Achieving the asymptotic tracking control of electrically driven robot manipulators is a challenging problem due to approximation/modelling error arising from parametric and non-parametric uncertainty. Thanks to the specific property of Taylor series systems as they are universal approximators, this research outlines two robust control schemes using an adaptive Taylor series system for robot manipulators, including actuators' dynamics. First, an indirect adaptive controller is designed such as to approximate an uncertain continuous function by using a Taylor series system in the proposed control law. Second, a direct adaptive scheme is established to employ the Taylor series system as a controller. In both controllers, not only a robustifying term is constructed using the estimation of the upper bound of approximation/modelling error, but the closed-loop stability, as well as the asymptotic convergence of joint-space tracking error and its time derivative, is ensured. Due to the design of the Taylor series system in the tracking error space, our technique clearly has an advantage over fuzzy and neural network-based control methods in terms of the small number of tuning parameters and inputs. The proposed methods are simple, model free in decentralized forms, no need for uncertainty bounding functions and perfectly capable of dealing with parametric and non-parametric uncertainty and measurement noise. Finally, simulation results are introduced to confirm the efficiency of the proposed control methods.

This paper discusses the concept and parameter design of a robust stair-climbing compliant modular robot, capable of tackling stairs with overhangs. Geometry modifications of the periphery of the wheels of our robot helped in tackling overhangs. Along with establishing a concept design, the robust design parameters are set to minimize performance variations. The Grey-based Taguchi method is applied to provide an optimal setting for the design parameters of the robot. The robot prototype is shown to have successfully scaled stairs of varying dimensions, with overhang, thus corroborating the analysis performed.

A Q-learning approach is often used for navigation in static environments where state space is easy to define. In this paper, a new Q-learning approach is proposed for navigation in dynamic environments by imitating human reasoning. As a model-free method, a Q-learning method does not require the environmental model in advance. The state space and the reward function in the proposed approach are defined according to human perception and evaluation, respectively. Specifically, approximate regions instead of accurate measurements are used to define states. Moreover, due to the limitation of robot dynamics, actions for each state are calculated by introducing a dynamic window that takes robot dynamics into account. The conducted tests show that the obstacle avoidance rate of the proposed approach can reach 90.5% after training, and the robot can always operate below the dynamics limitation.

This paper introduces a novel 2R1T parallel manipulator redundantly actuated by pneumatic muscles for lower-limb rehabilitation. First, the conceptual design of the proposed 3-DOF parallel mechanism is presented. Then, the inverse kinematics and the generalized Jacobian analysis are carried out. Based on the generalized Jacobian and the constraint characteristics of the mechanism, the force/motion transmissibility of the redundantly actuated parallel mechanism is investigated via four individual cases without actuation redundancy, leading to a suitable local transmission index for the evaluation of kinematic performance of the proposed mechanism. Finally, the design variables are optimized by maximizing the mean value of the local transmission index with the aid of genetic algorithm. The numerical result shows that the proposed parallel mechanism can achieve a good kinematic performance in its task workspace.

This paper reports an extended state observer (ESO)-based robust dynamic surface control (DSC) method for triaxial MEMS gyroscope applications. An ESO with non-linear gain function is designed to estimate both velocity and disturbance vectors of the gyroscope dynamics via measured position signals. Using the sector-bounded property of the non-linear gain function, the design of an $\mathcal{L}_2$-robust ESO is phrased as a convex optimization problem in terms of linear matrix inequalities (LMIs). Next, by using the estimated velocity and disturbance, a certainty equivalence tracking controller is designed based on DSC. To achieve an improved robustness and to remove static steady-state tracking errors, new non-linear integral error surfaces are incorporated into the DSC. Based on the energy-to-peak ($\mathcal{L}_2$-$\mathcal{L}_\infty$) performance criterion, a finite number of LMIs are derived to obtain the DSC gains. In order to prevent amplification of the measurement noise in the DSC error dynamics, a multi-objective convex optimization problem, which guarantees a prescribed $\mathcal{L}_2$-$\mathcal{L}_\infty$ performance bound, is considered. Finally, the efficacy of the proposed control method is illustrated by detailed software simulations.

This paper presents an optimal trajectory planning method for industrial robots. The paper specially focuses on the applications of path tracking. The problem is to plan the trajectory with a specified geometric path, while allowing the position and orientation of the path to be arbitrarily selected within the specific ranges. The special contributions of the paper include (1) an optimal path tracking formulation focusing on the least time and energy consumption without violating the kinematic constraints, (2) a special mechanism to discretize a prescribed path integration for segment interpolation to fulfill the optimization requirements of a task with its constraints, (3) a novel genetic algorithm (GA) optimization approach that transforms a target path to be tracked as a curve with optimal translation and orientation with respect to the world Cartesian coordinate frame, (4) an integration of the interval analysis, piecewise planning and GA algorithm to overcome the challenges for solving the special trajectory planning and path tracking optimization problem. Simulation study shows that it is an insufficient condition to define a trajectory just based on the consideration that each point on the trajectory should be reachable. Simulation results also demonstrate that the optimal trajectory for a path tracking problem can be obtained effectively and efficiently using the proposed method. The proposed method has the properties of broad adaptability, high feasibility and capability to achieve global optimization.

This paper presents the design, analysis, and experimental validation of a miniature modular inchworm robot (MMIR). Inchworm robots are capable of maneuvering in confined spaces due to their small size, a desirable characteristic for surveillance, exploration and search and rescue operations. This paper presents two generations of the MMIR (Version 1—V1 and Version 2—V2) that utilize anisotropic friction skin and an undulatory rectilinear gait to produce locomotion. This paper highlights design improvements and a multi-body dynamics approach to model and simulate the system. The MMIR V2 incorporates a slider-crank four-bar mechanism and a relative body revolute joint to produce high-frequency relative translation and rotation to increase forward velocity and enable turning capabilities. Friction analysis and locomotion experiments were conducted to assess the systems performance on various surfaces, validate the dynamic model and simulation results, and measure the maximum forward velocity. The MMIR V1 and V2 were able to achieve maximum forward velocities of 12.7 mm/s and 137.9 mm/s, respectively. These results are compared to reported results of similar robots published in the literature.

The paper introduces the use of radial basis functions (RBFs) to generate smooth point-to-point joint trajectories for robot manipulators. First, Gaussian RBF interpolation is introduced taking into account boundary conditions. Then, the proposed approach is compared with classical planning techniques based on polynomial and trigonometric models. Also, the trajectory planning problem involving via-points is solved using the proposed RBF interpolation technique. The obtained trajectories are then compared with those synthesized using algebraic and trigonometric splines. Finally, the proposed method is tested for the six-joint PUMA 560 robot in two cases (minimum time and minimum time-jerk) and results are compared with those of other planning techniques. Numerical results demonstrate the advantage of the proposed technique, offering an effective solution to generate trajectories with short execution time and smooth profile.

Previous researches on analysis of stability of haptic simulation largely assume that the virtual environment is passive. But the virtual environment can become non-passive due to various reasons including discretization errors and interaction dynamics between virtual tools and objects. This paper provides an analysis of the stability and performance of the haptic simulation involving non-passive virtual environment. The dynamic interaction between the virtual tools and the objects is modeled using the two-port networks. The analysis is carried out using a velocity and force mapping matrix for six-DOF simulation. New stability condition resulting from the analysis is applied to an example simulation of a one-DOF virtual wall. Maximum stiffness satisfying the stability condition established in the previous literature, and the proposed condition is compared with the maximum stiffness experimentally determined with various time steps. The newly proposed stability condition manifests less standard deviation of errors than the widely applied absolute stability condition.

Consider the practically relevant situation in which a robotic system is assigned a task to be executed in an environment that contains moving obstacles. Generating collision-free motions that allow the robot to execute the task while complying with its control input limitations is a challenging problem, whose solution must be sought in the robot state space extended with time. We describe a general planning framework which can be tailored to robots described by either kinematic or dynamic models. The main component is a control-based scheme for producing configuration space subtrajectories along which the task constraint is continuously satisfied. The geometric motion and time history along each subtrajectory are generated separately in order to guarantee feasibility of the latter and at the same time make the scheme intrinsically more flexible. A randomized algorithm then explores the search space by repeatedly invoking the motion generation scheme and checking the produced subtrajectories for collisions. The proposed framework is shown to provide a probabilistically complete planner both in the kinematic and the dynamic case. Modified versions of the planners based on the exploration–exploitation approach are also devised to improve search efficiency or optimize a performance criterion along the solution. We present results in various scenarios involving non-holonomic mobile robots and fixed-based manipulators to show the performance of the planner.