Aiming at the problems faced by robotic arms in realizing accurate grasping in complex environments

such as difficulties in data processing and feature extraction

excessive state space dimensionality

and low success rate of grasping

a robotic arm grasping method based on image quantization and sequence modeling is proposed.The acquired image information is quantized using a vector quantized generative adversarial network(VQ-GAN)

which maps the pixel values in the image to a finite number of discrete levels

reduces the data dimension

and decreases the complexity of the decision space.The quantized state images and action sequences are modeled using the Transformer model

which effectively establishes the correlation between images and improves the accuracy of action prediction.The experimental results show that compared with other end-to-end grasping methods

the proposed method significantly improves the success rate

with the highest reaching 93%;and it enables the robotic arm to understand the environment and target features more accurately

and complete accurate grasping in complex environments.