Recursive Methods in Modeling and Control of Modular Robotic Systems

This paper addresses the dynamical modeling and control of reconfigurable modular robots. The modular actuators (brushless DC motors with Harmonic Drive gears) for the robots under consideration are connected by rigid links. This way the robot can be assembled in different configurations by rearranging these components. For dynamical modeling the Projection Equation in Subsystem representation is used, taking advantage of its modular structure. Due to the lack of position sensors at the gearbox output shaft, deflections caused by the elasticities in the gears can not be compensated by the PD motor joint controller. Therefore, a correction of the motor trajectory is needed, which can be calculated as part of a flatness based feed-forward control using the exact model of the robot. With the recursive approach proposed in this paper the concept of reconfigurability is retained. For validation a redundant articulated robot arm with seven joints is regarded and results are presented. (© 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

On the Suitability of Different Online Gait Generation Methods for Pre-Defined Footsteps

Detailed dynamical modeling is the basis for simulation and model based control. In this contribution the Projection Equation is used for the modeling of a biped walking machine, resulting in the equations of motion which are needed for gait generation and verification of its stability. For biped robots one main issue is the generation of stable trajectories for the center of mass (CoM). Several different approaches based on the Zero Moment Point (ZMP) scheme have been presented in the past. Due to the complex dynamic structure of bipedal robots most of the considered algorithms use a linear inverted pendulum as a simplified model. This results in a decoupling of the ZMP equations in lateral and forward direction, but limits the trajectories to a constant height of the CoM. An extension of the well known LQR theory by future reference values has been proposed. This model based approach seems to perform quite well, but does not allow the consideration of constraints on the position of the ZMP. This limitation is removed by the use of Model Predictive Control (MPC) with inequality constraints. By extending this approach to a time invariant one the restriction to a constant height of the CoM is no longer necessary. Both methods as well as the time invariant approach for variable CoM heights have been evaluated in simulations and will be experimentally verified on a real robot soon. (© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

The Lightweight Robot ElRob: Interesting Aspects in Modeling and Control

The articulated robot ElRob, consisting of flexible links and joints, is considered in several publications. Recent developments are presented in this work. The overall goal of the research is to decrease the effects of structural elasticities in lightweight robots. For this purpose model-based control concepts are investigated and very accurate and efficient kinematic and dynamic models are necessary. The robot is split into groups of bodies, the so called subsystems, with separated describing velocities and coordinate systems. To obtain structured equations of motion the Projection Equation is used. The beams are modelled using the floating frame of reference formulation and a Ritz-approach. Because of its flexibility, the examined robot is an underactuated system leading to special difficulties. As an example is it not possible to compute the desired joint angles with respect to a reference path in task space for the flexible system (inverse kinematic problem). Different methods to solve this drawback and other problems resulting from flexibility are discussed with special focus on feed forward control and different feedback control concepts. The resulting end point error, the necessary control input and other interesting results for the laboratory experiment are presented and compared. (© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Modeling and Quasi-Static Trajectory Control of a Self-Balancing Two-Wheeled Vehicle

This paper introduces the kinematical and dynamical model as well as a quasi-static trajectory control of a self-balancing two-wheeled vehicle. The mobile robot is about 60cm tall, autonomous, unstable and driven by two wheels. Hence, it can be used for transport purposes. Due to the nonholonomic constraints only few modeling techniques are feasible. In this case, the modeling is based on the Projection Equation, followed by the derivation of various control strategies. In order to allow a desired velocity and to stabilize the inclination angle of the robot a partial feedback linearization in combination with a LQR controller is applied. The quasi-static trajectory controller, which is based on the kinematical model, uses a flatness based approach in order to remain on the desired path. Continuous curvature paths, composed by clothoids, enable good performance results in simulation and experiment. (© 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Time Optimal Path Planning for Industrial Robots: A Dynamic Programming Approach Considering Torque Derivative and Jerk Constraints

This paper presents a dynamic programming approach for calculating time optimal trajectories for industrial robots, subject to various physical constraints. In addition to path velocity, motor torque, joint velocity and acceleration constraints, the present contribution also shows how to deal with torque derivative and joint jerk limitations. First a Cartesian path for the endeffector is defined by splines using Bernstein polynomials as basis functions and is parameterized via a scalar path parameter. In order to compute the belonging quantities in configuration space, inverse kinematics is solved numerically. Using this and in combination with the dynamical model, joint torques as well as their derivatives can be constrained. For that purpose the equations of motion are calculated with the help of the Projection Equation. As a consequence of the used optimization problem formulation, the dynamical model as well as the restrictions have to be transformed to path parameter space. Due to the additional consideration of jerk and torque derivative constraints, the phase plane is expanded to a phase space. The parameterized restrictions lead to feasible regions in this space, in which the optimal solution is sought. Result of the optimization is the time behavior of the path parameter and subsequently the feed forward torques for the optimal motion on the spatial path defined by previously mentioned splines. Simulation results as well as experimental results for a three axes industrial robot are presented. (© 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Modeling Techniques for Coupled,Rotating Beams

Beams are parts of many industrial applications, like robot links, rolls in paper industry and turbo charger. In this work, a rotordynamical problem, the powertrain for a mill stand, is under consideration. Torsional and bending vibrations are used to describe the dynamical behavior. There are several methods for deriving the dynamical equations of motion. In this paper, the Projection Equation, a synthetical method, is used, leading to partial differential equations for the distributed parameter system. A simplification can be done by using the Ritz approximation method. This method requires the fulfillment of the geometric boundary conditions. For our example, a combination of rigid body modes and elastic modes is chosen. Also models for the gear box system and bearings are included. The solutions for the overall example are nonlinear ordinary differential equations which can be integrated numerically. The system is excited by constant torques and forces. Simulation results for this elastic multibody system are presented. (© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Energy interaction between linear and nonlinear oscillators (energy transfer through the subsystems in a hybrid system)

The analysis of the energy transfer between subsystems coupled in a hybrid system is an urgent problem for various applications. We present an analytic investigation of the energy transfer between linear and nonlinear oscillators for the case of free vibrations when the oscillators are statically or dynamically connected into a double-oscillator system and regarded as two new hybrid systems, each with two degrees of freedom. The analytic analysis shows that the elastic connection between the oscillators leads to the appearance of a two-frequency-like mode of the time function and that the energy transfer between the subsystems indeed exists. In addition, the dynamical linear constraint between the oscillators, each with one degree of freedom, coupled into the hybrid system changes the dynamics from single-frequency modes into two-frequency-like modes. The dynamical constraint, as a connection between the subsystems, is realized by a rolling element with inertial properties. In this case, the analytic analysis of the energy transfer between linear and nonlinear oscillators for free vibrations is also performed. The two Lyapunov exponents corresponding to each of the two eigenmodes are expressed via the energy of the corresponding eigentime components. Published in Ukrains'kyi Matematychnyi Zhurnal, Vol. 60, No. 6, pp. 796–814, June, 2008.     

Connectivity estimations of errors of linearization of essentially nonlinear systems

We consider a nonlinear dynamical system with several connectivity components. It includes subsystems which can be switched off or on in the operation process, i.e., the system undergoes structural changes. It is well-known that such systems are stable with respect to the connectivity. This property is known as the connectivity stability. In this paper we find an upper bound for the solution of the initial multiply connected domain of a nonlinear dynamical system and obtain a connectivity estimation for its linearization error.     

A modular and efficient approach to computational modeling and sensitivity analysis of robot and human motion dynamics

In this paper a new class library for the computation of the forward multi-body-system (MBS) dynamics of robots and biomechanical models of human motion is presented. By the developed modular modeling approach the library can be flexibly extended by specific modeling elements like joints with specific geometry or different muscle models and thus can be applied efficiently for a number of dynamic simulation and optimization problems. The library not only provides several methods for solving the forward dynamics problem (like articulated body or composite rigid body algorithms) which can transparently be exchanged. Moreover, the numerical solution of optimal control problems, like in the forward dynamics optimization of human motion, is significantly facilitated by the computation of the sensitivity matrix with respect to the control variables. Examples are given to demonstrate the efficiency of the approach. (© 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Synthesis of a control in a non-linear dynamical system based on decomposition

A non-linear controllable dynamical system with many degrees of freedom, described by Lagrange equations of the second kind, is considered. Geometric constraints are imposed on the magnitudes of the controls. It is assumed that, in the equations of motion, the kinetic energy matrix is close to a certain constant diagonal matrix. It is possible, for example, to reduce the equations of motion of robots, the drives of which have large gear ratios, to a system of this kind. A problem is formulated on the transfer of a system in a finite time from a specified initial state to a final state with zero velocities. The method of decomposition [1] is used to construct the equations. Sufficient conditions are found subject to which the maximum values of the non-linear terms in the equations of motion do not exceed the permissible magnitudes of the controls. In this case, non-linearities are treated as limited perturbations and the system is decomposed into independent, linear, second-order subsystems. A feedback control is specified for these subsystems which guarantees that each of them is brought into the final state for any permissible perturbations. The control has a simple structure. Applications of the proposed approach to problems in the control of manipulating robots are considered.     

Two Methods of Simulator Coupling

Modelling and simulation of complex engineering systems are often relieved by a modular approach in which the global system is decomposed into subsystems. Advantages arise from independent and parallel modelling of subsystems over easy exchange of the resulting modules to the use of different software for each module. However, the modular simulation of the global system by coupling of simulators may result in an unstable integration, if an algebraic loop exists between the subsystems. This numerical phenomenon is analyzed and two methods of simulator coupling which guarantee stability for general systems including algebraic loops are introduced. Numerical results of the modular simulation of a multibody system are presented.     

On Ramanujan’s modular identities

For Ramanujan’s modular identities connected with his well-known partition congruences for the moduli 5 or 7, we had given, in an earlier paper, natural and uniform proofs through the medium of modular forms. Analogous (modular) identities corresponding to the (more difficult) case of the modulus 11 are provided here, with the consequent partition congruences; the relationship with relevant results of N J Fine is also sketched.     

Experimental investigation of an oscillator with discontinuous support considering different system aspects

Non-smooth characteristics are, in general, the source of difficulties for the modeling and simulation of natural systems. These characteristics are usually related to either the friction phenomenon or the discontinuous behavior as intermittent contacts. This article develops an experimental investigation concerning non-smooth systems with discontinuous support. An experimental apparatus is developed in order to analyze the nonlinear dynamics of a single-degree of freedom system with discontinuous support. The apparatus is composed by an oscillator constructed by a car, free to move over a rail, connected to an excitation system. The discontinuous support is constructed considering mass–spring systems separated by a gap to the car position. This apparatus is instrumented to obtain all the system state variables. System dynamical behavior shows a rich response, presenting dynamical jumps, bifurcations and chaos. Different configurations of the experimental set up are treated in order to evaluate the influence of the internal impact within the car and also support characteristics in the system dynamics.     

On minimal decomposition of p-adic polynomial dynamical systems

A polynomial of degree ?2 with coefficients in the ring of p-adic numbers Zp is studied as a dynamical system on Zp. It is proved that the dynamical behavior of such a system is totally described by its minimal subsystems. For an arbitrary quadratic polynomial on Z₂, we exhibit all its minimal subsystems.     

Estimates of reachable sets of multidimensional control systems with nonlinear interconnections

The paper is devoted to the problem of constructing external estimates for the reachable set of a multidimensional control system by means of vector estimators. A system is considered that permits a decomposition into several independent subsystems with simple structure (for example, linear subsystems), which are connected to each other by means of nonlinear interconnections. For each of the subsystems, an external estimate of the reachable set is assumed to be known; this estimate is representable in the form of a level set of some function satisfying a differential inequality. An estimate for the reachable set of the combined system is constructed with the use of estimates for subsystems. The method of deriving the estimates is based on constructing comparison systems for analogs of vector Lyapunov functions (value functions).     

Model Order Reduction of a Drill-String-Model with Self-Excited Stick-Slip Vibrations

The dynamical behavior of a drill-string is defined by its small diameter-to-length ratio, which makes the string vulnerable to torsional vibrations. In combination with the nonlinear friction characteristic at the drill bit, this can lead to self-excited stick-slip vibrations which are detrimental to the drilling process. The string can be modeled by the Finite Element Method or as a Multi-Body system to represent the distributed character of the system. The analysis of the resulting high-dimensional model is, however, elaborate and time-consuming. We show that through Galerkin Projection onto the first two Characteristic Functions gained from Karhunen-Loève-Transformation, a reduced system can be obtained which reproduces the essential dynamical properties of the original system, e.g. the stick-slip motion. With the reduced system, the linear stability of the drill-string can be analyzed. We show that by reducing the inertia of the rotary table the system can be stabilized. (© 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Lyapunov formulation of the ISS cyclic-small-gain theorem for hybrid dynamical networks

This paper presents a Lyapunov-based cyclic-small-gain theorem for the hybrid dynamical networks composed of input-to-state stable (ISS) subsystems whose motions may be continuous, impulsive or piecewise constant on the time-line. On the one hand, it is shown that hybrid dynamic networks with interconnection gains less than the identity function are ISS by means of Lyapunov functions. Additionally, an ISS-Lyapunov function for the total network is constructed using the ISS-Lyapunov functions of the subsystems. On the other hand, a novel result of this paper shows that a hybrid dynamic network satisfying the cyclic-small-gain condition can be transformed into one with interconnection gains less than the identity. In sharp contrast with several previously known results, the impulses of the subsystems are time triggered and the impulsive times for different subsystems may be different.     

Efficient reconfiguration of lattice-based modular robots

Modular robots consist of many identical units (or atoms) that can attach together and perform local motions. By combining such motions, one can achieve a reconfiguration of the global shape of a robot. The term modular comes from the idea of grouping together a fixed number of atoms into a metamodule, which behaves as a larger individual component. Recently, a fair amount of research has focused on algorithms for universal reconfiguration using Crystalline and Telecube metamodules, which use expanding/contracting cubical atoms.From an algorithmic perspective, this work has achieved some of the best asymptotic reconfiguration times under a variety of different physical models. In this paper we show that these results extend to other types of modular robots, thus establishing improved upper bounds on their reconfiguration times. We describe a generic class of modular robots, and we prove that any robot meeting the generic class requirements can simulate the operation of a Crystalline atom by forming a six-arm structure. Previous reconfiguration bounds thus transfer automatically by substituting the six-arm structures for the Crystalline atoms. We also discuss four prototyped robots that satisfy the generic class requirements: M-TRAN, SuperBot, Molecube, and RoomBot.