Research On Fruit And Vegetable Griping Based On Slippage Detection For Agricultural Robot

When the agricultural robot were used for picking or processing fruit and vegetable, the end effector usually need to grip all kinds of objects. The suitable gripping force should not only be able to keep the fruit and vegetable from slipping, but manage to prevent or reduce the mechanical damage of fruit and vegetable. However, the fruit and vegetable vary in shape, weight and maturity, so the agriculture robot must be able to adjust the griping force according to the operation object in real time. Therefore, this paper took the grasping system of the picking robot as the research platform, the nondestructive grasping techniques were further investigated based on the sensor technology and the adjustment schemes of gripping force. The main research works were as follows:1) The contact relation between a parallel hand and fruit,vegetable was analyzed, then the stable grasping theory and the mechanical damage mechanism of fruit and vegetable were discussed. A kind of new sensitive slippage sensor was developed by using piezoresistor, and a slippage observation system was constructed. The technique of STFT was applied to analyze the spectral information of slippage signal. Then slippage signal was distinguished with the signal resulted from different normal gripping force by using the result of DWT. The experimental results showed that the slippage signal was able to be effectively extracted regardless of different normal gripping force. The detail coefficients of DWT were used to adjust the gripping force of a parallel hand, the experimental results demonstrated the effectiveness of the slippage sensor and the slippage detection algorithm.2) An intelligent controller based on subtractive clustering and self-adaptive fuzzy neural-network was developed to adjust the gripping force. The inputs of the controller were the current griping force and the detail coefficients of DWT when slippage occurred, the output was the closing distance of the end effector. First, subtractive clustering was used to generate a T-S fuzzy model, then the radius of a cluster center was adjusted to select optimal fuzzy rules. At last, a hybrid algorithm consisting of a gradient descent algorithm and least square algorithm was implemented to tune the antecedent parameters and consequent part of the model. As a result, the fuzzy controller obtained only needed4fuzzy rules, which could infer on line and make the robot more flexible. The actual experimental results demonstrated the controller was capable of manipulating unknown fruit and vegetable stably and nondestructively.3) Two slippage sensors were installed on a parallel hand, and a controller based on subtractive clustering and self-adaptive fuzzy neural-network was developed to adjust the gripping force. At last, an actual experiment on grasping tomato was carried out, and the weight loss of tomato after being loaded was studied. The experimental results showed that loading had a certain effect on the internal structure of tomato, but it hadnt a significant effect on the mechanical damage of tomato, which demonstrated the rationality of the controller.