AI Large Models + SLAM: When a Wheeled Robot Learns to "Think" & "Navigate"
Hiwonder JetAuto wheeled robot is more than just a device that follows commands—it's an intelligent agent capable of perception, comprehension, and decision-making. By integrating AI large models with SLAM navigation technology, it truly combines the ability to “think” with the skill to "find its way.”
In the past, if we wanted a robot to “go to the soccer field to play ball,” we might have needed to manually set coordinates, plan paths, and adjust its posture. Today, by incorporating multimodal AI large models, JetAuto can directly understand users’ voice commands and break down tasks at a semantic level, interpreting the intent behind them. Whether it's “Go to the zoo to see what animals are there” or “Head to the space center and look for items,” it can translate these casual, scenario-based instructions into specific navigation targets and action sequences—just like a robotic companion that understands human language and perceives its surroundings.
Deploying Multimodal AI Large Models
JetAuto supports the deployment of various mainstream AI large models, including DeepSeek, Qwen... and can flexibly call them via APIs such as Alibaba Cloud. Whether it's text generation, language translation, scene understanding, or object recognition, it relies on these "brains" to achieve high-level human-robot interaction and task execution. Combined with a built-in six-microphone array that enables clear voice capture and sound source localization, JetAuto can engage in smooth, natural voice conversations, interacting with you while carrying out tasks—making human-robot collaboration more intuitive and intelligent.
💡Everything for JetAuto robotic tutorials and STEAM education, all in one place.
JetAuto Takes You to the "Zoo" to See Animals
Imagine this scenario: You say to JetAuto, “Go to the zoo and see what animals are there.” A traditional robot might require you to preset the coordinates for the "zoo," but JetAuto is different—it genuinely understands that "zoo" refers to a location and independently initiates the SLAM mapping and navigation process. While moving, it uses lidar and a 3D camera to perceive the environment in real time, build a map, and dynamically plan its route. Upon reaching the "zoo" area, JetAuto doesn’t stop working—it performs in-depth semantic analysis of the scene through visual large models, recognizing and understanding what "animals" are in front of it, and might even tell you via voice interaction: “I see giraffes and zebras.” This complete loop from voice command to scene understanding to interactive feedback is exactly the kind of "embodied intelligence" experience made possible by the fusion of AI large models and SLAM.
When SLAM Mapping Meets AI Visual Understanding
SLAM technology allows robots to build maps in real time and locate themselves in unknown environments—this is the foundation of a robot's ability to "navigate." JetAuto takes this further by using its 3D depth camera and lidar to capture environmental data, which is then fed in real time to visual large models for deep semantic analysis. This means it can not only build a geometric map but also understand the "meaning" of each area on that map—here is the "zoo," there is the "space center," and over there is the "soccer field." This ability to perform multi-point navigation by understanding user voice commands through language models transforms navigation from a cold, coordinate-based movement into truly intelligent, context-aware cruising.
More Than Navigation: A Starting Point for Embodied Intelligence
From 3D spatial recognition and tracking to item identification and scene understanding based on visual large models; from multimodal autonomous patrolling and color tracking to expandable robotic arms for grasping and sorting—JetAuto is turning embodied intelligence from theory into practice. It not only offers a complete development experience for ROS learners, from beginner to advanced levels, in educational settings but also provides an experimental platform in the frontier field of AI-robotics integration—one that can perceive, reason, and act.