Humanoid Robot Embedded Vision System And Decision-making System

The humanoid robot is one of the hot issues on artificial intelligence and automation field research in recent years. the humanoid robot has a certain decision-making and operational capacity, it can autonomous or semi-autonomous complete a series of instructions. As its characteristics, such robots can help the human to complete a number of dangerous or heavy work. It has been applied to domestic service and extinguish fire, aerial work, deep exploration, defense and military and other fields.This paper mainly research and design a humanoid robot embedded vision systems and decision-making system with HIT-Ⅱ humanoid robot. This article firstly analyzes the HIT-Ⅱ humanoid robot hardware and software architecture. The embedded vision systemof HIT-Ⅱ humanoid robot is based on camera and PDA. Ths system obtain the image information by the camera installed in the head. Then the system pass the image information to a PDA for processing. Ultimately the system achieve the purpose of the perception of the surrounding environment and identify the target that provide information on support for the robot to make decisions. Decision-making system is based on MR3024controller. The system enables the robot to quickly find the ball, close to the ball, and complete shooting action when shooting. The system also enable the robot found the ball earlier in the goalkeeping and make a pre-sentence.These two systems work together. They make the robot to complete the image processing and targeting, and then complete the goalkeeping and shooting tasks in the robot soccer game.Embedded vision system is the basis of the whole system, its overall performance of the robot plays a decisive role. The system has achieved the target pixel extraction, image segmentation and object positioning. In order to meet the requirements of the game, the system has a strong real-time and accuracy, this visual system uses the HSI color space. And it use a recognition algorithm based on K-means clustering and a targeting algorithm based on monocular vision algorithm. This system has a good experimental results.Decision-making system is the core of the the whole system. Using the visual and location information provided by embedded vision system, the decision-making system make decisions and make action plansfor the robot. Decision-making system is designed based on finite automata theory. According to the task the robot needs to do in the game, This article summed up the transformation process of the state and make up the finite automata model.Finally, according to the finite automata model, This article design a corresponding algorithm to help the robot to make decisions and complete the action plan to make the robot to successfully complete the appropriate action in the game.