The rapid advancement of robotics technology has fueled an increasing demand for equipment capable of operating in harsh environments. In critical scenarios such as post-disaster search and rescue or resource exploration, robots with high mobility and robust stability are essential for ensuring personnel safety and mission efficacy. Traditional legged robots, particularly hexapods, have garnered significant attention due to their biomimetic design, offering superior adaptability and stability in unstructured terrains compared to wheeled or tracked counterparts. However, the predominant design paradigm for these robots remains a single terrestrial modality, which inherently limits their operational scope. This limitation becomes evident when facing insurmountable obstacles or the need for rapid area scanning, where aerial capabilities would be invaluable. Consequently, there is a pressing need for robotic systems that can transcend the constraints of a single environment. This paper introduces an innovative bionic robot that synergistically integrates the complementary strengths of aerial and terrestrial locomotion. The proposed system is a dual-mode bionic robot, fusing a quadrotor unmanned aerial vehicle (UAV) with a hexapod walking mechanism into a single, cohesive platform.
This integrated design strategy aims to overcome the fundamental limitations of unimodal systems. The hexapod configuration provides stable, high-traversability locomotion over complex, uneven ground, while the quadrotor system enables rapid aerial transit to bypass large obstacles, reconnoiter from elevated vantage points, or access otherwise unreachable areas. The key innovation lies not merely in the combination of two modalities but in the seamless, autonomous transition between them, facilitated by novel mechanical design and intelligent control algorithms. This paper details the comprehensive system design, encompassing the mechanical architecture, hierarchical control system, kinematic modeling, multi-modal gait planning, and the critical mechanisms for mode transition and posture adaptation. Experimental validation using a physical prototype demonstrates the system’s effectiveness in performing tasks across varied and challenging environments, highlighting the practical value of this dual-mode bionic robot for complex field operations.
System Integration of the Dual-Mode Bionic Robot
Overall Structural Design
The dual-mode bionic robot is architected around a central fuselage that houses core electronics, to which both the aerial and terrestrial locomotion systems are attached. This centralized design is crucial for managing the system’s center of gravity (CoG) across different modalities. The terrestrial subsystem is a radially symmetric hexapod. Each of the six legs is connected via a single servo to a protruding boss on the central fuselage. This design minimizes kinematic interference during leg motion and reduces lateral drift during straight-line walking. The leg segments themselves are constructed from lightweight, high-strength carbon fiber with strategic cutouts to minimize mass without compromising rigidity.
The aerial subsystem is mounted atop the central fuselage. Its base plate is constructed from aerospace-grade composite material and is fixed to the fuselage using length-adjustable aviation bolts. This adjustability is critical; it allows the elevation of the entire aerial assembly to a height where the fully extended hexapod legs cannot interfere with the rotors during ground operation. The aerial platform supports the brushless motors, electronic speed controllers (ESCs), a flight controller, and an upper platform for mounting additional payloads like sensors or mapping hardware. Each leg is actuated by three high-torque digital servos (45 kg·cm), providing the necessary force for stable terrestrial locomotion and posture adjustment.
Drivetrain and Control System Architecture
The actuation and control of this complex bionic robot are managed by a sophisticated, distributed architecture. The hexapod’s 18 leg joints are driven by high-precision encoded servo motors. These servos are organized via a serial bus interface, where each leg’s three servos are daisy-chained. This bus-based approach, using a broadcast protocol, offers a significant advantage over traditional PWM control. It enables high-speed, synchronous communication with response latencies below 2 ms and allows the central controller to receive real-time multi-dimensional feedback (position, torque, temperature, load) from each joint at a sampling frequency of 1 kHz. This rich data stream facilitates advanced closed-loop control at the joint level, resulting in highly accurate foot-end trajectory tracking with an error of less than 1.2 mm. The bus design also incorporates hot-swap redundancy, ensuring operational continuity even in the event of a single node failure.
The computational core of the bionic robot is an industrial-grade motor control board centered on an STM32G474VET6 microcontroller (170 MHz). This board implements a hierarchical, multi-tasking control architecture based on a real-time operating system (FreeRTOS). It serves as the nexus for heterogeneous data fusion, integrating commands from the flight controller (received via CAN bus) with data from various onboard sensors. The board features dedicated interfaces for brushless motor drivers (supporting Field-Oriented Control algorithms) and multiple serial ports, which drastically reduces wiring complexity within the central fuselage. Leveraging the microcontroller’s embedded CORDIC co-processor for single-cycle trigonometric operations and FMAC unit for floating-point matrix acceleration, the system achieves remarkably low latency. The computation for inverse kinematics of all leg endpoints is completed in under 0.8 ms, enabling high-frequency gait planning updates at 200 Hz or more. A dynamic priority scheduling strategy within the RTOS ensures deterministic real-time response even under complex operational conditions, with task-switching jitter under 5 µs.
Motion Scheme Design and Kinematic Analysis
Design of the Retractable Rotor Arm
A pivotal innovation enabling the dual-mode capability is the dynamically retractable rotor arm mechanism. To prevent mechanical interference between the sprawling rotor arms and the walking legs during terrestrial operation, the arms are designed to fold. Each brushless motor is mounted on the aerial platform base via a rigid motor mount. A flanged coupler connects the motor shaft to a uniquely shaped carbon fiber arm. The arm’s design and the elevated mounting ensure that at its stowed position, it clears the maximum height reached by the leg joints during walking, with a safety margin.
During mode transition, a dedicated geared DC motor drives the arm through a rotation to its deployed position for flight. An encoder provides real-time positional feedback to a PID closed-loop control system, achieving a rotation precision of approximately 0.1°. Upon reaching the preset deployment angle, an electromagnetic brake is engaged to lock the axial position. This braking is essential to counteract undesired yaw induced by propeller reaction torque during flight, thereby ensuring stable aerial attitude control. This mechanism allows the bionic robot to transform its physical footprint actively, optimizing it for either walking or flight.
Single-Leg Kinematic Analysis Using the D-H Model
The precise control of the hexapod’s locomotion relies on a rigorous kinematic model. The Denavit-Hartenberg (D-H) convention is employed to model a single leg. The leg chain is defined from the hip joint on the body (frame {0}) to the foot tip (frame {4}), with intermediate frames at the coxa ({1}), femur ({2}), and tibia ({3}) joints. The relevant D-H parameters are defined in the table below, where $l_1$, $l_2$, and $l_3$ are the lengths of the coxa, femur, and tibia links, respectively, and $\theta_1$, $\theta_2$, $\theta_3$ are the joint angles.
| i | $a_i$ | $\alpha_i$ | $d_i$ | $\theta_i$ |
|---|---|---|---|---|
| 1 | 0 | 0 | 0 | $\theta_1$ |
| 2 | $l_1$ | $\pi/2$ | 0 | $\theta_2$ |
| 3 | $l_2$ | 0 | 0 | $\theta_3$ |
| 4 | $l_3$ | 0 | 0 | 0 |
The homogeneous transformation matrix between consecutive frames is given by:
$$ {}^{i-1}_iT = \begin{bmatrix}
\cos\theta_i & -\sin\theta_i\cos\alpha_i & \sin\theta_i\sin\alpha_i & a_i\cos\theta_i\\
\sin\theta_i & \cos\theta_i\cos\alpha_i & -\cos\theta_i\sin\alpha_i & a_i\sin\theta_i\\
0 & \sin\alpha_i & \cos\alpha_i & d_i\\
0 & 0 & 0 & 1
\end{bmatrix} $$
The transformation from the hip frame to the foot-tip frame is the product of these matrices:
$$ {}^0_4T = {}^0_1T \cdot {}^1_2T \cdot {}^2_3T \cdot {}^3_4T $$
The position of the foot tip $ {}^0P_4 = [x, y, z]^T $ in the body frame is extracted from the fourth column of $ {}^0_4T $, yielding the forward kinematics equations:
$$
\begin{aligned}
x &= l_1\cos\theta_1 + l_2\cos\theta_1\cos\theta_2 + l_3\cos\theta_1\cos(\theta_2+\theta_3)\\
y &= l_1\sin\theta_1 + l_2\sin\theta_1\cos\theta_2 + l_3\sin\theta_1\cos(\theta_2+\theta_3)\\
z &= -l_2\sin\theta_2 – l_3\sin(\theta_2+\theta_3)
\end{aligned}
$$
For gait planning, the inverse kinematics problem must be solved: finding joint angles $(\theta_1, \theta_2, \theta_3)$ for a desired foot position $(x, y, z)$. The solution is derived as:
$$
\begin{aligned}
\theta_1 &= \text{atan2}(y, x)\\
D &= \frac{x}{\cos\theta_1} – l_1 \quad \text{(for } \cos\theta_1 \neq 0\text{)}\\
\theta_3 &= \arccos\left( \frac{D^2 + z^2 – l_2^2 – l_3^2}{2 l_2 l_3} \right)\\
\theta_2 &= \text{atan2}(-z, D) – \text{atan2}\left( l_3 \sin\theta_3, l_2 + l_3 \cos\theta_3 \right)
\end{aligned}
$$
These equations form the foundation for calculating all leg joint trajectories during walking and posture adjustment for this sophisticated bionic robot.
Adaptive Terrestrial Gait Planning
The bionic robot employs adaptive gait planning to optimize mobility for different terrains. Gaits are defined by the sequence and timing of leg lift (swing phase) and ground contact (stance phase). The system uses data from an inertial measurement unit (IMU) to estimate terrain roughness and autonomously switch between two primary gait patterns.
On flat or moderately uneven ground, a tripod gait is used for speed. The six legs are divided into two groups of three (L1, L3, R2 and L2, R1, R3, where L/R denote left/right and numbers indicate front-to-back order). These groups alternate swing and stance phases with a duty factor of 0.5. When one group is swinging forward, the other provides a stable triangular support base. To account for the robot’s non-negligible mass, a dynamic phase compensation adjusts overlap in the stance phases to minimize body pitch/roll oscillations.
On steep slopes, rugged terrain, or under high load, the system switches to a more stable wave gait. In this pattern, only one leg is in the swing phase at any time, while the other five remain in stance. This results in a duty factor greater than 0.5, significantly increasing stability at the cost of reduced forward speed. The swing sequence typically follows a wave from the rear to the front legs on one side, then the other. This gait provides maximum stability, ensuring the bionic robot maintains traction and balance in challenging environments.
Takeoff Posture Self-Correction
Transitioning from ground to flight on uneven terrain requires a critical step: ensuring the aerial platform is level at takeoff to generate vertical lift and avoid collision with the ground. The bionic robot achieves this through active posture correction using its legs. The problem is framed as finding the joint angles that position all feet firmly on the sloping ground while orienting the central body (and attached flight platform) horizontally.
Let the vector from the body origin $O$ to a foot point on level ground be $ {}^0P_4 $. When the body is tilted by Euler angles $(\phi, \theta, \psi)$ (roll, pitch, yaw) and the foot must reach a point on the slope defined in a world frame, the required foot point in the body frame $ {}^0P_{4,\text{slope}} $ becomes:
$$ {}^0P_{4,\text{slope}} = {}^0_WR(\phi, \theta, \psi) \cdot ( {}^WP_{\text{ground}} – {}^WP_O ) $$
Where ${}^0_WR$ is the rotation matrix from world to body frame, derived from the Z-Y-X Euler angle sequence:
$$ {}^0_WR = R_z(\psi) R_y(\theta) R_x(\phi) = \begin{bmatrix}
c\psi c\theta & c\psi s\theta s\phi – s\psi c\phi & c\psi s\theta c\phi + s\psi s\phi\\
s\psi c\theta & s\psi s\theta s\phi + c\psi c\phi & s\psi s\theta c\phi – c\psi s\phi\\
-s\theta & c\theta s\phi & c\theta c\phi
\end{bmatrix} $$
(Here, $c\cdot$ and $s\cdot$ denote $\cos(\cdot)$ and $\sin(\cdot)$, respectively). For takeoff leveling, the desired body orientation is horizontal, so $\phi_{\text{desired}} = \theta_{\text{desired}} = 0$. The IMU provides the current orientation $(\phi_{\text{current}}, \theta_{\text{current}})$. The control system calculates the required foot displacement to achieve the level posture. The inverse kinematics solver then computes the necessary joint angles for all six legs to realize this displacement simultaneously. By driving the servos to these calculated angles, the bionic robot actively levels its body against the slope before initiating rotor spin-up, ensuring a safe and stable vertical takeoff.
Aerial Mode and Compact Flight Configuration
Once airborne, the hexapod legs are retracted close to the body. However, the spanning rotor arms may still present a wide profile, hindering navigation through narrow passages like gaps between ruins or dense foliage. To address this, the bionic robot employs a synchronized arm retraction maneuver. All four rotor arms can be symmetrically rotated inward around their mounting points, transforming the robot’s footprint from a large “X” shape to a more compact configuration, often resembling a narrower “H” or folded shape. This is achieved by the same motors used for deployment, now rotating in the opposite direction under coordinated control.
During this maneuver and subsequent flight, the flight controller manages stability. The net thrust $F_{total}$ and body torques $\tau$ are functions of individual motor thrusts $F_i$ and their positions relative to the CoG. For a standard quadrotor in level hover, the following condition holds for forces and moments:
$$
\begin{aligned}
F_{total} &= \sum_{i=1}^{4} F_i = mg \quad \text{(Lift balances weight)}\\
\tau_x &= l( -F_2 + F_4 ) \quad \text{(Roll torque)}\\
\tau_y &= l( F_1 – F_3 ) \quad \text{(Pitch torque)}\\
\tau_z &= \sum_{i=1}^{4} Q_i \quad \text{(Yaw torque, } Q_i \text{ is reaction torque)}
\end{aligned}
$$
Where $l$ is the effective moment arm. The symmetrical and synchronous nature of the arm rotation ensures that the CoG shift and inertia changes are minimal and predictable, allowing the flight controller to compensate and maintain stable flight even in the compact configuration. This capability significantly enhances the operational flexibility of this unique bionic robot in confined spaces.
Software Architecture and Program Flow
The operational intelligence of the dual-mode bionic robot is governed by a centralized software executive running on the main control board. The program follows a structured, cyclic flow to manage sensors, decision-making, and actuator control.
Initialization: Upon startup, the system performs a comprehensive self-check. This includes sensor calibration (IMU zeroing), establishing communication links with the flight controller and the remote receiver, initializing the RTOS, and checking the status of all 18 servo motors on the serial bus. The system enters a standby mode awaiting user command.
Continuous Health Monitoring: In the main control loop, the system continuously polls the servo bus, parsing feedback data (current, temperature, position) from each joint node. Safety protocols are active; for instance, if a current spike above 5A is detected—indicating a potential stall or obstacle during leg movement—a dynamic torque limiting routine is triggered within 10 ms to prevent damage.
Terrain Assessment and Gait Selection: Data from the IMU and other optional terrain sensors (e.g., time-of-flight sensors on the legs) are fused to compute a terrain roughness index. Based on this real-time assessment, the gait planning module autonomously selects the optimal locomotion pattern—tripod for efficiency or wave for stability.
Mode Execution: The system constantly reads the command input from the remote receiver.
- In Terrestrial Mode: The robot ensures the rotor arms are stowed. It then executes the selected gait algorithm. Desired foot trajectories are generated based on the commanded velocity/direction, translated into joint angles via the inverse kinematics solver, and the resulting angle setpoints are broadcast to the respective servo motors.
- In Aerial Mode: The robot first commands the rotor arms to deploy. It then engages the takeoff posture self-correction routine, using leg adjustments to level the body if on a slope. Upon verification of a level state, the flight controller is armed, and takeoff proceeds. After becoming airborne, a command is sent to retract the legs to a streamlined position. The process is reversed for landing: legs are extended before touch-down to absorb impact and provide a stable base.
Loop Continuity: This cycle of sensor reading, data processing, decision-making, and actuator control repeats at high frequency, ensuring responsive and stable operation of the bionic robot in both modalities.
Experimental Validation and Performance Analysis
A functional prototype of the dual-mode bionic robot was constructed and subjected to a series of tests in various environments to validate its design principles and quantify its performance.
Terrestrial Mobility Performance
On flat grassland, the robot deployed its tripod gait. The transition from flight to walking mode, involving arm stowing, took approximately 2 seconds. The robot achieved a maximum straight-line speed of 0.72 m/s and could perform a 360° in-place turn in 4.2 seconds. On rugged, uneven terrain with small obstacles, the system autonomously switched to the wave gait. Here, the straight-line speed decreased to 0.45 m/s and the turning time increased to 6.1 seconds, accompanied by minor body oscillation (approx. 1.5° as measured by the IMU). The enhanced stability of the wave gait was confirmed by laser tracker measurements showing a reduced heading error over a 100m path: 2° for wave gait versus 5° for tripod gait.
Slope climbing tests were conducted on inclines. The robot maintained a wave gait on a 15° slope at 0.43 m/s. The maximum stable incline angle was found to be 32°, limited by the mechanical clearance between the leg’s second joint and the stowed rotor arm (maintaining >8mm gap). The following table compares key terrestrial performance metrics of this bionic robot’s hexapod mode with a conventional, similarly sized hexapod robot.
| Parameter | This Bionic Robot (Hexapod Mode) | Conventional Hexapod |
|---|---|---|
| Dimensions (L×W×H) | 0.668 m × 0.557 m × 0.402 m | 0.8 m × 0.6 m × 0.5 m |
| Max Speed (Tripod, Flat) | 0.72 m/s | ~0.8 m/s |
| Max Speed (Wave, Slope) | 0.43 m/s | ~0.3 m/s |
| Max Slope Angle | 32° | ~30° |
| Servo Response Time | 0.06–0.1 s | 0.1–0.2 s |
| Straight-line Error (per 100m) | 1–7° | 2–10° |
The data indicates that despite a 37.5% reduction in volume, this bionic robot matches or exceeds conventional designs in key mobility metrics, particularly on slopes, thanks to its advanced servo control and adaptive gait algorithms. Furthermore, while a typical hexapod may overcome obstacles up to 15-18 cm, this system’s dual-mode nature allows it to fly over arbitrarily large ground obstacles, offering unprecedented operational flexibility.
Aerial and Transition Performance
The aerial capabilities and mode transition were rigorously tested. The complete transition sequence from grounded hexapod mode to stable hover (including posture leveling, arm deployment, and takeoff) took approximately 7 seconds, both on level and sloped ground. The posture self-correction algorithm successfully maintained a level flight platform during takeoff from slopes up to the tested maximum of 24°. The retractable arm mechanism proved highly effective: the minimum lateral clearance required for flight was reduced from 0.6 m with arms deployed to 0.4 m with arms retracted, a 38% reduction in effective cross-sectional area. The robot achieved a maximum horizontal flight speed of 8 m/s. Its flight endurance in hover was approximately 22 minutes, and the maximum safe operational altitude tested was 500 m. The table below summarizes key aerial performance parameters.
| Parameter | Value (Aerial Mode) |
|---|---|
| Rotor Span (Deployed) | 0.608 m |
| Max Flight Speed | 8 m/s |
| Hover Endurance | 22 min |
| Full Mode Transition Time | 7 s |
| Max Takeoff Slope Angle | 24° |
| Operational Altitude | Up to 500 m |
Conclusion
This paper presented the design, analysis, and experimental validation of a novel dual-mode bionic robot that seamlessly integrates quadrotor aerial mobility with hexapod terrestrial locomotion. The system addresses the inherent limitations of unimodal robots by providing complementary capabilities: stable, adaptive walking for detailed ground-level operation and rapid, obstacle-clearing flight for reconnaissance and transit. Key innovations include the mechanically synchronized retractable rotor arm system that minimizes the robot’s footprint for confined-space navigation, an adaptive gait planning algorithm that switches between tripod and wave gaits based on real-time terrain assessment, and a takeoff posture self-correction routine that enables stable vertical takeoff from uneven or sloped surfaces.
The hierarchical control architecture, built around a high-performance microcontroller and a bus-based high-torque servo network, provides the computational power and precise actuation required for complex bionic motion. Kinematic modeling using the D-H convention forms the analytical backbone for accurate leg control. Physical prototype testing confirmed the system’s practical viability, demonstrating competent performance in both modalities and a smooth, rapid transition between them.
This work validates the concept of a tightly integrated aerial-terrestrial bionic robot as a viable and powerful platform for missions in complex, unstructured environments such as disaster zones, industrial inspection sites, and wilderness exploration. The dual-mode bionic robot represents a significant step towards creating truly versatile robotic systems capable of overcoming a wide spectrum of environmental challenges.