A multi-armed robot for assisting with agricultural tasks

Humans often use one hand to grasp the branch for better accessibility, while the other hand is used to perform primary tasks like branch pruning and hand pollination of the flower. In their paper "Force Aware Branch Manipulation To Assist Agricultural Tasks", presented at IROS 2025, Madhav Rijal, Rashik Shrestha, Trevor Smith, and Yu Gu proposed a methodology to safely manipulate branches to aid various agricultural tasks.

The Challenge of Occlusion

Madhav Rijal (MR) explains that the work is motivated by StickBug, a multi-armed robotic system for precision pollination in greenhouse environments. A main challenge is that many flowers are partially or fully hidden within the plant canopy, making them difficult to detect and reach. This also applies to fruit harvesting, where targets may be occluded by foliage.

To address this, the team studied how one robot arm can safely manipulate branches so that occluded flowers can be brought into the field of view or reachable workspace of another robot arm. This is challenging because branches are deformable, fragile, and vary significantly. Unlike pick-and-place tasks, branches remain attached to the plant, imposing motion constraints. Excessive force can easily damage the plant.

Force-Aware Motion Planning

The proposed approach combines motion planning that accounts for branch constraints with real-time force feedback. First, they generate a feasible manipulation path using an RRT* (rapidly exploring random tree) algorithm-based planner. Branches are modeled as deformable linear objects, and geometric heuristics identify safe configurations.

During execution, the system monitors interaction force via a sensor. If the force exceeds a predefined threshold, the system re-plans the motion online to find an alternative path that reduces stress while still achieving the task. The robot does not plan only for reachability; it adapts based on the physical response of the branch.

Experimental Results

The method was evaluated through 50 trials using five different starting poses. A trial was successful if the robot brought the grasp point within 5 cm of the goal. The system achieved a 78% success rate.

In terms of safety, constraint-aware planning reduced manipulation force from above 100 N to below 60 N. Online force-aware replanning further reduced it below the 40 N threshold. These findings highlight the value of force-aware planning for practical robotic manipulation in agricultural environments.

Future Directions

Future work includes learning safe force thresholds automatically from branch geometry or visual cues, rather than defining them in advance. The team also aims to improve grasp-point selection and design compliant grippers with integrated sensing. Ultimately, this method will be integrated into multi-arm robots where arms collaborate to perform pollination, pruning, or harvesting by actively exposing hidden targets within the canopy.