A comprehensive framework for the analysis and synthesis of 3D human gait is presented. The framework consists of a realistic morphological representation of the human body involving 40 degrees of freedom and 17 body segments. Through the analysis of human gait, the joint reaction forces/moments can be estimated and parameters associated with postural stability can be quantified. The synthesis of 3D human gait is a complicated problem due to the synchronisation of a large number of joint variables. Herein, the framework is employed to reconstruct a dynamically balanced gait cycle and develop sets of reference trajectories that can be used for either the assessment of human mobility or the control of mechanical ambulatory systems. The gait cycle is divided into eight postural configurations based on particular gait events. Gait kinematic data is used to provide natural human movements. The balance stability analysis is performed with various ground reference points. The proposed reconstruction of the gait cycle requires two optimisation steps that minimise the error distance between evaluated and desired gait and balance constraints. The first step (quasi-static motion) is used to approximate the postural configurations to a region close to the second optimisation step target while preserving the natural movements of human gait. The second step (dynamic motion) considers a normal speed gait cycle and is solved using the spacetime constraint method and a global optimisation algorithm. An experimental validation of the generated reference trajectories is carried out by comparing the paths followed by 19 optical markers of a motion tracking system with the paths of the corresponding node points on the model.

In this work, a new translational robot formed with two different parallel manipulators with a common control point is introduced. An asymmetric parallel manipulator provides three translational degrees of freedom to the proposed robot while the orientation of the end-effector platform is kept constant by means of a Delta-like manipulator. An exact solution is easily derived to solve the forward displacement analysis while a semi-closed form solution is available for solving the inverse displacement analysis. The infinitesimal kinematics of the robot is approached by applying the theory of screws. Finally, a numerical example that consists of solving the inverse/forward displacement analysis as well as the forward acceleration analysis of the end-effector platform is presented. The example also includes the computation of the workspace and the direct/inverse singularities of the example.

Endowing robots with the ability of skill learning enables them to be versatile and skillful in performing various tasks. This paper proposes a neuro-fuzzy-based, self-organizing skill-learning framework, which differs from previous work in its capability of decomposing a skill by self-categorizing it into significant stimulus-response units (SRU, a fundamental unit of our skill representation), and self-organizing learned skills into a new skill. The proposed neuro-fuzzy-based, self-organizing skill-learning framework can be realized by skill decomposition and skill synthesis. Skill decomposition aims at representing a skill and acquiring it by SRUs, and is implemented by stages with a five-layer neuro-fuzzy network with supervised learning, resolution control, and reinforcement learning to enable robots to identify a sufficient number of significant SRUs for accomplishing a given task without extraneous actions. Skill synthesis aims at organizing a new skill by sequentially planning learned skills composed of SRUs, and is realized by stages, which establish common SRUs between two similar skills and self-organize a new skill from these common SRUs and additional new SRUs by reinforcement learning. Computer simulations and experiments with a Pioneer 3-DX mobile robot were conducted to validate the self-organizing capability of the proposed skill-learning framework in identifying significant SRUs from task examples and in common SRUs between similar skills and learning new skills from learned skills.

General-purpose autonomous robots must have the ability to combine the available sensorimotor knowledge in order to solve more complex tasks. Such knowledge is often given in the form of movement primitives. In this paper, we investigate the problem of sequencing of movement primitives. We selected nonlinear dynamic systems as the underlying sensorimotor representation because they provide a powerful machinery for the specification of primitive movements. We propose two new methodologies which both ensure that consecutive movement primitives are joined together in a continuous way (up to second-order derivatives). The first is based on proper initialization of the third-order dynamic motion primitives and the second uses online Gaussian kernel functions modification of the second-order dynamic motion primitives. Both methodologies were validated by simulation and by experimentally using a Mitsubishi PA-10 articulated robot arm. Experiments comprehend pouring, table wiping, and carrying a glass of liquid.

Hybrid robots, composed of a parallel platform and serial wrist, achieve a compromise of stiffness and dexterity. Thus, they are well suited for applications such as aircraft component machining and automotive assembly, where high accuracy and large workspace movements are required. However, their forward kinematics can be highly coupled and be nonlinear. To reduce the time required to define the forward kinematics of a robot with parallel–serial structure, this paper introduces the use of neural networks. Two radial basis function networks are trained to learn the parallel and serial kinematics separately, and then integrated into a complete model. The error of this network model is analyzed and identified by a particle swarm optimization algorithm. Simulation and experiment results are obtained from the hybrid robot, Exechon, which shows that the developed kinematic model is able to produce accurate position and orientation estimates of the end-effector. The computation time of the neural network model is greatly reduced when compared to the time achieved by the numerical model.

Exact knowledge of the position and proper calibration of robots that move by wheels form an important foundation in mobile robot applications. In this context, a variety of sensory systems and techniques have been developed for accurate positioning of differential drive mobile robots. This paper, first, provides a brief overview of mobile robots positioning techniques and then, presents a new benchmark method capable of calibrating mobile robots with differential drive mechanisms to correct systematic errors. The proposed method is compared with the commonly used University of Michigan Benchmark (UMBmark) odometry method. Two sets of comparisons are conducted on six prototyped robots with differential drives. The first set of tests establishes the workability and accuracy that can be achieved with the new method and compares them with the ones obtained from the UMBmark technique. The second experiment compares the performance of a mobile robot, calibrated with either the UMBmark or the new method, for an unseen path. It is demonstrated that the proposed method of calibration is simple to implement, and leads to accuracy comparable to the UMBmark method. Specifically, while the error corrections in both methods are within ±5% of each other, the proposed method requires single straight line motion for calibration, which is believed to be simpler and less timely to implement than the square path motion required by the UMBmark technique. The method should therefore be considered seriously as a new tool when calibrating differential drive mobile robots.

This paper presents an adaptive force/position controller for a parallel robot executing constrained motions. This controller, based on an MLPNN (or Multi-Layer Perceptron Neural Network), does not require the inverse dynamic model of the robot to derive the control law. A neural identification of the dynamic model of the robot is proposed to determine the principal components of the MLPNN input vector. The latter is used to compensate the dynamic effects arising from the robot–environment interaction and its parameters are adjusted according to an adaptation law based on the Lyapunov-analysis methodology. The proposed controller is evaluated experimentally on the C5 parallel robot. This method is capable of tracking accurately the force/position trajectories and its stability robustness is proved.

We have proposed a new repetition framework for vision-based behavior imitation by a sequence of multiple humanoid robots, introducing an on-line method for delimiting a time-varying context. This novel approach investigates the ability of a robot “student” to observe and imitate a behavior from a “teacher” robot; the student later changes roles to become the “teacher” for a naïve robot. For the many robots that already use video acquisition systems for their real-world tasks, this method eliminates the need for additional communication capabilities and complicated interfaces. This can reduce human intervention requirements and thus enhance the robots' practical usefulness outside the laboratory. Articulated motions are modeled in a three-layer method and registered as learned behaviors using color-based landmarks. Behaviors were identified on-line after each iteration by inducing a decision tree from the visually acquired data. Error accumulated over time, creating a context drift for behavior identification. In addition, identification and transmission of behaviors can occur between robots with differing, dynamically changing configurations. ITI, an on-line decision tree inducer in the C4.5 family, performed well for data that were similar in time and configuration to the training data but the greedily chosen attributes were not optimized for resistance to accumulating error or configuration changes. Our novel algorithm, OLDEX identified context changes on-line, as well as the amount of drift that could be tolerated before compensation was required. OLDEX can thus identify time and configuration contexts for the behavior data. This improved on previous methods, which either separated contexts off-line, or could not separate the slowly time-varying context into distinct regions at all. The results demonstrated the feasibility, usefulness, and potential of our unique idea for behavioral repetition and a propagating learning scheme.

To achieve desired configuration, a scheme for state adjustment of a redundant robot manipulator with no end-effector task explicitly assigned and referred to as a state-adjustment scheme is proposed in this paper. Owing to the physical limits in an actual robot manipulator, both joint and joint-velocity limits are incorporated into the proposed scheme for practical purposes. In addition, the proposed state-adjustment scheme is formulated as a quadratic program and resolved at the joint-velocity level. A numerical computing algorithm based on the conversion technique of the quadratic program to linear variational inequalities is presented to address the robot state-adjustment scheme. By employing the state-adjustment scheme, the robot manipulator can automatically move to the desired configuration from any initial configuration with the movement kept within its physical limits. Computer simulation and experimental results using a practical six-link planar robot manipulator with variable joint-velocity limits further verify the realizability, effectiveness, accuracy, and flexibility of the proposed state-adjustment scheme.

We study a problem of K-barrier coverage by employing a network of self-deployed, autonomous mobile robotic sensors. A decentralized coordination algorithm is proposed for the robotic sensors to address the coverage problem. The algorithm is developed based on some simple rules that only rely on local information. By applying the algorithm to the robotic sensors, K layers of sensor barriers are formed to cover the region between two given points. To illustrate the proposed algorithm, numerical simulations are carried out for a number of scenarios.

This paper presents the results of a study on the dynamical modeling, analysis, and control of a spherical rolling robot. The rolling mechanism consists of a 2-DOF pendulum located inside a spherical shell with freedom to rotate about the transverse and longitudinal axis. The kinematics of the model has been investigated through the classical methods with rotation matrices. Dynamic modeling of the system is based on the Euler–Lagrange formalism. Nonholonomic and highly nonlinear equations of motion have then been decomposed into two simpler subsystems through the decoupled dynamics approach. A feedback linearization loop with fuzzy controllers has been designed for the control of the decoupled dynamics. Rolling of the controlled mechanism over linear and curvilinear trajectories has been simulated by using the proposed decoupled dynamical model and feedback controllers. Analysis of radius of curvature over curvilinear trajectories has also been investigated.

This paper considers the magnitude of the gripping power, i.e., the internal force that depends on the grasping posture or object orientation in a two-dimensional grasp by two contact points with friction. Expressing the effect of variations in the object posture as the direction of an external force, we propose an “internal force diagram.” The internal force necessary to create a statically stable grasp is depicted in the object coordinate frame. Then, a polar coordinate system is introduced in which the orientation represents the direction of the external force, while the distance from the origin represents the minimum necessary internal force. We demonstrate a method based on friction cone configurations to manually draw the internal force diagram, using only a ruler and a compass. The validity of this drawing method is confirmed by a comparison with computer-generated plots. Finally, the characteristics of the internal force diagram are discussed.

This paper presents a wheel–terrain interaction model, which enables efficient modeling of wheeled locomotion in soft soil and numerical simulations of off-road mobile robots. This modular model is derived based on wheel kinematics and terramechanics and the main focus is on describing the stress distributions along the wheel–terrain interface and the reaction forces exerted on the wheel by the soil. When the wheels are steered, the shear stresses underneath the wheel were modeled based on isotropic assumptions. The forces and torques contributed by the bulldozing effect of the side surfaces is also considered in the proposed model. Furthermore, the influence of grousers, commonly used on smaller mobile robots, was modeled by (1) averaging the normal pressures contributed by the grousers and the wheel concave portion, and (2) assuming that the shear phenomenon takes places along the grouser tips. By integrating the model with multibody system code for vehicle dynamics, simulation studies of various off-road conditions in three-dimensional environments can be conducted. The model was verified by using field experiment data, both for a single-wheel vehicle and a whole vehicle.

Traditionally, robot motion planners use Voronoi Diagrams for generating admissible paths that connect an initial with a final configuration. When dynamic scenarios are involved, these techniques imply a heavy computational cost. The novelty of the technique presented here is that it can provide fast and particularly suitable routes without considering the full scenario if environmental changes appear. The new method is designed to work with a post-process technique in order to provide admissible paths for car-like robots coping with kinematic constraints. This new approach is capable of supplying continuous paths and also elaborate maneuvers if cluttered environments are involved.

Kinematic reconfigurable mobile robots have the ability to change their structure to increase stability and decrease the probability of tipping over on rough terrain. If stability increases without decreasing center of mass height, the robot can pass more easily through bushes and rocky terrain. In this paper, an improved sample return rover is presented. The vehicle has a redundant rolling degree of freedom. A genetic algorithm utilizes this redundancy to optimize stability. Parametric motion equations of the robot were derived by considering Iterative Kane and Lagrange's dynamic equations. In this research, an optimal reconfiguration strategy for an improved SRR mobile robot in terms of the Force–Angle stability measure was designed using a genetic algorithm. A path-tracking nonlinear controller, which maintains the robot's maximum stability, was designed and simulated in MATLAB. In the simulation, the vehicle and end-effector paths and the terrain are predefined and the vehicle has constant velocity. The controller was found to successfully keep the end-effector to the desired path and maintained optimal stability. The robot was simulated using ADAMS for optimization evaluation.

In this paper we address the problem of stabilization and local positioning of a four-rotor rotorcraft using computer vision. Our approaches to estimate the orientation and position of the rotorcraft combine the measurements from an Inertial Measurement Unit (IMU) and a vision system composed of a single camera. In the first stage, the vision system is used to estimate the position and yaw angle of the rotorcraft, while in the second stage the vision system is used to estimate the translational velocity of the flying robot. In both cases the IMU gives the pitch and roll angles at a higher rate. The technique used to estimate the position of the rotorcraft in the first stage combines the homogeneous transformation approach for the camera calibration process with the plane-based pose method for estimating the position. In the second stage, a navigation system using the optical flow is also developed to estimate the translational velocity of the aircraft. We present real-time experiments of stabilization and location of a four-rotor rotorcraft.