Possibility of developing the the Bittle X control integrated with LLMs

Hi Guys. I am a senior student. My graduation design teacher requires me to deploy a large model based on the Petoi Bittle X platform. According to the content input by voice, the large model will give detailed instructions after analysis, and the visual module will realize environmental interaction to clamp and pick up the required workpieces, ultimately achieving intelligent handling. I think there are many difficulties in this process: 1. The deployment of a large language model at the edge. I can think of two ways: deploying it on a host computer such as a Raspberry Pi or directly calling an API over the network. 2. The design of the gripper. I saw that you posted a video on Bilibili called “Robot Dog Mining Race” which uses Bittle R. This kind of robotic arm may be helpful for my work. Can this kind of robotic arm be purchased as an accessory and installed on Bittle X? 3. I currently have Intelligent Camera Module. May I ask if there is any way to deploy algorithms such as YoLov8 on this ESP32-based module? I will try to solve the remaining difficulties myself first.

Dear developers, with your knowledge, as a beginner in the fields of embedded development, large language models, and image detection, and relying on my own knowledge of the internet and with the help of the forum, am able to give a finished result, really according to the large language model to intelligently transport goods? I would like to hear your opinion, and if it is indeed beyond my ability, I may be able to quit.