Romansy 19 – Robot Design, Dynamics and Control

Wire-driven parallel robot (WDPR) is a special class of parallel robot in which the rigid legs are replaced by wires, with potential advantages in terms of intrusivity and workspace. Although the study of WDPR seems to be a well-addressed subject, we will show that there are still numerous challenging open issues in this field.

A translational parallel manipulator with three degrees of freedom and three kinematic chains is considered. Each kinematic chain contains five revolute joints. Kinematics, workspace, singularities and dynamics of the proposed mechanism are discussed.

A

N

−1 wire-driven parallel robot is a robot for which all the

N

≥3 wires are connected at the same point of the platform, allowing to control the location of this point. We are interested in the positioning accuracy of such a robot. If the wires are not elastic we show that the influence on the accuracy of the co-location errors of the wire anchor points on the platform is moderate, although a full analysis is a very difficult task. If the wires are elastic we study the influence of the the wire lengths measurement errors and inaccurate estimation of the stiffness of the wires. Again we show a moderate influence but very large changes in the tensions in the wires that probably prohibit the use of the redundancy to optimize the tension in the wires. In all cases the complexity of the forward kinematics of such a robot makes accuracy analysis a very demanding task that requires an in-depth investigation.

This paper deals with the analysis of a spherical parallel manipulator (3RCC) to determine the error on the pose of the end effector as a function of the manufacturing errors of the different links and the presence of a clearance in the joints. The obtained model allowed us to identify the error on the platform in three cases, i.e., only manufacturing errors were considered, then only clearance in the joints was considered and finally the case of both sources of error were present in the system. It was shown, in particular, that the axial displacement in the C joints is quite important. The second result is the fact that the superposition principle does not work when we consider both the manufacturing errors and the clearance despite the assumption of small displacements.

The object of this article is to obtain in a systematic form all the families of parallel manipulators with 3 degrees of freedom considering solely fully parallel manipulators whose legs are structurally identical. The motion pattern of the end-effector that should be taken into account are: 3T, 2T1R, 1T2R and 3R, where T and R refer to the character of translational and rotational degrees of freedom.The methodology that follows utilizes the concepts of the theory of groups of displacements applied to the structural synthesis of parallel manipulators.

Massive parallel arrays of discrete actuators are forceregulated robots that undergo continuous motions despite being commanded through a large but finite number of states only. Realtime control of such systems requires fast and efficient methods for solving their inverse static analysis, which is a challenging problem. Artificial intelligence methods are investigated here for the on-line computation of the inverse static analysis of a planar parallel array featuring eight three-state force actuators and possessing one degree of revolute motion.

The design of shaking-force balanced linkages can be approached by deriving these linkages from balanced linkage architectures. When desired, a possible step is to add degrees-of-freedom (dof), for instance by substituting a link with a n-dof equivalent linkage for which the balanced design of the other links is not affected. This paper shows how the coupler link of a shaking-force balanced 4R four-bar linkage, applied as a 5R five-bar linkage, can be substituted with an equivalent 2-dof pantograph.

The development of robotic systems has faced many challenges: First came what Freudenstein called “the Mount Everest” of kinematics. Thereafter came the challenge of finding all forwardkinematics solutions of a six-degree-of-freedom parallel robot. The two foregoing problems are largely solved by now. The new challenge is the development of ever faster pick-and-place four-degree-of-freedom robots. The limit of the serial version thereof was reached in the late nineties, with a record speed of two cycles per second. This called for the development of parallel versions of the same. Some industrial robots of this kind, carrying three to four limbs, are out in the market. With the purpose of simplifying their morphology and reducing their footprint, two-limb robots have started emerging. The challenge here is the transmission of force and motion from the two actuators of each limb, mounted on a common base, to produce two independent motions, normally pan and tilt. Discussed in this paper are the theoretical and practical hurdles that the robot designer faces in this quest.

Solar tracking devices try to increase the area exposed to direct radiation of the sun. The main drawback of these kind of devices is the energy consumed when following the sun. This work presents the design of a mechanism able to accurately follow the sun motion minimizing the energy consumption during its operation. The objective is achieved by means of a mechanism with parallel kinematics architecture.

An identification of the model parameters for a parallel cable-driven robot (8 cables for 6 degrees of freedom) is performed by using both a calibration and a self-calibration approach. Additionally, advanced tools and algorithmic improvements are presented to perform the parameter identification. A complete experimentation validates the robot accuracy improvement.

Most of the papers dealing with the dynamic parameters identification of parallel robots are based on simple models, which take only the dynamics of the moving platform into account. Moreover the actuator drive gains are not calibrated which leads to identification errors. In this paper a systematic way to derive the full dynamic identification model of parallel robots is proposed in combination with a method that allows the identification of both robot inertial parameters and drive gains.

Usually, identification models of parallel robots are simplified and take only the dynamics of the moving platform into account. Moreover the input efforts are estimated through the use of the manfucaturer’s actuator drive gains that are not calibrated thus leading to identification errors. In this paper a systematic way to derive the full dynamic identification model of the Orthoglide parallel robot in combination with a method that allows the identification of both robot inertial parameters and drive gains.

Parallel robots can be modelled as serial robots after cuttingopen the robot at some of its joints. Internal forces appear at joints which have been cut open. In this paper, the coordinate partitioning method and the velocity projection method are employed to make the internal forces disappear such that a dynamic parameter identification model can be obtained in terms of generalized independent coordinates. The influence of these methods for obtaining the identification model when exciting trajectories are used for experiments is investigated. In addition, the effect of the parameter identification solution is also considered.

This paper deals with the optimal dynamics of the 3-RRR decoupled robot in which the linear displacements of the platform are decoupled from its orientation. The particularity of the 3-RRR decoupled robot is that an optimal choice of the leg’s geometric parameters allows an unlimited platform rotation, which can be attractive for many industrial applications. However, a singular configuration is necessarily encountered during full platform rotation. In order to generate a stable motion in the presence of singularities, optimal dynamic conditions are disclosed. The suggested optimal conditions ensuring unlimited platform rotation are illustrated and validated by numerical simulations with ADAMS.

This paper presents a new approach for gravity compensation of a 3-UPU parallel manipulator. The conventional method of computing the effect of gravity force on the end-effector works properly for serial manipulators. However, employing Newton-Euler approach for the parallel robots is computationally expensive and it cannot satisfy the requirements in this work. In order to overcome this difficulty, the new algorithm based on Lagrengian method is proposed. This model is established based on total potential energy of the system as a scalar value and, the position of the end-effector. This paper presents this new algorithm which is more efficient in sense of computation and more proper for Real-Time purposes in parallel robots.

A portable motion replication system for posture guidance of upper extremity is presented in this paper. The system can capture user’s postures, compare the postures with corresponding ones in database, and give the user proper visual and verbal feedback for posture modifying in real time. The experiment of user test shows that the subjects under both visual and verbal feedback have better performance.

In this article a novel actuation concept of robotic manipulators is presented. The concept is based on an innovative extension of gravity balancing techniques using passive elements. It allows to exert, at the level of the end effector, a force of variable intensity and generic spatial orientation without any motor torque required during any movement of the manipulator, except when a change of the exerted force’s intensity and/or orientation is requested. So, a new hybrid machine can be realized combining the benefits of active and passive existing systems without their drawbacks, with possible very high potentials in robotics applications.

The authors discuss the design and implementation of a precision flexure mechanism for a non-contact bio-cellular experimental apparatus called magnetic tweezer (MT). MT manipulates a magnetic particle of several-micron diameter by magnetomotive force, diverged from magnetic poles to stimulate a target cell to measure its mechanical property. The developed flexure stage was designed to satisfy a strict size limitation to be assembled in a conventional microscope, and to make MT a practical experimental apparatus. The design concept and series of analysis were discussed to satisfy practical requirements on working range and mechanical stiffness. Motion control experiment using a 3D-visual servoing was performed to evaluate the control accuracy and to demonstrate validity and effectiveness of the developed system.

This paper describes the development of a new expressive robotic head for bipedal humanoid robot. Through a preliminary experiment, the authors defined the representative facial expressions for 6 basic emotions. To realize these facial expressions, 24-DoFs of mechanisms that allow wide moveable range, and facial color were needed on the face. We designed compact mechanisms that fit into the head of which major dimensions are based on average of adult Japanese female’s size. We conducted a questionnaire surveys to evaluate the facial expression recognition rates. The results show that facial expression recognition rates for 6 basic emotions were increased compared to the old head.

This paper presents a novel design of a single degree of freedom planar 12-link mechanism for a finger exoskeleton. The mechanism is sized to each finger of the human hand and attached to the phalanges to control the flexion/extension movements while generating the finger desired grasping trajectory.

Several novel interaction peripherals recently came to the market. Tactile interfaces and mocap systems deeply changed the way we interact with computers. On the other side, the continuous fall in prices of VR technologies and the development of force feedback haptic interfaces changed the way people develop new products, moving from real to digital mock-ups. Some limitations remain however as no commercially available interface allows natural dexterous interactions in 3D space. Tactile devices are limited to 2D. Data-gloves are difficult to calibrate and do not give any feedback to the user, the later being also true for mocap systems. Haptic devices are equiped with handles and limit the user’s dexterity. Multi-finger haptic interfaces are required to overcome these limitations. This paper makes a short review of such devices, gives more details on a two and a three fingers interfaces recently developed for VR applications and gives guidelines for their design.

In the future, we will need service robots capable of natural interaction with people, which often involves the use of hands. In this paper, we aim to evaluate the impression of the interaction between a human and a humanoid robot with anthropomorphic hands. In particular, we focus on the influence of the grip of robot’s hands during handshaking. First, we measured pressured places and forces. Based on this result, we selected motors that have enough torque for handshaking. Finally, we developed a new anthropomorphic soft robot hand named WSH-1RII with the soft material structure, enough grip strength, and force sensing. The experimental results confirmed that the hand gives the users a good impression.

This paper describes the development of a distributed control system and a small motor controller unit modularized to a controller module and motor driver modules for a humanoid robot. The humanoid robot, KOBIAN, had a centralized control system. It was difficult to increase the number of motors because of PC’s I/O channel limit. We installed the distributed control system in the new head to increase the number of motors. Moreover, the modularized motor controller is developed to use in the various parts of the body. The small flexible motor controller is composed of a controller module and motor driver modules. Therefore, changing motor driver modules allows control of various components with the same controller module. This small motor controller unit can control 4 DC motors and read 8 channel of analog sensory data. We evaluated the performance of the developed motor controller and controlled the new robot’s head.

In this paper two approaches for the correct task execution during null-space impedance control of a kinematically redundant robot are presented. The algorithms guarantee safe and dependable reaction of the robot during deliberate or accidental interaction with the environment, thanks to null-space impedance control. Moreover, the correct execution of the task assigned to the end-effector is ensured by control laws relying on two different observers. One is based on task space information and the other on the generalized momentum of the robot. The performance of the proposed control is verified through numerical simulations on 7R KUKA lightweight robot arm.

Methodologies for investigating the effect of wire/actuator failures on the wrench capability of wire-actuated parallel manipulators are presented that result in minimum norm solution for the correctional wire tension and overall tension vectors. A planar manipulator is simulated to illustrate the proposed methodologies.

In this paper, the wrench recovery of parallel manipulators after actuator failure is investigated. To achieve the desired wrench capability, the mobile platform task is divided into the recoverable and non-recoverable subtasks. The presented work is based on the projection of the lost wrench due to total or partial actuator failure onto the range space of the reduced Jacobian matrix. The force/torque of the healthy joints is adjusted such that the secondary goal comprising the 2-norm of the error of non-recoverable wrench and the 2-norm of the vector of the overall forces/torques of the healthy joints is minimized without affecting the recovered components of the wrench.

We propose a fast motion controller for a robot which has a flexible arm using IDCS control scheme. We test IDCS with the controller suitable for vibration suppressing: DMM to the lift table and consider the target tracking of IDCS with suppressing the vibration of the flexible board.

In recent years, Japanese society has been aging, engendering a labor shortage of young workers. Robots are therefore expected to be useful to perform tasks such as day-to-day work support for elderly people. Consequently, a tendon-driven balloon actuator has been developed for a robot hand in such environments. This study evaluated stroke control characteristics of a balloon actuator using a predictive functional control (PFC) system. Predictive functional control, a model based on predictive control (MPC) schemes, predicts the future outputs of the actual plant over the prediction horizon and computes the control effort over the control horizon at every sampling instance. Herein, PFC control performance of a one-link finger using a pneumatic balloon actuator is reported.

The domestic robot platform ACROBOTER exploits a novel concept of ceiling based locomotion. A climber unit moves on the almost obstacle free ceiling, while carries a swinging unit with a system of suspending and orienting cables. The objective of the robot is the fine positioning of the swinging unit that accomplishes path following or pick and place tasks. Its motion is controlled by ducted fan actuators additionally to the variable length suspending cables. The complexity of the mechanical structure induces the use of natural coordinates for the kinematical description. An algorithm is proposed to control this underactuated and also redundant manipulator, which can be characterized as a control-constraint based computed torque control strategy.

In this paper, we present some results about dexterous manipulation planning with an anthropomorphic hand. The task is to drive the grasped object from a start to a goal configuration. The planning algorithm automatically computes the finger motion and the required regrasping motion i.e. when and how to relocate the contacts on the object’s surface. The planner relies on a previously presented method but some extensions were added in order to make it applicable to the hand used in the experiments (a four-fingered hand with a human-inspired kinematics). Some experiments, conducted on the real platform, are presented.

This paper presents a new control algorithm, based on angular momentum, for balancing a planar inverted double pendulum robot having one degree of underactuation. The robot may either pivot about a fixed point, or roll with a curved foot over a flat ground. The controller is able to stabilize the robot in any unstable balanced configuration, and to follow arbitrary motion trajectories without losing balance. The latter necessarily involves some tracking error. Several simulation results are presented.

This paper explores the attributes of a double-pendulum robot that determine its ability to balance. A new measure is defined, called the velocity gain, that expresses the degree to which the robot’s centre of mass will move in response to motion of the robot’s actuated joint. This measure can be used both to analyse a robot’s performance and to design robot mechanisms for improved performance. Some properties of the velocity gain are explained, and several examples of both good and bad balancing robots are presented. The significance of this work is that a robot mechanism’s intrinsic ability to balance sets a physical upper limit to the robot’s attainable performance at balancing tasks, which is independent of the choice of control system.

In Japan, after the Great Hanshin Earthquake, attention became focused on rescue activities by robots on behalf of disaster victims. Each feature of the rescue robots was investigated. Attention was devoted to the peristalsis performed as a movement mechanism replaced with them when earthworm is movement. This study assesses a robot that can move forward by contact with the ground and by development of a mechanism for movement in pipes.

This paper proposes an algorithm dedicated to the control of off-road mobile robots at high speed. Based on adaptive and predictive principles, it first proposes a control law to preserve a high level of accuracy in the path tracking problem. Next, the dynamic model used for grip condition estimation is considered to address also robot integrity preservation thanks to the velocity limitation.

This paper presents an efficient algorithm for on-line obstacle avoidance that accounts for robot dynamics and actuator constraints. The robot trajectory (path and speed) is generated on-line by avoiding obstacles, optimally, one at a time. The trajectory is generated recursively using a

basic

algorithm that plans trajectory segments to intermediate goals. The use of intermediate goals ensures safety and convergence to the global goal. This approach reduces the original problem of avoiding

m

obstacles to

m

simpler problems of avoiding

one

obstacle each, producing a planner that is linear, instead of exponential, in the number of obstacles.

This paper describes a control algorithm that optimises the distribution of joint torques of a polyarticulted robot while performing obstacle clearance of a large step. In this work, a special class of polyarticulted locomotion systems known as hybrid wheellegged robots is considered. This type of system is usually redundantly actuated, involving internal forces that could be exploited to improve the tipover stability and the traction forces needed to address more challenging obstacles. The proposed algorithm is based on the forces distribution model including internal forces. Its objective is to optimize a criterion representing the maximum allowable disturbance while respecting the frictional contact conditions. The performance of this controller is evaluated in simulation.

Some aspects of the motion dynamics of a body with moveable internal masses in a resisting medium have been studied. The system consists of a body with two internal masses, that can move relative to the body. One of the internal masses vibrates in the horizontal direction and the other in the vertical direction. To provide the asymmetry of the friction force, necessary for the progressive motion in a given direction, it suffices to apply bristles to the contact surfaces. The bristles are attached to the body by means of spiral springs. The equations of motion are given and solved numerically. A prototype applying these principles has been constructed and proved positive.

Pushing/Pulling tasks is an important part of work in many industries. Usually, most researchers study the Push/Pull tasks by analyzing different posture conditions, force requirements, velocity factors, etc. However few studies have reported the effects of muscle fatigue. Fatigue caused by physical loading is one of the main reasons responsible for MusculoSkeletal Disorders (MSD). In this paper, muscle groups of articulation is considered and from joint level a new approach is proposed for muscle fatigue evaluation in the arms Push/Pull operations. The objective of this work is to predict the muscle fatigue situation in the Push/Pull tasks in order to reduce the probability of MSD problems for workers. A case study is presented to use this new approach for analyzing arm fatigue in Pushing/Pulling.

Using the

essential parameters

of the human body, we propose to calculate the LS solution with SVD factorization, which is the closest in 2 norm of a set of a priori anatomic values given by literature database. This solution keeps both the same minimum norm error given by the

essential

parameters and the physical anatomical meaning of the a priori values when the measuring noise and errors are small. Experimental results are presented.

Many industrial tasks cannot be executed by a robot alone. A way to help workers in order to decrease the risk of musculoskeletal disorders is to assist them with a collaborative robot. Yet assessing its usefulness to the worker remains costly because it usually requires a prototype. We propose a dynamic simulation framework to model the performing of a task jointly by a virtual manikin and a robot. It allows to measure physical quantities in order to perform an ergonomic assessment of the robot. Experiments are carried out on two different robots. The results show that the proposed simulation framework is helpful for designing collaborative robots. Further work includes enhancing the simulation realism and validation on a real robot.

This paper proposes a method to estimate muscle activity with taking account the volumetric effects of muscles. We analyze muscles as elastic bodies by using finite element method combined with computation technique in robotics. We describe the way to estimate muscle force and deformation as the problem to find muscle activity which produce joint torque calculated by inverse dynamics subject to equilibrium equation of elastic bodies. This problem is solved as sequential quadratic programming with finite element analysis in parallel for each muscle.

The Robotic Gait Rehabilitation (RGR) Trainer has been designed to address secondary gait deviations in stroke survivors undergoing rehabilitation. In this paper we describe the operating principle of the RGR Trainer and the systems ability to record the pelvic obliquity patterns (during normal gait and during hip hiking simulated by healthy subjects). Furthermore, we present results of experiment designed to teach new gait pattern in healthy subjects.

The goal of the ROBOSKIN project is to develop and demonstrate a range of new robot capabilities based on the tactile feedback provided by a robotic skin covering large areas of the robot body. So far, a principled investigation of these issues has been limited by the lack of tactile sensing technology enabling large scale experimental activities. As a matter of fact, skin based technology and embedded tactile sensors have been mostly demonstrated only at the stage of prototypes. The new capabilities are expected to improve the ability of robots to effectively and safely operate in unconstrained environments, as well as to communicate and cooperate with each other and with humans.

Estimation of the gravitational vertical is a fundamental problem faced by locomoting biological systems and robots alike. A robotic model of a vestibular system is suggested with the purpose of explaining an observed phenomenon—head stabilization during locomotion. The mechanical model of the vestibular system comprises a damped inclinometer and an inertial measurement unit which are mounted on an actuated orienting platform (a robotic head). Generic linear control is employed to stabilize the headplatform while the vestibular system exercises an extended Kalman filter algorithm to estimate the gravitational direction in space. It is demonstrated that stabilization of the head-platform is essential in achieving accurate verticality estimation as it attenuates the disturbances generated by locomotion and simplifies state observation in a non-inertial frame, without the need for fixed external beacons.

This work develops a method for tactilely mapping unknown harsh environments such as oil wells, using a manipulator equipped with only position sensors. Because contact with the environment may occur anywhere on the manipulator, determining the contact location is challenging. Here a method is developed, based on a probabilistic classification of the data according to the contact location on the manipulator, and the reconstruction of the surface using such classified data. The approach effectiveness is demonstrated in several case studies and laboratory experiments.

In this paper the problem of BIM based indoor navigation is considered. The purpose of the project is to develop the semantic navigation system of an autonomous robot using BIM (Building Information Modeling). The described representation enables semantic robot navigation with a goal specified at a various levels of abstraction. The concept of hierarchical action planning is presented, where the plan is a time-optimized path combined with a sequence of actions required for robot movement across the whole building. The navigation process is supported by semantic localization which utilizes two methods: object detection based on point clouds (the 3D camera data acquired and converted into a point cloud) and visual object detection (based on the image taken from two color cameras placed on the sides of the robot).

This paper describes an overload protection mechanism for a 6-axis force/torque sensor. It can limit the load applied to the force sensor by contacting a sole part with a top plate placed on the force sensor. Specifically, compression springs are arranged between the force sensor and the top plate, and the springs are precompressed. When more loads than the preload are applied to the top plate, the top plate touches the sole part by compressing the springs, and it is possible to prevent the 6-axis force/torque sensor from being over-loaded. Verification of the proposed mechanism is conducted through experiments with a human-carrying biped robot, WL-16RV.