안내 및 배달을 위한 지능형 서비스 로봇의 자율 주행 기술 개발 및 서비스 로봇을 위한 진화 연산과 신경망 기반의 동시 위치 인식 및 지도 작성에 관한 연구

Abstract

This paper addresses path planning, navigation, localization and map building method for intelligent service robot for guidance and delivery service. First, we propose a path planning algorithm that works in line segment-based map. Portals which serve as passages for robot’s movement are generated and portal tree is constructed using these portals. By inspecting portal tree, we can search shortest path from start to goal position. Moreover, path optimization techniques are applied to the generated path to assure that the path is reliable for robot’s navigation. Experimental results are demonstrated using third floor map in existing LG building. In addition to this, proposed path planning method is able to cope with unexpected obstacles by replanning the travel path in the real time when the robot’s traveling path is totally blocked because of obstacles. Second, we developed a gap based obstacle avoidance method for safe navigation. For avoiding obstacles, five different behaviors for mobile robot are presented and these behaviors are selected by the robot, depending on the environment. Experiments were carried out in several simulated and real environments including a narrow and complex environment, a dead-end trap, and a rapidly changing situation. The robot was able to navigate in a corridor without collisions. So our method can make the robot navigate in very dense, cluttered, and complex scenarios which are a challenge for many other methods. Third, we address a method for improving accuracy of Simultaneous Localization and Mapping (SLAM) by compensating for odometric error of the robot. The Neural Network (NN) is used for estimating the odometric error and online learning of NN is implemented by augmenting the synaptic weights of the NN as the elements of state vector in the EKF-SLAM process. Due to this trainability, the NN could adapt to systematic error of the robot without any prior knowledge and the proposed NN aided EKF-SLAM is very effective compared to the standard EKF-SLAM method under the colored noise or systematic bias error. Experimental results are presented to validate that our NN aided EKF-SLAM generates more accurate feature map than conventional EKF-SLAM. Lastly, we presents a novel approach to SLAM, called NeoSLAM (Neuro-Evolutionary Optimization SLAM) that can implicitly learn by itself not only nonlinear motion model but also the noise statistics from the actual measurement data by employing a neural network. It casts SLAM as a global optimization problem in which the objective is to maximize the quality of robot trajectory and the feature positions in the world frame. In our algorithm, the neural network is trained via evolutionary optimization (a kind of stochastic optimization) to learn the residual motion model or motion error which is then added to the odometry motion to form the full motion model estimate. Then, we applied proposed NeoSLAM method to generating a feature map from sonar readings. The various experimental results demonstrate that the neural network based SLAM guarantees a consistent environmental map under the sonar readings which in general are known to have poor bearing accuracy and resolution.