miniAuto Pro
An open-source AI-powered smart car developed by Hiwonder, integrating the ESP-Claw agent module, an AI voice assistant, and a visual large language model. It supports natural-language instruction control of mechanical actions and is designed for embedded developers and AI enthusiasts.
Timeline (last 90 days)
Hiwonder's miniAuto Pro is an open-source robot platform that integrates the ESP-Claw agent, the AI voice assistant "Xiao Huan," and the robot body.
miniAuto Pro can parse continuous complex natural language instructions and convert them into a series of precise mechanical action sequences.
miniAuto Pro's multi-step task decomposition capability is based on the Function Calling mechanism of large language models.
miniAuto Pro can combine real-time data from ultrasonic sensors to perform if-else conditional branch judgments; in actual testing, it detected a distance of 52 centimeters and correctly judged it to be greater than 30 centimeters and executed the corresponding action.
miniAuto Pro delegates conditional judgment at the decision layer to the large model for dynamic generation, so developers do not need to exhaustively enumerate all judgment branches in advance and only need to describe the rules in natural language.
miniAuto Pro is connected to a visual large model with scene description and object recognition capabilities, and can recognize panda, zebra, small tree, and Eiffel Tower models on a desktop.
miniAuto Pro can combine visual recognition with mechanical claw control to complete the full process of recognition, grasping, and transportation.
All Facts (7)
Hiwonder's miniAuto Pro is an open-source robot platform that integrates the ESP-Claw agent, the AI voice assistant "Xiao Huan," and the robot body.
50%UnverifiedminiAuto Pro can parse continuous complex natural language instructions and convert them into a series of precise mechanical action sequences.
50%UnverifiedminiAuto Pro's multi-step task decomposition capability is based on the Function Calling mechanism of large language models.
50%UnverifiedminiAuto Pro can combine real-time data from ultrasonic sensors to perform if-else conditional branch judgments; in actual testing, it detected a distance of 52 centimeters and correctly judged it to be greater than 30 centimeters and executed the corresponding action.
50%UnverifiedminiAuto Pro delegates conditional judgment at the decision layer to the large model for dynamic generation, so developers do not need to exhaustively enumerate all judgment branches in advance and only need to describe the rules in natural language.
50%UnverifiedminiAuto Pro is connected to a visual large model with scene description and object recognition capabilities, and can recognize panda, zebra, small tree, and Eiffel Tower models on a desktop.
50%UnverifiedminiAuto Pro can combine visual recognition with mechanical claw control to complete the full process of recognition, grasping, and transportation.
50%