INTELLIGENT CONTROL OF ROBOT MANIPULATORS USING SOFT COMPUTING TECHNIQUES

Abstract

The rapid development of factory automation involves industrial robot manipulators, which are mostly used in manufacturing processes to increase the productivity and quality of the products. The current thrust of research in robotics is to control robots which can operate in dynamic and partially known environments. Control of an industrial robot includes nonlinearities, uncertainties and external perturbations that should be considered in the design of control laws. Most non-adaptive control schemes for the control of robot manipulator systems, usually assume a full knowledge of the system dynamics. This is an unrealistic assumption in many cases as these complex systems are subjected to the presence of uncertainties. If not dealt appropriately, these uncertainties may have a remarkable effect on the controller performance and even induce instability. The ability of learning provides the robot manipulator with autonomous intelligence to handle such situations. Recent studies strongly stipulate that the learning based intelligent controllers reduce the adverse effects of time-varying behaviour and unmodeled dynamics when the analytical design is tedious or is not satisfactory. Hence, developing model-free adaptive control structure using computationally intelligent soft computing techniques has become an interesting research topic. The main objectives of the present work are; (a) investigation of the ability of soft computing approaches for addressing various design tasks and issues associated with the control of robot manipulators, (b) development of high performance tracking control algorithms in the presence of time-varying structured and unstructured uncertainties using computationally intelligent soft computing based controllers, (c) integration and hybridization of fuzzy logic, neural network, genetic algorithm, bacterial foraging algorithm, and particle swarm optimization with conventional controllers to overcome each others weakness leading to new approach for solving robotic control problems, (d) investigation of new hybrid adaptive neuro-fuzzy control algorithms for manipulator control with structured and unstructured uncertainties, (e) development ofnovel algorithm based on the foraging behavior of E-coli bacteria and particle swarm optimization to optimize fuzzy precompensated control of two link flexible manipulator, and (f) study of hybrid fuzzy logic based precompensation scheme consisting of fuzzy PD precompensator and aconventional PD Controller for superior steady state and transient performance, good stabilization and tracking performance. Obtaining the joint variables that result in a desired position of the robot endeffector called as inverse kinematics is one of the most important problems in robot kinematics and control. As the complexity of robot increases, obtaining the inverse kinematics solution is difficult and computationally expensive. In this work, using the ability of Adaptive Neuro-Fuzzy Inference System (ANFIS), an implementation of a representative fuzzy inference system using aBP neural network-like structure, the inverse kinematics problem with limited mathematical representation of the system is solved. Computer simulation conducted on two degree of freedom (DOF) and three DOF robot manipulators shows the effectiveness of the developed approach to provide fast and acceptable solutions of the inverse kinematics and thereby making ANFIS as an alternate approach to map the inverse kinematics solution. Tracking control of a six degrees of freedom robot arm (PUMA Robot) using hybrid Fuzzy PD plus conventional Icontroller is carried out as anew approach. The three input and one output fuzzy system is too complex to construct the PID controller. It is very difficult to decide the fuzzy control rules intuitively. Proposed fuzzy controller is mainly focused to mitigate such problems. Complexity of the proposed Fuzzy PD plus conventional I controller is minimized and only two design variables are used to adjust the rate of variations of the proportional gain and derivative gain. Numerical simulation using the dynamic model of six DOF robot arm shows the effectiveness of the approach in trajectory tracking problems. Comparative evaluation with respect to PID and Fuzzy PID controllers are presented to validate the controller design. The results presented emphasize that a satisfactory tracking precision could be achieved using Fuzzy PD plus conventional controller structure when compared to their respective fuzzy PID only or conventional PID counterparts. To improve further, the hybrid Fuzzy PD plus conventional I controller is extended with hybrid Adaptive Neuro Fuzzy PD plus I controller. Simulation results show the usefulness of the approach in terms of better tracking than the Fuzzy PD+Icontroller. The performance of fuzzy logic controllers for controlling multi input - multi output and complex systems strongly depends on the scaling factors, membership functions, and rule base parameters. In the present work, Genetic algorithm tuned Fuzzy PID controller (GAFPID) is designed to trace the desired trajectory for a three DOF robot arm. Fuzzy PID controller performs well compared to PD and PID controllers and the problem of tuning of scaling factors of these controllers is overcome by genetic algorithm. Numerical simulation using the dynamic model of three DOF robot arm is carried out and the effectiveness of the approach in trajectory tracking problems is shown to achieve low tracking errors for high speed trajectories. Comparative evaluation with respect to PD, PID and Fuzzy PID controls are presented to validate the controller design in terms of integral absolute error (IAE), integral square error (ISE), and integral absolute time error (ITAE).

Neuro-Fuzzy techniques have been emerged from the fusion of neural networks and fuzzy inference systems form apopular framework for solving real world problems. In this work, utility and effectiveness of soft computing constituents based controllers (neural, fuzzy, and neuro-fuzzy) for the control of six degree of freedom robot arm (Puma 560) with structured and unstructured uncertainties are investigated. Comparative analysis of simulated performance of some conventional controllers (PID control, computed torque control, feed forward inverse dynamics control, and critically damped inverse dynamics control) is presented to validate the controller design. The results obtained show that soft computing based controllers perform better with uncertainties of the dynamics parameters and friction changes during manipulator operation. It is observed that the hybrid neurofuzzy controller performs better than only fuzzy or only neural or only conventional controllers under uncertainties in terms of IAE and maximum errors. Three novel hybrid adaptive neuro-fuzzy control algorithms have been proposed for manipulator control with uncertainties. These hybrid controllers consist of adaptive neurofuzzy controllers and conventional controllers. The output of these controllers is applied to produce the final actuation signal based on current position and velocity errors. Numerical simulation using the dynamic model of six DOF puma robot arm with uncertainties shows the effectiveness of the approach in trajectory tracking problems. Performance indices of root mean square error, maximum error are used for comparison. It is established through simulation results that the hybrid controllers perform better than only fuzzy or only adaptive neuro-fuzzy or only conventional/adaptive controllers. The Bacterial Foraging Optimization Algorithm (BFO) and Particle swarm Optimization (PSO), is currently gaining popularity in the community of researchers, for its

effectiveness in solving certain real world optimization problems. In this work, a novel algorithm based on the foraging behavior of E-coli bacteria and particle swarm optimization to optimize Precompensated Fuzzy PD with proportional velocity feedback control of two link flexible manipulator is developed. The proposed algorithm performs local search through the chemotactic movement operation of BFO whereas the global search over the entire search space is accomplished by a PSO operator. Here, PSO is used for high convergence of the randomly initialized direction vector optimization. In this way it balances between exploration and exploitation enjoying best of both the techniques. The fuzzy logic based precompensation used in the work has superior steady state and transient performance, good stabilization, and tracking performance compared to a conventional PD controller. Numerical simulation using the dynamic model of two link rigid flexible robot arm shows the effectiveness of the approach in trajectory tracking problems. Comparative evaluation of integral square errors with respect to genetic algorithm, particle swarm optimization and bacterial foraging optimization is presented to validate the controller design. The results of the developed intelligent algorithms have been very encouraging and would certainly open new opportunities for dealing with robot control of complex structures in manufacturing industry to increase the productivity and quality.