Autonomous weeding robots offer significant potential for reducing labor and environmental costs in precision agriculture. However, most systems decouple chassis locomotion from manipulator control, limiting efficiency in unstructured environments. We present LSTM-PPO-Weeding, an end-to-end framework unifying mobility and manipulation via memory-augmented reinforcement learning (RL). A high-speed region of interest (ROI) filter abstracts camera input into three-dimensional weed/crop poses, enabling the system to focus solely on task-relevant features. To mitigate challenges arising from occlusion and partial observability, we integrate long short-term memory (LSTM) into the proximal policy optimization (PPO) algorithm, establishing the proposed LSTM-PPO framework. The agent learns coordinated actions through a custom reward function balancing speed and precision. Experiments in simulation and real-world greenhouse settings show that LSTM-PPO-Weeding improves throughput while maintaining high accuracy. Compared to baseline methods, our approach reduces average weeding time by up to 74% with minimal loss in success rate, demonstrating its robustness for agricultural mobile manipulation.

This paper presents the design and analysis of Pegasus, a quadrupedal wheeled robot grounded in biomimicry principles. Pegasus offers two distinct motion modes, including a wheeled motion and a hybrid wheeled-legged motion, enabling adaptability across various tasks and environmental conditions. The robot draws inspiration from the joint structures of quadruped animals and incorporates biomimetic features. At the robot’s ankle joint, we imitate the articulation of a radiusulna joint to enhance the wheeled motion’s agility. Additionally, we establish comprehensive mathematical models for adaptive dynamics model, providing a robust theoretical foundation for subsequent motion planning and high-precision control. A novel telescopic vehicle mode is also proposed for complex wheel-leg hybrid motion, offering optimized solutions for intricate robot locomotion. Furthermore, we employ parallel underactuated MPC controllers for each leg at the control level, contributing to heightened motion precision and stability. Extensive validation through physical platform experiments highlights the effectiveness and feasibility of the proposed controllers, offering substantial support for real-world applications in robotics.