Efficient fuzzy control of a rotary inverted pendulum based on LQR mapping

This paper develops fuzzy methods for control of the rotary inverted pendulum, an underactuated mechanical system. Two control laws are presented, one for swing up and another for the stabilization. The pendulum is swung up from the vertical down stable position to the upward unstable position in a controlled trajectory. The rules for the swing up are heuristically written such that each swing results in greater energy build up. The stabilization is achieved by mapping a stabilizing LQR control law to two fuzzy inference engines, which reduces the computational load compared with using a single fuzzy inference engine. The robustness of the balancing control is tested by attaching a bottle of water at the tip of the pendulum.

Strong structural controllability and the multi-link inverted pendulum

This technical note investigates the controllability of the linearized dynamics of the multilink inverted pendulum as the number of links and the number and location of actuators changes. It is demonstrated that, in some instances, there exist sets of parameter values that render the system uncontrollable and so usual methods for assessing controllability are difficult to employ. To assess the controllability, a theorem on strong structural controllability for single-input systems is extended to the multiinput case.

Strong structural controllability and the multilink inverted pendulum

This technical note investigates the controllability of the linearized dynamics of the multilink inverted pendulum as the number of links and the number and location of actuators changes. It is demonstrated that, in some instances, there exist sets of parameter values that render the system uncontrollable and so usual methods for assessing controllability are difficult to employ. To assess the controllability, a theorem on strong structural controllability for single-input systems is extended to the multiinput case.

Non-linear dynamics of a tower orchard sprayer based on an inverted pendulum model

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

From inverted pendulum to N-link chain: inertial sensors-based assessment of movement kinematics and dynamics for functional evaluation and rehabilitation

Despite several clinical tests that have been developed to qualitatively describe complex motor tasks by functional testing, these methods often depend on clinicians' interpretation, experience and training, which make the assessment results inconsistent, without the precision required to objectively assess the effect of the rehabilitative intervention. A more detailed characterization is required to fully capture the various aspects of motor control and performance during complex movements of lower and upper limbs. The need for cost-effective and clinically applicable instrumented tests would enable quantitative assessment of performance on a subject-specific basis, overcoming the limitations due to the lack of objectiveness related to individual judgment, and possibly disclosing subtle alterations that are not clearly visible to the observer. Postural motion measurements at additional locations, such as lower and upper limbs and trunk, may be necessary in order to obtain information about the inter-segmental coordination during different functional tests involved in clinical practice. With these considerations in mind, this Thesis aims: i) to suggest a novel quantitative assessment tool for the kinematics and dynamics evaluation of a multi-link kinematic chain during several functional motor tasks (i.e. squat, sit-to-stand, postural sway), using one single-axis accelerometer per segment, ii) to present a novel quantitative technique for the upper limb joint kinematics estimation, considering a 3-link kinematic chain during the Fugl-Meyer Motor Assessment and using one inertial measurement unit per segment. The suggested methods could have several positive feedbacks from clinical practice. The use of objective biomechanical measurements, provided by inertial sensor-based technique, may help clinicians to: i) objectively track changes in motor ability, ii) provide timely feedback about the effectiveness of administered rehabilitation interventions, iii) enable intervention strategies to be modified or changed if found to be ineffective, and iv) speed up the experimental sessions when several subjects are asked to perform different functional tests.

Fuzzy Optimal Control for Double Inverted Pendulum

In this paper a fuzzy optimal control for stabilizing an upright position a double inverted pendulum (DIP) is developed and compared. Modeling is based on Euler-Lagrange equations. This results in a complicated nonlinear fast reaction, unstable multivariable system. Firstly, the mathematical models of double pendulum system are presented. The weight variable fuzzy input is gained by combining the fuzzy control theory with the optimal control theory. Simulation results show that the controller, which the upper pendulum is considered as main control variable, has high accuracy, quick convergence speed and higher precision.

Parametric resonances in a base-excited double pendulum

Two parametrically-induced phenomena are addressed in the context of a double pendulum subject to a vertical base excitation. First, the parametric resonances that cause the stable downward vertical equilibrium to bifurcate into large-amplitude periodic solutions are investigated extensively. Then the stabilization of the unstable upward equilibrium states through the parametric action of the high-frequency base motion is documented in the experiments and in the simulations. It is shown that there is a region in the plane of the excitation frequency and amplitude where all four unstable equilibrium states can be stabilized simultaneously in the double pendulum. The parametric resonances of the two modes of the base-excited double pendulum are studied both theoretically and experimentally. The transition curves (i.e., boundaries of the dynamic instability regions) are constructed asymptotically via the method of multiple scales including higher-order effects. The bifurcations characterizing the transitions from the trivial equilibrium to the periodic solutions are computed by either continuation methods and or by time integration and compared with the theoretical and experimental results.

Effects of joint immobilization on standing balance

We investigated the effect of joint immobilization on the postural sway during quiet standing. We hypothesized that the center of pressure (COP), rambling, and trembling trajectories would be affected by joint immobilization. Ten young adults stood on a force plate during 60 s without and with immobilized joints (only knees constrained, CK; knees and hips, CH; and knees, hips, and trunk, CT). with their eyes open (OE) or closed (CE). The root mean square deviation (RMS, the standard deviation from the mean) and mean speed of COP, rambling, and trembling trajectories in the anterior-posterior and medial-lateral directions were analyzed. Similar effects of vision were observed for both directions: larger amplitudes for all variables were observed in the CE condition. In the anterior-posterior direction, postural sway increased only when the knees, hips, and trunk were immobilized. For the medial-lateral direction, the RMS and the mean speed of the COP, rambling, and trembling displacements decreased after immobilization of knees and hips and knees, hips, and trunk. These findings indicate that the single inverted pendulum model is unable to completely explain the processes involved in the control of the quiet upright stance in the anterior-posterior and medial-lateral directions. (C) 2009 Elsevier B.V. All rights reserved.

Postural control during kneeling

Postural control was studied when the subject was kneeling with erect trunk in a quiet posture and compared to that obtained during quiet standing. The analysis was based on the center of pressure motion in the sagittal plane (CPx), both in the time and in the frequency domains. One could assume that postural control during kneeling would be poorer than in standing because it is a less natural posture. This could cause a higher CPx variability. The power spectral density (PSD) of the CPx obtained from the experimental data in the kneeling position (KN) showed a significant decrease at frequencies below 0.3 Hz compared to upright (UP) (P < 0.01), which indicates less sway in KN. Conversely, there was an increase in fast postural oscillations (above 0.7 Hz) during KN compared to UP (P < 0.05). The root mean square (RMS) of the CPx was higher for UP (P < 0.01) while the mean velocity (MV) was higher during KN (P < 0.05). Lack of vision had a significant effect on the PSD and the parameters estimated from the CPx in both positions. We also sought to verify whether the changes in the PSD of the CPx found between the UP and KN positions were exclusively due to biomechanical factors (e.g., lowered center of gravity), or also reflected changes in the neural processes involved in the control of balance. To reach this goal, besides the experimental approach, a simple feedback model (a PID neural system, with added neural noise and controlling an inverted pendulum) was used to simulate postural sway in both conditions (in KN the pendulum was shortened, the mass and the moment of inertia were decreased). A parameter optimization method was used to fit the CPx power spectrum given by the model to that obtained experimentally. The results indicated that the changed anthropometric parameters in KN would indeed cause a large decrease in the power spectrum at low frequencies. However, the model fitting also showed that there were considerable changes also in the neural subsystem when the subject went from standing to kneeling. There was a lowering of the proportional and derivative gains and an increase in the neural noise power. Additional increases in the neural noise power were found also when the subject closed his eyes.

Dynamic equilibrium through reinforcement learning

Reinforcement Learning is an area of Machine Learning that deals with how an agent should take actions in an environment such as to maximize the notion of accumulated reward. This type of learning is inspired by the way humans learn and has led to the creation of various algorithms for reinforcement learning. These algorithms focus on the way in which an agent’s behaviour can be improved, assuming independence as to their surroundings. The current work studies the application of reinforcement learning methods to solve the inverted pendulum problem. The importance of the variability of the environment (factors that are external to the agent) on the execution of reinforcement learning agents is studied by using a model that seeks to obtain equilibrium (stability) through dynamism – a Cart-Pole system or inverted pendulum. We sought to improve the behaviour of the autonomous agents by changing the information passed to them, while maintaining the agent’s internal parameters constant (learning rate, discount factors, decay rate, etc.), instead of the classical approach of tuning the agent’s internal parameters. The influence of changes on the state set and the action set on an agent’s capability to solve the Cart-pole problem was studied. We have studied typical behaviour of reinforcement learning agents applied to the classic BOXES model and a new form of characterizing the environment was proposed using the notion of convergence towards a reference value. We demonstrate the gain in performance of this new method applied to a Q-Learning agent.

Analysis of ground reaction force and electromyographic activity of the gastrocnemius muscle during double support

O documento em anexo encontra-se na versão post-print (versão corrigida pelo editor).

Estudi: microcontrolador LPC2119 - ARM. Cas pràctic: anàlisi, disseny i construcció d'un robot amb pèndul invertit

Estudi de l'arquitectura i prestacions del microcontrolador LPC2119 tot implementant la proposta d’un cas pràctic. En la besant teòrica, es fa una anàlisi acurada del dispositiu LPC2119, enumerant les principals característiques i exposant les seves parts, aprofundint sobretot en l’arquitectura i core ARM que incorpora. En l'àmbit pràctic, s'introdueix el problema del pèndul invertit com a proposta per a ser integrada sobre un robot que exploti les funcionalitats del dispositiu integrat presentades a l'estudi teòric.

Disseny i construcció d'un pèndol invertit

Aquest projecte consisteix en l’estudi del problema clàssic del pèndol invertit en l’àrea de control. Per tal de poder realitzar aquest estudi s’ha construït un prototip que simuli el comportament d’un pèndol invertit. Seguidament es dissenyen uns controladors PID i LQR per aquest prototip. Finalment s’escull el controlador LQR, que és per al qual es realitzen les simulacions i es programa el prototip real.

Preliminary results from a real time control system using optical information

Successful results from training an adaptive controller to use optical information to balance an inverted pendulum are presented in comparison to the training requirements using traditional controller inputs. Results from research into the psychology of the sense of balance in humans are presented as the motivation for the investigation of this new type of controller. The simulated model of the inverted pendulum and the virtual reality environments used to provide the optical input are described The successful introduction of optical information is found to require the preservation of at least two of the traditional input types and entail increased training time for the adaptive controller and reduced performance (measured as the time the pendulum remains upright).

Results of applying sensor fusion to a control system using optic flow

The results from applying a sensor fusion process to an adaptive controller used to balance all inverted pendulum axe presented. The goal of the sensor fusion process was to replace some of the four mechanical measurements, which are known to be sufficient inputs for a linear state feedback controller to balance the system, with optic flow variables. Results from research into the psychology of the sense of balance in humans were the motivation for the investigation of this new type of controller input. The simulated model of the inverted pendulum and the virtual reality environments used to provide the optical input are described. The successful introduction of optical information is found to require the preservation of at least two of the traditional input types and entail increased training till-le for the adaptive controller and reduced performance (measured as the time the pendulum remains upright)