With the rapid advancement of Industry 4.0, the demand for flexible and intelligent production systems has intensified. Technologies such as 5G communication and digital twins have made manufacturing processes more dynamic, but they also require execution terminals with enhanced environmental adaptability. Traditional robotic systems, including wheeled and tracked robots, often struggle in complex, irregular industrial environments like narrow passages, stairs, or rugged terrain. In contrast, quadruped intelligent robots, inspired by biological locomotion mechanisms, exhibit superior adaptability to complex terrains and multifunctional operational potential. However, current applications primarily treat these intelligent robots as mere mobile platforms, neglecting their intrinsic state monitoring. If a failure occurs in the quadruped intelligent robot’s locomotion system, core operations are disrupted. Additionally, localized control methods for task execution become inefficient when multiple intelligent robots need to coordinate movements. To address these issues, we adopt a cloud-network-device framework and integrate digital twin technology, enabling real-time state observation and remote task distribution for quadruped intelligent robots in smart manufacturing systems.
Our research focuses on developing a comprehensive system where quadruped intelligent robots perform discrete tasks such as inspection and transportation, with their status integrated into a digital twin environment. This approach not only enhances operational efficiency but also lays the foundation for multi-robot collaboration. The system architecture, communication networks, and practical implementations are detailed in this article, providing a scalable solution for intelligent factories. By leveraging digital twins, we achieve synchronized virtual and physical representations, facilitating proactive maintenance and optimized task scheduling for these intelligent robots.
The core of our system involves a hierarchical architecture that decouples the digital twin, wireless communication, and the quadruped intelligent robots themselves. The quadruped intelligent robots execute tasks and perform environmental detection, while wireless networks ensure stable data exchange between the robots and the digital twin system. The digital twin handles task scheduling, video display of the work environment, and data analysis. Communication between the digital twin and the intelligent robots utilizes the WebSocket protocol, with the digital twin acting as the server and the robots as clients. Control commands are sent downstream, and status data, including position, posture, and task progress, are uploaded upstream periodically by the intelligent robots. For video streaming, the RTSP protocol is employed to transmit real-time environmental feeds to the digital twin, allowing remote monitoring by operators.
In designing the digital twin system, we create virtual models that mirror the physical geometry of the quadruped intelligent robots, enabling dynamic mapping through real-time data acquisition and bidirectional interaction. This system incorporates key functionalities such as perceptual synchronization, 3D modeling, behavior simulation, data analysis, and remote control. The front-end interface, built with Vue 2, uses component-based structures and responsive interactions to display device status and facilitate user control. Web-based 3D visualization, powered by Babylon.js, renders the positions, postures, and action trajectories of the quadruped intelligent robots. Models are created in Blender and exported in GLB format, with WebSocket ensuring real-time data synchronization. The backend processes data from physical devices, supporting functions like operation monitoring, video surveillance, and task management. This integrated approach forms a closed-loop system from perception to feedback, enhancing the adaptability of intelligent robots in industrial settings.
Wireless communication is critical for reliable operation. We deploy a local area network (LAN) where the quadruped intelligent robots serve as servers, broadcasting sensor data and status via WebSocket. Front-end clients, running in browsers, establish stable connections to receive and render this data, while also sending control commands. To ensure coverage across large areas, we use multiple wireless access points (APs) to extend signal range and reduce network latency. For instance, in a 200 m × 50 m factory setup, APs like the TP-Link AX3000 are employed to maintain connectivity as the intelligent robots move. In cases where the robot and digital twin operate on different subnets, we implement IP static routing to bridge the networks. For example, if the digital twin’s IP segment is 192.168.3.0 and the robot’s is 192.168.2.1, we use commands such as sudo ip route add 192.168.3.0/24 via 192.168.2.1 dev wlan0 to enable data routing. Video streams are forwarded using tools like socat, ensuring seamless integration.
The application scenario involves deploying quadruped intelligent robots in a factory environment for automated inspection tasks. These intelligent robots, with specifications including a standing size of 1000 mm × 715 mm × 470 mm, weight of 60 kg, and maximum speed of 1.5 m/s, can navigate obstacles up to 20 cm high and slopes up to 40 degrees. They are equipped with sensors such as cameras, IMUs, and LiDAR for perception. In practice, an inspection task covering the factory area takes approximately 8 minutes, demonstrating efficiency gains. The digital twin system displays real-time status information, as shown in the table below, which includes metrics like battery level, joint temperatures, and task progress. This setup not only automates inspections but also provides insights for predictive maintenance of the intelligent robots.
| Robot Type | Terrain Adaptability | Max Load (kg) | Typical Speed (m/s) | Key Applications |
|---|---|---|---|---|
| Wheeled Intelligent Robot | Flat surfaces only | >100 | 1.0-2.0 | Material handling in warehouses |
| Tracked Intelligent Robot | Mud, sand, slopes up to 45° | 100-500 | 0.5-1.5 | Heavy-duty transport in harsh environments |
| Quadruped Intelligent Robot | Stairs, rugged terrain, narrow spaces | 20-30 | 1.5 | Inspection, maintenance, and logistics |
To model the locomotion of quadruped intelligent robots, we use dynamics equations based on reinforced learning and control algorithms. The general form of the motion equation can be expressed as:
$$ \tau = M(q) \ddot{q} + C(q, \dot{q}) \dot{q} + G(q) $$
where \( \tau \) represents the joint torques, \( M(q) \) is the inertia matrix, \( C(q, \dot{q}) \) accounts for Coriolis and centrifugal forces, and \( G(q) \) is the gravitational vector. For task scheduling, we implement optimization algorithms in the digital twin to minimize execution time. The objective function for task allocation among multiple intelligent robots is:
$$ \min \sum_{i=1}^{n} \left( t_i + \alpha \cdot d_i \right) $$
where \( t_i \) is the time for robot \( i \) to complete its task, \( d_i \) is the distance traveled, and \( \alpha \) is a weighting factor. This ensures efficient coordination of the intelligent robots.
In terms of state monitoring, the quadruped intelligent robots periodically upload data to the digital twin. The table below summarizes key state variables collected during operation, which are crucial for real-time diagnostics and performance analysis of these intelligent robots.
| State Category | Specific Metric | Value | Unit |
|---|---|---|---|
| Task Information | Current Task ID | 5 | – |
| Total Tasks Executed | 7 | – | |
| Inspection Progress | 71.43 | % | |
| Power and Mobility | Battery Level | 36 | % |
| Current Mileage | 7940 | cm | |
| Total Mileage | 2635794 | cm | |
| Joint Temperatures | Left Front Swing Joint | 48.06 | °C |
| Right Front Hip Joint | 47.52 | °C | |
| Left Rear Knee Joint | 61.33 | °C | |
| Pose and Velocity | Roll Angle | 0.0 | rad |
| Yaw Angle | 1.682 | rad | |
| Linear Velocity (x) | 0.285 | m/s |
Our implementation demonstrates that integrating quadruped intelligent robots with digital twins significantly enhances the flexibility and intelligence of manufacturing systems. By addressing network challenges through static IP routing and wireless extenders, we ensure reliable communication. The use of WebSocket and RTSP protocols facilitates real-time data flow, enabling the digital twin to act as a centralized coordinator for task distribution and status monitoring. This approach reduces human intervention and operational costs, as seen in the automated inspection tasks that complete efficiently without manual guidance.
Looking ahead, several key technologies require further development to maximize the potential of quadruped intelligent robots in smart manufacturing. First, improving battery life is essential; solutions like wireless charging stations, dynamic power management, and advanced battery materials (e.g., sodium or solid-state batteries) can extend operational time. Second, enhancing human-robot interaction through models like Ultralytics YOLOv11 for gesture recognition and microphone arrays for voice control in noisy environments will make these intelligent robots more intuitive to use. Third, safety mechanisms must be strengthened, including tactile e-skin for collision detection, LiDAR-based obstacle avoidance, and multi-layer battery protection to prevent hazards. Fourth, advancing multi-robot autonomy through distributed decision-making will enable collaborative task execution, such as coordinated inspections or transport. Finally, achieving localization of core controllers, currently dominated by Intel and NVIDIA architectures, is vital for technological independence and supply chain security. These advancements will drive the evolution of intelligent robots toward higher levels of autonomy and integration in industrial automation.
In conclusion, our work presents a practical framework for deploying quadruped intelligent robots in smart manufacturing systems using digital twin technology. By enabling real-time state monitoring and remote task management, we address existing limitations and pave the way for scalable multi-robot applications. This research not only boosts the intelligence of industrial processes but also serves as a reference for building smart factories. As we continue to refine these systems, the role of quadruped intelligent robots is expected to expand, contributing to more adaptive and efficient manufacturing environments.