In recent years, the field of robotics has seen significant advancements, particularly in the development of bionic robots that mimic biological organisms. Among these, quadruped bionic robots have garnered attention due to their superior mobility and adaptability in complex environments. This article presents a comprehensive study on the application of a quadruped bionic robot for express delivery tasks, based on our participation in national robotics competitions. The focus is on the hardware design, software implementation, and control strategies that enable the bionic robot to navigate various terrains and perform delivery functions efficiently. Through this research, we aim to demonstrate the practicality of bionic robots in real-world scenarios, such as logistics and disaster response, where traditional wheeled robots may fail.
The inspiration for this work stems from the growing need for automated delivery systems in urban and remote areas. Bionic robots, with their legged locomotion, offer advantages in traversing uneven surfaces like stairs, grass, and slopes. Our project involves designing and programming a quadruped bionic robot to complete an express delivery course, which includes obstacles like speed bumps, narrow bridges, and staircases. The robot must identify delivery locations using visual cues and deposit packages accurately. This study details our approach, from the mechanical structure to the algorithmic solutions, and validates the design through competitive performance. Throughout this article, we emphasize the term “bionic robot” to highlight the biomimetic aspects of our system.
To provide context, we first outline the competition rules that guided our development. The express delivery task requires the bionic robot to start from a dispatch area, pick up packages (represented by colored balls), and deliver them to specific住户 areas while overcoming terrain challenges. The course is 6000 mm × 5000 mm in size, with a 400 mm wide track marked by white lines. Key obstacles include减速带 (speed bumps), steps, narrow bridges, slopes, and grass patches. The bionic robot must complete the deliveries within 20 minutes, adhering to size and weight constraints: it must weigh less than 3.0 kg, with dimensions under 400 mm × 240 mm × 260 mm when standing, and a leg length under 200 mm. These rules ensure that the bionic robot is compact and agile, mimicking mammalian locomotion for stability.
We summarize the obstacle specifications in Table 1 to clarify the challenges faced by the bionic robot. This table helps in understanding the environmental demands and designing appropriate control strategies.
| Obstacle Type | Dimensions (mm) | Material | Purpose |
|---|---|---|---|
| Speed Bump | Height: 10, Width: 400 | White density board | Test balance and gait adaptation |
| Steps | Height: 50, Depth: 200 | White wooden board | Assess climbing ability |
| Narrow Bridge | Width: 200, Length: 1000 | White density board | Evaluate precision and stability |
| Slope | Incline: 15°, Length: 800 | White density board | Test downhill control |
| Grass Patch | Thickness: 5, Area: 500×500 | Artificial turf | Simulate rough terrain navigation |
The hardware platform of our quadruped bionic robot is meticulously designed to meet these challenges. It consists of five main components: servo groups, a Banana Pi development board, a motion control board, a CCD camera, and attitude sensors. Each component plays a crucial role in ensuring the bionic robot’s functionality. The servo groups, for instance, enable articulated leg movements, with three servos per leg—upper hip, lower hip, and knee joints—allowing a wide range of motion for adaptive gaits. The Banana Pi board serves as the primary processor, running an Android operating system for high-level computations, while the STM32-based motion control board handles real-time servo commands. The CCD camera provides visual input for navigation, and the attitude sensors offer inertial data for balance correction. This integration is key to the bionic robot’s performance.
We detail the hardware specifications in Table 2, which encapsulates the technical aspects of our bionic robot. This table underscores the careful selection of components to achieve lightweight yet robust design.
| Component | Specifications | Function |
|---|---|---|
| Servo Groups | 12 servos total; ranges: head (60°), upper hip (100°), lower hip (180°), knee (150°) | Provide leg articulation for walking and obstacle traversal |
| Banana Pi M2 Board | Allwinner A31S quad-core CPU, 1GB RAM, Android OS, GPIO interfaces | Execute software algorithms and process sensor data |
| Motion Control Board | STM32 microcontroller, TTL serial communication | Translate high-level commands into servo actions |
| CCD Camera | 5-megapixel USB camera | Capture environmental images for visual guidance |
| Attitude Sensor | MEMS-based, outputs 3D angular velocity and acceleration | Monitor orientation and balance for stable locomotion |
The software design is equally critical, as it enables the bionic robot to autonomously navigate and perform tasks. Our software pipeline includes image processing, line tracking, color recognition, and command issuance. Image processing involves resizing, median filtering, and binarization to extract useful features from camera input. For example, we reduce image resolution to 80×60 pixels to optimize processing speed, then apply median filtering to remove noise. Binarization converts the image to black-and-white based on color thresholds, facilitating line detection. This process is essential for the bionic robot to follow the track accurately.
We model the image processing steps mathematically. Let \( I(x,y) \) represent the original image captured by the bionic robot’s camera, where \( x \) and \( y \) are pixel coordinates. The resizing operation can be expressed as:
$$ I_{\text{resized}}(u,v) = \text{resize}(I, (80,60)) $$
where \( u \) and \( v \) are new coordinates. Median filtering is applied to reduce noise:
$$ I_{\text{filtered}}(u,v) = \text{median}\{I_{\text{resized}}(u+i, v+j) : i,j \in [-k,k]\} $$
with \( k \) defining the kernel size. Binarization uses a threshold \( T \) for a specific color range:
$$ I_{\text{binary}}(u,v) = \begin{cases} 255 & \text{if } I_{\text{filtered}}(u,v) \in [C_{\text{min}}, C_{\text{max}}] \\ 0 & \text{otherwise} \end{cases} $$
Here, \( C_{\text{min}} \) and \( C_{\text{max}} \) define the color bounds for track detection. This algorithmic approach ensures that the bionic robot can efficiently process visual data in real-time.
Line tracking is implemented using a single-edge strategy, where the bionic robot follows the inner side of the track to avoid external interference. We scan each row of the binarized image to detect the edge position, compute the mean distance from the edge, and use a PID controller to adjust the robot’s movement. The control error \( e(t) \) at time \( t \) is given by:
$$ e(t) = d_{\text{target}} – d_{\text{measured}}(t) $$
where \( d_{\text{target}} \) is the desired distance from the edge, and \( d_{\text{measured}}(t) \) is the actual distance derived from image analysis. The PID output \( u(t) \) controls the steering:
$$ u(t) = K_p e(t) + K_i \int_0^t e(\tau) d\tau + K_d \frac{de(t)}{dt} $$
with \( K_p \), \( K_i \), and \( K_d \) as tuning parameters. This allows the bionic robot to maintain a consistent path, even on curved sections.
Color recognition is vital for identifying delivery locations. The bionic robot scans for specific colored markers in the environment, such as red, green, or blue areas, which correspond to different住户 zones. After binarization for a target color, we calculate the percentage of white pixels \( P \) in the image:
$$ P = \frac{N_{\text{white}}}{N_{\text{total}}} \times 100\% $$
where \( N_{\text{white}} \) is the count of white pixels (value 255) and \( N_{\text{total}} \) is the total pixels. If \( P \geq 30\% \), the bionic robot recognizes the color and executes the associated delivery action. This simple yet effective method enables reliable task completion.
Command issuance involves sending serial packets from the Banana Pi board to the motion control board. The packet format includes headers, servo IDs, position values, and checksums. For instance, to command a walking gait, we send a sequence of angles for each servo. The bionic robot’s gait can be described using kinematic equations. Let \( \theta_1, \theta_2, \theta_3 \) represent the angles of the upper hip, lower hip, and knee servos, respectively. The foot position \( (x_f, y_f, z_f) \) relative to the body can be computed via forward kinematics:
$$ x_f = L_1 \cos(\theta_1) + L_2 \cos(\theta_1 + \theta_2) + L_3 \cos(\theta_1 + \theta_2 + \theta_3) $$
$$ y_f = L_1 \sin(\theta_1) + L_2 \sin(\theta_1 + \theta_2) + L_3 \sin(\theta_1 + \theta_2 + \theta_3) $$
$$ z_f = \text{constant offset} $$
where \( L_1, L_2, L_3 \) are link lengths. By precomputing these positions for various gaits, we optimize the bionic robot’s movement over obstacles.
Our bionic robot’s performance was validated through participation in national competitions from 2019 to 2021. The results, summarized in Table 3, demonstrate the effectiveness of our design. Each year, the bionic robot achieved high rankings, confirming the robustness of the hardware and software integration. Notably, in 2020, it secured a national first prize, highlighting its superiority in express delivery tasks. These outcomes underscore the potential of bionic robots for practical applications beyond competitions.
| Year | Competition | Achievement | Significance |
|---|---|---|---|
| 2019 | China Robot Competition | National Second Prize | Initial validation of bionic robot design |
| 2020 | China Robot Competition | National First Prize | Enhanced control strategies for better navigation |
| 2021 | China Robot and AI Competition (Preliminary) | Provincial Second Prize | Improved color recognition and obstacle traversal |
| 2021 | China Robot and AI Competition (Final) | National Third Prize | Demonstrated reliability in complex courses |
In conclusion, this research showcases the development and application of a quadruped bionic robot for express delivery. The bionic robot successfully navigates diverse terrains, identifies targets through visual processing, and delivers packages with precision. Our hardware platform, featuring servo-driven legs and embedded systems, provides the physical foundation, while our software algorithms enable autonomous operation. The competition results affirm that bionic robots are viable for logistics and other fields requiring adaptable mobility. Future work may focus on enhancing AI capabilities, such as deep learning for object recognition, to further advance bionic robot technology. Ultimately, the bionic robot represents a step toward more intelligent and versatile robotic systems.
To delve deeper into the control aspects, we analyzed the stability of the bionic robot using dynamic models. The equation of motion for a quadruped bionic robot can be approximated as:
$$ M(q)\ddot{q} + C(q, \dot{q})\dot{q} + G(q) = \tau $$
where \( q \) is the vector of joint angles, \( M \) is the inertia matrix, \( C \) accounts for Coriolis forces, \( G \) represents gravitational effects, and \( \tau \) is the torque input from servos. By linearizing around a nominal gait, we design controllers that ensure the bionic robot remains balanced during obstacle traversal. This mathematical framework underscores the complexity involved in bionic robot locomotion.
Moreover, we evaluated energy efficiency, which is crucial for prolonged operation of the bionic robot. The power consumption \( P_{\text{total}} \) can be estimated as:
$$ P_{\text{total}} = \sum_{i=1}^{12} I_i V_i + P_{\text{electronics}} $$
where \( I_i \) and \( V_i \) are the current and voltage for each servo, and \( P_{\text{electronics}} \) covers the development board and sensors. Through optimization, we minimized energy use by tuning gait parameters, allowing the bionic robot to complete tasks within battery constraints. This aspect highlights the practical considerations in deploying bionic robots for real-world delivery services.
In summary, the quadruped bionic robot developed in this study embodies the fusion of mechanical design, electronic integration, and algorithmic intelligence. Its success in competitive environments validates the approach and paves the way for broader adoption of bionic robots in automation. As technology progresses, we anticipate that bionic robots will become increasingly capable, transforming industries and improving human life through innovative applications.