Navigation And Localization Technique For Dust Absorption Robot Based On 3D-reconstruction

Automaton technology development is the comprehensive result of present science and technology progress. It has a great effect on the developing of our economy and society, which involves mechanics, electronics, automation, computer-measuring, artificial intelligence, sensing technique, etc. Dust absorption robot is also named as cleaning robot or auto-dust-collecting machine or intelligent dust sucker. It is a popular researching subject in the field of electronic appliance. There has been dust absorption robot with high intelligence sold in the appliance market, but its price is surprisingly high, which is out of reach of ordinary consumers. Meanwhile, those imitated products with relatively lower price, yet, lack humanity in the too complicated, immature design. Most researchers adopt infrared and sonar as the sensor to collect surrounding environmental information. Studies on visual-sensor application in designing dust absorption robot are still on its way and considered as a challenging subject. This thesis makes an attempt to propose a navigation and localization technique for dust absorption robot based on 3D-reconstruction.The paper consists of three major parts: localization technique, map-building technique and path-planning technique. With theoretical discuss and experimental analysis, the following aspects are explored.1. The establishment of localization in different path-planning methodsBased on the mode of path-planning and characteristics of the visual sensor adopted for dust absorption robot in this paper, varieties of localization techniques are put forward after briefly introducing the present classification and development of localization technique. When searching for the best path-planning, wheel pulse simple localization combined with 2-feature-point accurate localization is adopted; when searching for the complete path-planning, line-controlling is chosen based on deviate angle to assure the robot move along a straight line. Simultaneous localization and map-building are as well discussed afterwards.2. Rectangle map-storing and Map-building based on searching array used for localizationIn this part, methods of map-building and their advantages and disadvantages are explained with the proposal of rectangle map-storing and map-building. Further explanations are given about map expression, how to build a map, map revision and perfection, data-storing, array-searching structure and rectangle map. Rectangle calculation is easily to be revised and rebuilt. When it is combined with searching array that can facilitate the finding of feature points, this computing method turns to be closer to perfection by the mutual compensation. It is a key point to make a decision on the size of rectangle, which is crucial to the accuracy of path-planning calculating methods. Therefore, a stimulating experiment is carried out in this paper, from the result of which the following conclusions are drawn. The smaller the rectangle is chosen to be, the higher the environmental distinguishing rate is, accompanying with the deduction of anti-disturbing capacity, larger storage of environmental information, slowness in decision-making speed. Vise verse, the larger the rectangle is, the anti-disturbing capability increases, the storage of environmental information is larger; the process of decision-making is swifter; however, there must be a concomitant decrease in distinguishing rate, i.e. the dust absorption robot is slower in seeking the path among intensive obstacles. In accordance to the above analysis, a 5*5cm rectangle is selected in the present thesis.3. On the basis of several complete coverage navigation path-planning methods, this paper attempts to improve the existing calculation better matches this system. And a stimulated experiment is implemented in the environment of C++ Builder6. Results from the experiment shows that the designed path-planning calculation achieved its expectation with a 95% coverage rate and 10% repetition at least. Besides the calculation itself, other factors like size of the house, complexity of the obstacles, distribution intense, etc. are also affected for the coverage rate. The path-planning under regional cleaning mode is equal to the path-planning of general mobile robots. This thesis integrates the global path-planning and the regional path-planning methods, which makes the shortest path and increases the flexibility of path-planning.4. This paper first makes a design of an entire systematic program. A new moving mechanism is proposed as two-wheel driving with back wheel following. This design is convenient for robot controlling. With an investigation into the theories,features and the choice-making principles of step motor,a suitable step motor is selected for the dust absorption robot. After giving a brief explanation to the sensor and dust-absorbing equipment, focus is put on exploring the hard ware electric circuit in controlling unit and the systematic design. A double CPU mode of DSP+FPGA is preferred in this research. FPGA functions in primarily processing the sensor information and its 3D-reconstruction, which is to be passed to DSP. DSP acts as the core of the controlling unit and is responsible for the domination of varieties of parts and the realization of designed calculations.This thesis is composed with the purpose of exploring vision- sensing-technology-based dust absorption robot, reducing the production cost, popularizing dust absorption robot among ordinary consumers, better serving the people and making a little contribution to the development in this researching field.