Autonomous Mobile Robot Control Based On Membrane Computing

Study of the theory and applications of membrane computing can provide a new approach for solving key issues such as intelligent planning,decision-making and control in autonomous walking mobile robots.At present,the theoretical research of membrane computing has been fruitful,and there is an urgent need for breakthroughs in the research of applications.Membrane systems(or P systems as they are generically known)are parallel distributed computing models mathematically abstracted from functioning of the living cells.It has powerful information processing and computing capabilities.Also,it is suitable for solving the motion planning and control problems of mobile robots.In this paper,the problems faced by the autonomous walking mobile robots are solved with the characteristics of membrane computing.A membrane optimization algorithm and a variety of membrane controllers are designed to improve the environmental adaptability of mobile robots when they walk autonomously.This paper first describes the characteristics of the information processing in membrane computing,and analyzes the reasons why membrane system is suitable to solve the key problem of autonomous mobile robot.Based on the comprehensive analysis of the commonness between the cognitive system of general intelligent agent and P systems construction,we proposed the hybrid control architecture of the autonomous mobile robots with P systems which is suitable for different types of applications.The specific problems in application are solved by the P systems and are discussed in different control layer,which lays the foundation of the application framework of P systems for the subsequent research.Aiming at the problems of slow convergence,weak local environment detection ability and difficulty in giving consideration to efficiency and effect,a particle swarm optimization algorithm with variable dimensions is proposed by analyzing the evolution laws of the individual path nodes in optimization process.Membrane rules such as dissolve,communication,transport,have been utilized fully to compute the auxiliary functions in the point repair algorithm,smoothing algorithm,direction adjustment algorithm,and realize the optimization of particle population dimension change and information exchange.This kind of multi-dimensional population can improve the searching efficiency of PSO.The evaluation and decision making methods in case of multiple targets are defined to produce more reasonable paths which can accelerate algorithm convergence and improve its adaptability.The trajectory tracking problems often face external disturbances,dramatic changes in parameters,and difficulty in accurate modeling.A two-layer trajectory tracking controller based on kinematics model and dynamic model is designed.At the kinematic layer,the kinematic tracking control law combining feedforward and feedback is designed in sections,which provides a more accurate reference path input for the dynamic model.At the dynamic layer,P systems are used to regularize the control model of the neural network PID.In parameter self-learning process,the combination of neural networks and expert knowledge in membrane regions,are realized by utilizing the characteristics of the enzyme variables.This flexible switching method can weaken the influence between control parameters,and achieve the effective controlling for strong time-varying disturbance systems.Aiming at the problems of easy oscillation and falling into the minimum trap,while real-time navigation,local environment mode classifier,multi-behavior selection strategy and multi-behavior coordination membrane controller are designed respectively.While autonomous robots can explore the unknown environments,accurate understanding of the environments is conducive to make more correct behavioral response.Various local environmental patterns are also defined in binarized way to reduce the effect of the sensor noise.Moreover,a P system is introduced to design the environmental classifier,which can accurately and rapidly realize the identification of environmental patterns.In order to facilitate the coordination among multiple behaviors,the physical characteristics of the robot,the control laws of goal reaching,obstacle avoidance,wall following and corridor walking were designed respectively.In this paper,a multi-behavior coordination strategy is proposed to get rid of the local minimum trap.The multi-behavior coordination membrane controller is designed to help the mobile robot to get out of the complex maze successfully with excellent performance.A mobile robot experimental platform based on P system is built.Several groups of experiments verify that the membrane controllers proposed in this paper have satisfactory results in motion planning and motion control.