Vincent Hugel

Towards efficient implementation of quadruped gaits with duty factor of 0.75

Deals with the implementation of an efficient walking pattern for quadruped robots. The main objective is to give the robot the ability to walk in every direction as quickly as possible. Not only forward motion, but also backward motion, turning gaits of variable curvature including rotation around the body platform center of gravity (COG) are considered separately. For this purpose the well known crawl gait with duty factor of 3/4 represents the starting point of the experimental study. From this point the regular symmetric gait is improved thanks to sideways motion, turning gaits are adapted from the forward crawl gait and a special rotation motion is designed. All transitions between forward, backward, left and right turning and rotation motions are detailed. Experiments have been carried out using the Sony quadruped pet robot prototype which look part in the Paris Robocup-98 legged robot exhibition.

Bird terrestrial locomotion as revealed by 3D kinematics.

Most birds use at least two modes of locomotion: flying and walking (terrestrial locomotion). Whereas the wings and tail are used for flying, the legs are mainly used for walking. The role of other body segments remains, however, poorly understood. In this study, we examine the kinematics of the head, the trunk, and the legs during terrestrial locomotion in the quail (Coturnix coturnix). Despite the trunk representing about 70% of the total body mass, its function in locomotion has received little scientific interest to date. This prompted us to focus on its role in terrestrial locomotion. We used high-speed video fluoroscopic recordings of quails walking at voluntary speeds on a trackway. Dorso-ventral and lateral views of the motion of the skeletal elements were recorded successively and reconstructed in three dimensions using a novel method based on the temporal synchronisation of both views. An analysis of the trajectories of the body parts and their coordination showed that the trunk plays an important role during walking. Moreover, two sub-systems participate in the gait kinematics: (i) the integrated 3D motion of the trunk and thighs allows for the adjustment of the path of the centre of mass; (ii) the motion of distal limbs transforms the alternating forward motion of the feet into a continuous forward motion at the knee and thus assures propulsion. Finally, head bobbing appears qualitatively synchronised to the movements of the trunk. An important role for the thigh muscles in generating the 3D motion of the trunk is suggested by an analysis of the pelvic anatomy.

Qualitative Adaptive Reward Learning With Success Failure Maps: Applied to Humanoid Robot Walking

In the human brain, rewards are encoded in a flexible and adaptive way after each novel stimulus. Neurons of the orbitofrontal cortex are the key reward structure of the brain. Neurobiological studies show that the anterior cingulate cortex of the brain is primarily responsible for avoiding repeated mistakes. According to vigilance threshold, which denotes the tolerance to risks, we can differentiate between a learning mechanism that takes risks and one that averts risks. The tolerance to risk plays an important role in such a learning mechanism. Results have shown the differences in learning capacity between risk-taking and risk-avert behaviors. These neurological properties provide promising inspirations for robot learning based on rewards. In this paper, we propose a learning mechanism that is able to learn from negative and positive feedback with reward coding adaptively. It is composed of two phases: evaluation and decision making. In the evaluation phase, we use a Kohonen self-organizing map technique to represent success and failure. Decision making is based on an early warning mechanism that enables avoiding repeating past mistakes. The behavior to risk is modulated in order to gain experiences for success and for failure. Success map is learned with adaptive reward that qualifies the learned task in order to optimize the efficiency. Our approach is presented with an implementation on the NAO humanoid robot, controlled by a bioinspired neural controller based on a central pattern generator. The learning system adapts the oscillation frequency and the motor neuron gain in pitch and roll in order to walk on flat and sloped terrain, and to switch between them.

Walking, paddling, waddling: 3D kinematics anatidae locomotion (Callonetta leucophrys).

Walking and paddling motions were studied in a semiaquatic bird, the ringed teal (Callonetta leucophrys), to investigate the motions associated with movements in two environments with radically divergent physical properties. A three-dimensional (3D) kinematic reconstruction based on nonsynchronous biplanar cineradiographic data was used to quantify the 3D trajectories of the body and hind limb segments. Our study revealed that two subsystems interact to provide propulsion in water and on land. During paddling, the trunk, the femur, and the tibiotarsus are in a stable position and play the role of the hull. The femur and tibiotarsus are positioned laterally and parasagittaly and the intertarsal joint is fixed and positioned caudally allowing large amplitude movements of the paddle (tarsometatarsus and palmate foot). During walking, the center of mass is held above the medially oriented foot, providing stability during the single support phase. During stance, the foot is medially oriented because of the lateral and parasagittal positions of the tibiotarsus and tarsometatarsus during both walking and paddling. This position of the foot during walking imposes trunk translation and results in the typical waddling motion of Anatidae. This study provides new insights into how waddling motion relates to semiaquatic birds ability to move in both terrestrial and aquatic environments.

Kinematic Modeling of Bird Locomotion from Experimental Data

We present the design of a bird-like kinematics model for a biped robot as an alternative to the human model. The starting point of the research consists of analyzing the walking motion of quail birds using biological data obtained by X-ray radiography. The 3-D-motion analysis allows identification of the number of degrees of freedom (DOF) and the rotation mechanism for each leg, especially the main rotation axis. Leg joints are located at the hip, knee, ankle, and foot. The ankle is off the ground. Using this analysis, we have designed a biped kinematics model with a minimum of actuated joints and with the original orientation of hip and ankle main rotary joints, which are not horizontally and vertically oriented as in classical biped robotics. Given a reference-foot trajectory, we carry out simulations to compare internal-joint trajectories with the ones obtained from biological measurements. We show that the proposed model can be used to reproduce the kinematics of bird locomotion with a minimum of four actuated joints per leg, i.e., two at the hip and one at the knee, and one at the ankle, which is less than the usual six joints per leg that drive anthropomorphic legs.

Pivoting Manipulation of a Large Object: A Study of Application using Humanoid Platform

Pivoting manipulation can be an alternative to pushing operation when the floor is not flat enough, or when the object to manipulate is too heavy. The technique of pivoting is used by humans to move large and bulky furniture from one place to another. The decomposition of the task of pivoting has already been studied, in particular with the use of two fingers of a robotic arm. This work intends to apply the technique of pivoting using an humanoid platform. The robot should be able to pivot the object and to walk with it to displace it to a specific remote location. The research achievements proposed here take place in a more long term objective aimed at improving the dexterity and the autonomy of humanoid robots. As a matter of fact, such robots should be able to handle objects and move around in the environment in an autonomous way. This paper describes the algorithm designed to perform the displacement of a large object using the pivoting technique. It also presents the results of the dynamic simulation and the results of the real hardware experiment of the HRP-2 platform performing the task.

Humanoids learn object properties from robust tactile feature descriptors via multi-modal artificial skin

This paper presents new methods for the recognition and categorization of object properties such as surface texture, weight, and compliance using a multi-modal artificial skin mounted on both arms of a humanoid. In addition, it introduces two novel feature descriptors, which are useful for providing high-level information to learning algorithms. The artificial skin has built-in 3-axis accelerometer, normal force, proximity, and temperature sensors. To explore different surface textures and weights, objects were left sliding between the NAO humanoids arms. The caused vibration was detected by accelerometers. Surface texture and weight recognition models were learned from the extracted features of the vibration signals thanks to two learning algorithms, namely the support vector machine (SVM) and the Expectation Maximization (EM). In order to recognize objects having different compliances, SVM and EM took into account total amount of forces applied by the arms to hold the object firmly. The experimental results show that the humanoid can distinguish between different objects having different surface textures and weights with a recognition rate of 100%. Furthermore, it can categorize objects with hard and soft surfaces and classify objects having similar compliance with 100% and 70% accuracy rates respectively.

Motion planning for whole body tasks by humanoid robot

In this paper we present motion planning for tasks that require whole body motion of a humanoid robot. We address two such typical tasks, stepping over obstacles and manipulating an object, with the help of resolved momentum control (RMC) to guarantee the robot stability. For the first task, we plan the trajectories of the feet and the waist according to obstacle size. The motion of upper body is determined using RMC to keep balance. This novel planning method is adaptive to obstacle sizes and hence oriented to autonomous stepping-over of humanoid robots guided by vision. The second task is pivoting manipulation of a large object, which is suitable for precise and stable transportation of large objects. To perform this manipulation, first the object is moved by two arms using force and position control together with body balancing control while robot stays in the same place. Next the movement of the humanoid robot itself is performed by stepping using RMC. The effectiveness of proposed motion planning methods is verified by experiments and simulations.

Influence of rotation of humanoid hip joint axes on joint power during locomotion

Most of the fully actuated legs of existing humanoids have a straight kinematics, i.e. roll, pitch and yaw joint rotary axes are aligned, respectively, along longitudinal, lateral and vertical axes of the robot’s body. This work explores different orientations of the hip joint axes where the yaw joint and the pitch joint axes are rotated about the longitudinal axis, keeping the serial order – yaw, roll and pitch – from the pelvis. Two walking movements are simulated dynamically: forward walk and turn-in-place. The measurements of joint mechanical power (JMP) consumption are used to compare the different kinematics configurations. The results obtained from simulation show that the inward and outward rotations of hip yaw and pitch joint axes about the longitudinal axis bring a significant reduction of pitch JMP, and lead to a better JMP distribution among the hip joint actuators compared with the classical straight configuration. In addition, the outward configuration presents a significant reduction of JMP consumption of the hip pitch joint of the supporting leg during acceleration and regular steps, compared with the inward configuration. Graphical Abstract

Influence of the number of humanoid vertebral column pitch joints in flexion movements

This paper deals with 2D simulations of a humanoid robot equipped with back bone pitch joints to study the advantages of having such a mechanism for daily human-like movements. The movements under investigation here involve knee flexion for sitting down on a chair or picking up objects on the floor. The model used for the humanoid robot is based on a kid-sized human body. The trunk is decomposed into a thorax and a lumbar part. As the lumbar region is the most mobile part in the human vertebral column, vertebrae are only placed in the robots lumbar part. Simulations are carried out in the sagittal plane to investigate the influence of the number of vertebra pitch joint on the movements. Results show that a number of two pitch joints is a good tradeoff in matter of work at hip and thorax inclination.