Research On Design Of Control System And Environment Modeling For A Prototype Of Mobile Robot

This dissertation is supported by the key project of the National Natural Science Foundation of China under grant no.60234030, Research on Theories and Methods of Navigation Control for Mobile Robots under Unknown Environments. As one part of the project, this dissertation aims at designing and developing a prototype control system in order to provide an experimental platform for probing special issues on the navigation of mobile robots under unknown environments. The prototype system is the reduction and condensation model of all kinds of mobile robots that pays emphasis on control architectures and control strategies. So the study in this dissertation focuses on two key points as follows:(1) An open architecture of software and hardware is the base of extendable intelligence of a mobile robot. Designing an experimental platform with better expansibility is the requirements of going deep into researches and expanding the system function. This paper proposes a distributed control architecture of a mobile robot prototype which realizes the reactive intelligence, deliberative intelligence and global environment modeling with cooperation among the layered software agents in parallel and asynchronously.(2) Spatial representation provides a frame of organizing and fusing sensors data. A cognitive map is the body of knowledge of a large-scale environment that is acquired by integrating observations gathered over time. In order to realize the coordination of real-time navigation and the complex environment modeling or global planning in a large-scale environment, this paper presents a method of spatial knowledge representation through adopting two layers for perception and feature respectively. The intuitive metric representation is adopted in the perceptual layer to fuse the multi-sensor information and provide the direct support of real-time navigation, while topological environment model is used in the feature layer to describe the skeleton feature of free area and realize the global planning and decision in large-scale environment.Considering the control performance and the intelligence among reactive behavior, dynamic deliberation and global cognition, several aspects around these key problems are studied in this dissertation, and main contribution and work are described as follows:A prototype system of mobile robots is designed and developed. An open architecture of distributed control system based on network is provided for deep research and development. It realizes a three-layer system frame composed of feature integration, deliberation and control layers corresponding to spatial representation type respectively. The navigation control in real time is implemented with the cooperation of control layer and deliberation layer, and the spatial knowledge is expanded through environmental modeling in the feature layer.For lateral stabilization in path tracking control, the state feedback equation is constructed in polar coordinates instead of in the global coordinates, and the back-stepping technique based on Lyapunov law is adopted to design the controller toward acceleration control.? A hybrid strategy combined the reactive control and deliberative planning for the mobile robot navigation is researched. A local planning method based on simulated annealing and a Brown motion model of disturbance is provided. The real-time information from a laser scanner is used to optimize the local planning in the rolling window and an improved inverse D* algorithm for deliberative planning is presented to search the middle object so as to approximate the global goal and decrease the blindness of global search under unknown environment.? A 3-D perceptual system based on a laser scanner under unstructured environment is designed. It utilized a 2-D laser scanner mounting on a high precision rotating platform with horizontal and pitch rotation to measure the 3-D terrain. By analyzing the cause of system error and sensor noise, the dynamic adaptive filter is presented to obtain the local environmental information and to reconstruct the 3-D terrain map so as to judge the free area and obstacle area under unstructured environment.? An environment modeling method for mobile robots based on approximate Voronoi boundary network (AVBN) is proposed. It relies on the local sensed information to build the topological map of traversable area. This method builds an environment model with less nodes compared with Generalized Voronoi Graph, therefore the complexity of path planning is reduced largely. At last, the genetic algorithms (GAs) based on the Elitist rule is applied to plan the global path of AVBN model.