With the rapid advancement of bionic robotics, a wide variety of bionic robots have emerged. Combining biological and robotic characteristics, bionic robots have shown promising application prospects in environments unsuitable for human tasks, such as counter-terrorism, aerospace, military reconnaissance, resource exploration, underwater detection, medical examination, and disaster relief. Existing bionic robots primarily employ wheeled, legged, serpentine crawling, or novel material-driven propulsion methods. Among these, legged robots are widely used due to their strong adaptability to terrain. However, most current legged bionic robots simplify the degrees of freedom (DOFs) in legs and torso during design, failing to closely mimic the contours of their biological counterparts and focusing only on functional biomimicry. Consequently, these bionic robots are suitable only for regular terrain and exhibit weak adaptability to complex environments. Moreover, their structural differences from biological models prevent them from blending into operational environments, limiting their ability to perform tasks requiring high camouflage and concealment, such as reconnaissance and monitoring. This significantly narrows the scope of bionic robot applications.
Therefore, this paper proposes a highly biomimetic robot inspired by the salamander, an organism known for its adaptability to complex environments. We design a bionic robot with multi-DOF legs and torso, aiming to achieve both structural and functional fidelity. From structural design, force analysis, fabrication, assembly, to prototype testing, we systematically validate the correctness of the robot’s design. The development of such a bionic robot enhances the potential for deployment in diverse, unstructured settings.
The mechanical structure forms the foundation of any robotic system. To ensure excellent dynamic characteristics and biomimetic structure, we conducted a thorough analysis of salamander morphology and locomotion mechanisms based on bionics theory. This analysis informed the overall contour and dimensions of the robot, matching the salamander’s head, torso, and tail proportions. Observing salamander crawling, we determined a total of 25 DOFs for the robot: each leg employs a 4-DOF structure, the head has a transverse swinging DOF, the torso uses 5 DOFs to enable body twisting, and the tail, which exhibits transverse and longitudinal swings during rapid crawling, is simplified to 2 active transverse swinging DOFs for easier control.
According to the salamander’s structure and movement principles, the bionic robot is modularized into four components: legs, tail, torso, and head. Each module is designed to meet specific functional requirements while maintaining biomimetic aesthetics.
The legs of a quadruped bionic robot are core components for support and locomotion. Their structure must satisfy: (1) sufficient strength and stiffness to support the body and withstand loads; (2) compliance with biomimetic gait requirements, with adequate DOFs and range of motion for crawling; and (3) simplicity in control. Based on these criteria, we designed a serial 4-DOF mechanical leg, with 2 DOFs each at the hip and knee joints. The overall structure comprises four servo motors, a hip joint motor connector, thigh joint connector, and shank joint connector. All components are linked in an open-chain configuration, providing greater motion space and enhanced terrain adaptability. The servo motor parameters are summarized in Table 1.
| Parameter | Value |
|---|---|
| Weight | 56 g |
| Operating Voltage | 16–7.4 V |
| Dimensions | 40 mm × 20 mm × 39 mm |
| Rated Torque | 55 N·cm |
| Stall Torque | 165 N·cm |
| No-load Speed | 150 mA at 7.4 V |
The torso and tail primarily consist of bionic vertebral connectors, forelimb connectors, hindlimb connectors, tail passive plates, and servos. The salamander vertebral structure comprises 13 joints between forelimbs and hindlimbs. Due to the robot’s compact torso volume, it is impractical to incorporate too many active DOFs; moreover, increased DOFs raise control complexity. Balancing biomimetic functional requirements and control simplicity, the torso employs a 5-DOF structure, with driving servos connected in series via bionic vertebral connectors. The tail uses 2 DOFs to achieve transverse swinging motion.
The head’s external structure is designed according to the salamander’s actual shape, while internally providing space for sensors such as vision, radar, and lidar. This enables functions like obstacle avoidance, photography, and monitoring.
Assembling all designed components yields the complete three-dimensional model of the bionic robot, comprising head, torso, quadruped, and tail modules. This integrated design ensures a high degree of biomimicry both in form and function.
Static analysis is a crucial step in the design process of any bionic robot. Considering the complex shapes of servos, joint connectors, and other small parts, and since the salamander bionic robot features rigid connections, we use rigid links to represent joint structures for force analysis. The front and rear legs have identical structural dimensions and are rigidly connected to the torso. The front leg support points are farther from the torso’s central plane than the rear legs, so we analyze the joint forces of a front leg as an example.
We simplify the front leg structure in its initial state, as shown in the schematic. The assumptions are: (1) The robot’s forward direction is the Y-direction; the hip, thigh, and shank are simplified as links; lateral friction between the foot and ground is neglected. (2) Points A, B, C, D and E, F, G, H represent the four rotational joint positions of the front leg; M and N are ground contact points. The distance from the hip joint to the torso central plane, and the lengths of the hip, thigh, and shank are denoted as L, L1, L2, L3, respectively. (3) Taking leg 1 for analysis, the rotational axes at joints A and C in the current static state are not in the plane, and rotation at these joints mostly occurs during leg lifting motions. Therefore, joints A and C are equivalent to connectors, and we only analyze forces and moments at joints B and D. The simplified single-leg force analysis is depicted.
Performing force analysis on this single-leg state:
For link L1:
$$ \frac{1}{2} m_g g – F_{Bg} + m_1 g = 0 $$
$$ \frac{1}{2} m_g g \times \frac{L}{2} – M_B = 0 $$
For link L2:
$$ F_{Bg} + m_2 g – F_{Dg} = 0 $$
$$ F_{Bg} \times \frac{L_2}{2} \cos \alpha + F_{Dg} \times \frac{L_2}{2} \cos \alpha + M_D – M_B = 0 $$
For link L3:
$$ m_3 g – F_{Dg} + F_{Eg} = 0 $$
$$ F_{Eg} \times \frac{L_3}{2} \cos \beta + F_{Dg} \times \frac{L_3}{2} \cos \beta – M_D = 0 $$
Where:
– \( m_g g \) is the weight of the robot torso.
– \( m_1 g, m_2 g, m_3 g \) are the gravitational forces of the respective links, assuming uniform density.
– \( F_{Bg}, F_{Dg}, M_B, M_D \) are the forces and moments at joints B and D.
– \( F_{Eg} \) is the ground support force at the foot.
– \( \alpha, \beta \) are the angles of the thigh and shank relative to the horizontal plane.
When the bionic robot adopts a diagonal gait, the single leg bears maximum load. The ground support force at the foot of leg 1 is:
$$ F_{Eg} = \frac{m_g}{2} g + (m_1 + m_2 + m_3) g $$
Solving equations (1) to (7) simultaneously yields:
$$ F_{Bg} = \frac{m_g}{2} g + m_1 g $$
$$ F_{Dg} = \frac{m_g}{2} g + m_1 g + m_2 g $$
$$ M_B = \frac{L}{2} \times \frac{m_g}{2} g $$
$$ M_D = \frac{L}{2} \times \frac{m_g}{2} g – \frac{L_2}{2} \cos \alpha \left( \frac{m_g}{2} g + 2 m_1 g + m_2 g \right) $$
Substituting parameters from Table 2 into these equations, we compute maximum values: \( F_{Bmax} = 7.41 \, \text{N} \), \( F_{Dmax} = 8.15 \, \text{N} \), \( M_{Bmax} = 51.04 \, \text{N·cm} \), \( M_{Dmax} = 38.44 \, \text{N·cm} \). These joint torque values are below the servo’s rated torque of 55 N·cm and far less than its peak stall torque of 165 N·cm. Thus, the motor selection for the quadruped bionic robot meets requirements.
| Parameter | Value | Parameter | Value |
|---|---|---|---|
| L (mm) | 1361 | m (g) | 150 |
| α (°) | 15–120 | β (°) | 0–30 |
| L1 (mm) | 60.62 | m1 (g) | 66 |
| L2 (mm) | 73.50 | m2 (g) | 61 |
| L3 (mm) | 75.62 | m3 (g) | 105 |
During the design process, support components and shell parts feature biomimetic designs with complex shapes and numerous curved surfaces. These were fabricated using 3D printing. After assembling each module, we performed overall integration to obtain the robot prototype. The completed bionic robot prototype is controlled via an STM32 board. Legs are labeled: left front (1), right front (2), left rear (3), right rear (4). Considering robustness in complex environments, we adopted a static gait as the primary gait for experimental testing. The stepping sequence permutations total \( A_4^3 = 24 \) types, with different duty cycle coefficients. McGhee’s analysis on gait stability indicates only six statically stable gaits for quadruped robots. Combining the salamander’s crawling leg-lift sequence, we used sequence 4-2-3-1: each leg swings once in this order constitutes one gait cycle.
During experiments, movement speed varies with step length. Setting the periodic movement distance to 1/4 body length, we tested the bionic robot on a platform. Initially, the robot powers into a default state. Then, with legs 1, 2, 3 as support, leg 4 swings forward while the torso bends inward toward the swing side; all leg joints coordinate synchronously to increase step length. Next, with legs 1, 3, 4 supporting, leg 2 swings, and the torso adjusts from bent toward straight. Subsequently, with legs 1, 2, 4 supporting, leg 3 extends forward, torso bends inward. Finally, with legs 2, 3, 4 supporting, leg 1 swings, torso swings back near straight. Repeating this sequence enables crawling. The maximum measured speed, ensuring stability, is 0.08 m/s, limited by servo performance.
We placed the bionic robot in outdoor complex terrain, equipped with ultrasonic and camera modules on the head for testing. During experiments, the robot and all modules operated normally, demonstrating capability for outdoor detection, rescue, and similar tasks. This validates the bionic robot’s practical utility in real-world scenarios.
In summary, employing bionic mechanics, we designed a quadruped bionic robot with highly biomimetic structure and motion mechanisms. Structurally, it features biomimetic head, torso, tail, and quadruped modules, with multi-DOF legs and torso enabling superior biomimetic form and movement. Force analysis provided insight into critical component stresses, guiding motor selection. Fabrication and assembly yielded a functional prototype; motion tests confirmed the robot operates as designed, verifying correctness. Future work may focus on enhancing control algorithms, integrating advanced sensors, and improving adaptability to dynamic environments. This bionic robot represents a step toward more lifelike and versatile robotic systems for challenging applications.
The development of such a bionic robot underscores the importance of holistic biomimicry—not just functional but also morphological. By closely mimicking biological counterparts, robots can better integrate into natural environments, expanding their operational scope. The multi-DOF design allows complex motions akin to living organisms, improving terrain negotiation. Static analysis ensures mechanical reliability, while experimental tests validate real-world performance. Continued research in bionic robotics will likely yield even more sophisticated machines, blurring lines between biological and artificial systems.
Further considerations include energy efficiency, autonomous navigation, and swarm coordination for the bionic robot. Incorporating machine learning could enable adaptive gait generation based on terrain sensing. Material advancements might allow softer, more compliant structures for enhanced biomimicry and safety. The potential applications are vast, from environmental monitoring to search-and-rescue missions where human presence is risky or impossible.
In conclusion, this work presents a comprehensive approach to designing and testing a multi-DOF bionic robot inspired by the salamander. Through detailed structural design, rigorous force analysis, and practical experimentation, we demonstrate a viable platform for high-fidelity biomimetic robotics. The insights gained contribute to the broader field of bionic robot development, paving the way for more adaptive and inconspicuous machines in diverse settings.