others. The paper compares existing ARTag, AprilTag, and CALTag systems utilizing a high fidelity camera, which is a main vision sensor of a full-size Russian humanoid robot AR-601M. In experiments the comparison of the three marker systems reliability and detection rate in occlusions of various types and intensities was verified. Finally, a preferable for AR-601M robot visual applications marker system was selected.

Walking Algorithm of Humanoid Robot on Uneven Terrain with Terrain Estimation

Humanoid robots are expected to achieve stable walking on uneven terrains. In this paper, a control algorithm for humanoid robots walking on previously unknown terrains with terrain estimation is proposed, which requires only minimum modification to the original walking gait. The swing foot trajectory is redesigned to ensure that the foot lands at the desired horizontal positions under various terrain height. A compliant terrain adaptation method is applied to the landing foot to achieve a firm contact with the ground. Then a terrain estimation method that takes into account the deformations of the linkages is applied, providing the target for the following correction and adjustment. The algorithm was validated through walking experiments on uneven terrains with the full-size humanoid robot Kong.

Biologically inspired design of a parallel actuated humanoid robot

This paper presents the comparison for the role of bi-articular and mono-articular actuators in human and bipedal robot legs, in particular the hip and knee joint, for driving the design of a humanoid robot with inspirations from the biological system. The various constraints driving the design of both systems are also compared. Additional factors particular to robotic system are identified and incorporated in the design process. To do this, a dynamic simulation is used to determine loading conditions and the forces and power produced by each actuator under various arrangements. It is shown that while the design principles of humans and humanoids are similar, other constraints ensure that robots are still merely inspired by humans, and not direct copies. A simple design methodology that captures the complexity and constraints of such a system in this paper is proposed. Finally, a full-size humanoid robot that demonstrates the newfound principle is highlighted.

Realizing whole-body tactile interactions with a self-organizing, multi-modal artificial skin on a humanoid robot

In this paper, we present a new approach to realize whole-body tactile interactions with a self-organizing, multi-modal artificial skin on a humanoid robot. We, therefore, equipped the whole upper body of the humanoid HRP-2 with various patches of CellulARSkin – a modular artificial skin. In order to automatically handle a potentially high number of tactile sensor cells and motors units, the robot uses open-loop exploration motions, and distributed accelerometers in the artificial skin cells, to acquire its self-centered sensory-motor knowledge. This body self-knowledge is then utilized to transfer multi-modal tactile stimulations into reactive body motions. Tactile events provide feedback on changes of contact on the whole-body surface. We demonstrate the feasibility of our approach on a humanoid, here HRP-2, grasping large and unknown objects only via tactile feedback. Kinesthetically taught grasping trajectories, are reactively adapted to the size and stiffness of different test objects. Our paper contributes the first realization of a self-organizing tactile sensor-behavior mapping on a full-sized humanoid robot, enabling a position controlled robot to compliantly handle objects.

Autonomous navigation of a humanoid robot over unknown rough terrain using a laser range sensor

The present paper describes the integration of laser-based perception, footstep planning, and walking control of a humanoid robot for navigation over previously unknown rough terrain. A perception system that obtains the shape of the surrounding environment to an accuracy of a few centimeters is realized based on input obtained using a scanning laser range sensor. A footstep planner decides the sequence of stepping positions using the obtained terrain shape. A walking controller that can cope with a few centimeters of error in terrain shape measurement is achieved by combining the generation of a 40-ms cycle online walking pattern and a ground reaction force controller with sensor feedback. An operational interface was developed to send commands to the robot. A mixed-reality display was adopted to realize an intuitive interface. The navigation system was implemented on the HRP-2, a full-size humanoid robot. The performance of the proposed system for navigation over unknown rough terrain and the accuracy of the terrain shape measurement were investigated through several experiments.

TEO: FULL-SIZE HUMANOID ROBOT DESIGN POWERED BY A FUEL CELL SYSTEM

This article deals with the design of the full-size humanoid robot TEO, an improved version of its predecessor Rh-1. The whole platform is conceived under the premise of high efficiency in terms of energy consumption and optimization. We will focus mainly on the electromechanical structure of the lower part of the prototype, which is the main component demanding energy during motion. The dimensions and weight of the robotic platform, together with its link configuration and rigidity, will be optimized. Experimental results are presented to show the validity of the design.

New 3-DOFs Hybrid Mechanism for Ankle and Wrist of Humanoid Robot: Modeling, Simulation, and Experiments

This paper deals with the design of a new class of hybrid mechanism dedicated to humanoid robotics application. Since the designing and control of humanoid robots are still open questions, we propose the use of a new class of mechanisms in order to face several challenges that are mainly the compactness and the high power to mass ratio. Human ankle and wrist joints can be considered more compact with the highest power capacity and the lowest weight. The very important role played by these joints during locomotion or manipulation tasks makes their design and control essential to achieve a robust full size humanoid robot. The analysis of all existing humanoid robots shows that classical solutions (serial or parallel) leading to bulky and heavy structures are usually used. To face these drawbacks and get a slender humanoid robot, a novel three degrees of freedom hybrid mechanism achieved with serial and parallel substructures with a minimal number of moving parts is proposed. This hybrid mechanism that is able to achieve pitch, yaw, and roll movements can be actuated either hydraulically or electrically. For the parallel submechanism, the power transmission is achieved, thanks to cables, which allow the alignment of actuators along the shin or the forearm main axes. Hence, the proposed solution fulfills the requirements induced by both geometrical, power transmission, and biomechanics (range of motion) constraints. All stages including kinematic modeling, mechanical design, and experimentation using the HYDROïD humanoid robot’s ankle mechanism are given in order to demonstrate the novelty and the efficiency of the proposed solution.

A humanoid robot capable of carrying heavy objects

SUMMARYA new type of 28-DOF (degree of freedom) full-size humanoid robot, driven by a closed-chain type of joint actuation system, is developed in this paper. Each leg of the robot is composed of six joints, where three are at the hip, one is at the knee, and two are at the ankle. The robot has six joints for each arm, one balancing joint, and three joints for the head, with two cameras. The weight of the robot is 75 kg, and its height is 168 cm.The actuation systems of the pitching joint for the arms and legs of the robot are designed based on a closed-chain mechanism composed of four bar links driven by a ball screw, and each leg of the robot is designed to support 95 kg weight to include a 20 kg payload that can be carried by the robot arms having very light designs (with weight 8.5 kg), but each capable of carrying a 10 kg payload.An analysis of the closed-chain joint actuation systems of a light arm capable of handling heavy objects is performed, and the light arm is designed via finite-element method analysis performed using ANSYS. In addition, the kinematic analysis and the detailed structure of the arm and leg of the robot are performed.The main controller uses the ARM processor and a distributed controller for the leg joints is developed using the TMS320c2407 processor with the communications between the main and joint controllers being performed via the CAN system.Good performances of the proposed robot is demonstrated by presenting several experimental results; these include (1) experimentally handling a 13 kg payload, (2) through walking experiments of the robot supporting a 85 kg load, and (3) measurements of the arm and leg joint motors while performing walking experiments.

System Design and Control of Anthropomorphic Walking Robot LOLA

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> This paper presents the 25-DOF full-size humanoid robot <emphasis emphasistype="italic">LOLA</emphasis> . Our goal is to realize fast and human-like walking. Furthermore, we want to increase the robot’s autonomous, vision-guided walking capabilities. <emphasis emphasistype="italic">LOLA</emphasis> is characterized by a redundant kinematic configuration, an extremely lightweight design, joint actuators with brushless motors and an electronics architecture using decentralized joint control. Special emphasis was put on an improved mass distribution to achieve good dynamic performance. Center of mass trajectories are calculated in real-time from footstep locations using a spline collocation method. Reference trajectories are modified by a stabilizing control system based on hybrid force/position control with an inner joint position control loop. </para>

REAL-TIME GAIT PLANNING FOR THE HUMANOID ROBOT Rh-1 USING THE LOCAL AXIS GAIT ALGORITHM

This paper presents the 3D foot and center of gravity motion planning for the humanoid robot called the "local axis gait" (LAG) algorithm. It permits walking on different kinds of surfaces, such as planes, ramps or stairs. Furthermore, continuous change of the step length and orientation in real time will be possible, due to the real-time linear dynamics model of the walking pattern of the humanoid. The robot model is based on the cart table formulation for planning the center of gravity (COG) and zero moment point (ZMP) motion. The proposed algorithm takes into account physical robot constraints such as joint angles, angular velocity and torques. Torques are computed by the Lagrange method under screws and Lie groups. The LAG is divided into several stages: computation of the footprints; the decision of the ZMP limits around the footprints; the dynamic humanoid COG motion generation based on the cart table model; and joining the footprints of the swing foot by splines. In this way it is possible to generate each step online, using the desired footprints as input. In order to compute the joint torque limits, the Lagrangian method is used under the Lie groups and screw theory. The paper presents and discusses some successful results on the LAG in the full-size humanoid robot Rh-1 developed in the Roboticslab of University Carlos III of Madrid.

The Rh-1 full-size humanoid robot: Design, walking pattern generation and control

This paper is an overview of the humanoid robot Rh-1, the second phase of the Rh project, which was launched by the Robotics Lab at the Carlos III University of Madrid in 2002. The robot mechanical design includes the specifications development in order to construct a platform, which is capable of stable biped walking. At first, the robots' weights were calculated in order to obtain the inverse dynamics and to select the actuators. After that, mechanical specifications were introduced in order to verify the robot's structural behaviour with different experimental gaits. In addition, an important aspect is the joints design when their axes are crossed, which is called 'Joints of Rectangular Axes' JRA. The problem with these joints is obtaining two or more degrees of freedom DOF in small space. The construction of a humanoid robot also includes the design of hardware and software architectures. The main advantage of the proposed hardware and software architectures is the use of standardised solutions frequently used in the automation industry and commercially available hardware components. It provides scalability, modularity and application of standardised interfaces and brings the design of the complex control system of the humanoid robot out of a closed laboratory to industry. Stable walking is the most essential ability for the humanoid robot. The three dimensional Linear Inverted Pendulum Model 3D-LIPM and the Cart-table models had been used in order to achieve natural and dynamic biped walking. Humanoid dynamics is widely simplified by concentrating its mass in the centre of gravity COG and moving it following the natural inverted pendulum laws 3D-LIPM or by controlling the cart motion Cart-table model. An offline-calculated motion pattern does not guarantee the walking stability of the humanoid robot. Control architecture for the dynamic humanoid robot walking was developed, which is able to make online modifications of the motion patterns in order to adjust it to the continuously changing environment. Experimental results concerning biped locomotion of the Rh-1 humanoid robot are presented and discussed.

Modular Architecture for Humanoid Walking Pattern Prototyping and Experiments

In this paper we describe the use of design patterns as a basis for creating humanoid walking pattern generator software having a modular architecture. This architecture enabled the rapid porting of several novel walking algorithms on a full-size humanoid robot, HRP-2. The body of work currently available allows extracting a general software architecture usable with inter-exchange between simulations and real experiments. The proposed architecture with the associated design patterns is described together with several applications: a pattern generator for a HRP-2 with passive toe joints, a pattern for dynamically stepping over large obstacles and a new quadratic problem (QP) formulation for the generation of the reference zero-momentum point. Thanks to the versatility and the modularity of the proposed framework, the QP method has been implemented and experienced within 4 days only.