CAU Robot Team
Team members
Xia Hong, Yu Nuo, Zheng Lei, Zhang Qi, Chu Yuxing, Yang Wenchao, Xia Junlin, Xu Guangkuo, Liu Yihao
Team captain’s name
Xia Hong
Instructor(s)
Associate Professor Xin Wang, College of Engineering, China Agricultural University, Beijing, China
Associate Professor Tian Wenbin, College of Engineering, China Agricultural University, Beijing, China
Participating in the FRE since
First time
Description of the team and robot
We developed a compact, high‒precision, fully autonomous field robot designed to complete official FRE competition tasks in complex, unstructured outdoor agricultural environments. The robot integrates mechatronic control, multi‒sensor fusion, and artificial intelligence algorithms, with all design, assembly, and debugging independently completed by our student team.
Robot specifications
W x L x H (cm):
30 x 50 x 60
Weight (kg):
20
Commercial or prototype:
SCOUT MINI
Total no. of wheels / no. driven wheels:
4/4
Drivetrain concept/ max. speed (m/s):
3
Turning radius (cm):
0
Battery type / capacity (Ah):
15
Total motor power (W):
600
No. of sensors internal/
external:
Sensor type:
Internal： 10 External： 4
Sensor Type：
IMU, Encoders, BMS, Motor Current and Temperature Sensors, RGB-D camera
Controller system software description:
Perception & Data Analysis: For environmental sensing, the system relies exclusively on a forward-facing RGB camera. The perception pipeline extracts the navigation path through a real-time adaptive image binarization algorithm (e.g., ExG and Otsu’s thresholding) to segment crop rows from the soil background. Shape and centroid analysis are applied to the binarized image to dynamically calculate the robot's lateral offset and yaw error relative to the target pathway centre.
Controller system hardware description:
The computed trajectory errors are processed via an optimized PID controller to determine velocity commands (Twist), directly commanding the AgileX SCOUT mini base over an industrial CAN bus at 50Hz. The controller incorporates velocity smoothing to mitigate wheel slip on loose field soil and stabilize the camera's field of view. For fault-tolerant safety, the software continuously telemeters the SCOUT mini's built-in motor current and temperature sensors, triggering an immediate safety stop sequence if a vision blackout or motor overload (e.g., weed entanglement) is detected.
Short strategy description for navigation and applications:
This robot features a highly robust, purely vision-based farmland navigation system built upon ROS 2 and the Scout Mini skid-steer chassis.For Task 1 (Straight Crop Rows), the system utilizes Inverse Perspective Mapping (IPM) perspective stretching to compensate for missing-crop gaps. This is combined with HSV segmentation and quadratic polynomial fitting to extract the baseline, achieving centered cruising via a PD controller.For Task 2 (Complex Curved Rows), a cubic polynomial is employed to reconstruct the path curvature, incorporating a tangent pose feedforward control mechanism.At the low-level control layer, the system integrates a depth-vision-based near-field Artificial Potential Field (APF) with an Instantaneous Centers of Rotation (ICRs) model, enabling collision avoidance deceleration and high-precision slip compensation to ensure safe and smooth operations across all working conditions.
These are the commercial team sponsors & partners:
CLAAS