A multi-armed robotic for helping with agricultural duties

People usually use one hand to know the department for higher accessibility, whereas the opposite hand is used to carry out main duties like (a) department pruning and (b) hand pollination of the flower. (c) An summary of the method utilized by Madhav and colleagues, the place one robotic manipulates the department to maneuver the flower to the sphere of view of one other robotic by planning a force-aware path. Determine from Power Conscious Department Manipulation To Help Agricultural Duties.

Of their paper Power Conscious Department Manipulation To Help Agricultural Duties, which was offered at IROS 2025, Madhav Rijal, Rashik Shrestha, Trevor Smith, and Yu Gu proposed a strategy to soundly manipulate branches to help numerous agricultural duties. We interviewed Madhav to seek out out extra.

Might you give us an summary of the issue you have been addressing within the paper?

Madhav Rijal (MR): Our work is motivated by StickBug [1], a multi-armed robotic system for precision pollination in greenhouse environments. One of many essential challenges StickBug faces is that many flowers are partially or totally hidden inside the plant cover, making them tough to detect and attain straight for pollination. This problem additionally arises in different agricultural duties, reminiscent of fruit harvesting, the place goal fruits could also be occluded by surrounding branches and foliage.

To handle this, we examine how one robotic arm can safely manipulate branches in order that these occluded flowers will be introduced into the sphere of view or reachable workspace of one other robotic arm. This can be a difficult manipulation downside as a result of plant branches are deformable, fragile, and range considerably from one department to a different. As well as, in contrast to pick-and-place duties, the place objects transfer freely in house, branches stay hooked up to the plant, which imposes extra movement constraints throughout manipulation. If the robotic strikes a department with out accounting for these constraints and security limits, it could possibly apply extreme drive and injury the department.

So, the core downside we addressed on this paper is: how can a robotic safely manipulate branches to disclose hidden flowers whereas remaining conscious of interplay forces and minimizing injury?

How did your method go about tackling the issue?

MR: Our method [2] combines movement planning that accounts for department constraints with real-time drive suggestions.

First, we generate a possible manipulation path utilizing an RRT* (quickly exploring random tree) algorithm-based planner within the workspace. The planner respects the geometric constraints of the department and the duty necessities. We mannequin branches as deformable linear objects and use a geometrical heuristic to establish configurations which are safer to control.

Then, throughout execution, we monitor the interplay drive utilizing a drive sensor mounted on the manipulator. If the measured drive exceeds a predefined protected threshold, the system doesn’t proceed alongside the identical path. As an alternative, it re-plans the movement on-line and searches for another path or purpose configuration that may cut back department stress whereas nonetheless attaining the duty.

So, the important thing concept is that the robotic doesn’t plan just for reachability. It additionally adapts its movement based mostly on the bodily response of the department throughout manipulation.

Madhav with the multi-armed pollination robotic, StickBug.

What are the primary contributions of your work?

MR: The primary contributions of our work are:

1. A geometrical heuristic mannequin for department manipulation that doesn’t require branch-specific parameter tuning or bodily probing.
2. A movement planning technique for department manipulation that respects each workspace and department constraints, utilizing the geometric heuristic to information RRT* and incorporating on-line replanning based mostly on drive suggestions.
3. An experimental demonstration exhibiting that drive feedback-based movement planning can shield branches from extreme drive throughout manipulation.
4. Generalization throughout totally different department sorts, because the methodology depends totally on department geometry and may adapt on-line to compensate for mannequin inaccuracies.

Might you speak concerning the experiments that you simply carried out to check the method?

MR: We evaluated the proposed methodology by a set of department manipulation experiments utilizing 5 totally different beginning poses, all focusing on a typical purpose area. Every configuration was examined 10 instances, leading to a complete of fifty trials. A trial was thought of profitable if the robotic introduced the grasp level to inside 5 cm of the purpose level. For all trials, the planning time restrict was set to 400 seconds, and the allowable interplay drive vary was −40 N to 40 N. Throughout the 50 trials, 39 have been profitable and 11 failed, comparable to a hit charge of about 78%. The typical variety of replanning makes an attempt throughout all eventualities was 20.

When it comes to drive discount, the outcomes present a transparent development in security. Constraint-aware planning lowered the manipulation drive from above 100 N to under 60 N. Constructing on this, on-line force-aware replanning additional lowered the drive from about 60 N to under the specified 40 N threshold. This means that security consciousness by geometric heuristics, which mannequin branches as deformable linear objects, along with force-aware on-line replanning, can successfully decrease interplay forces throughout manipulation.

General, the experiments display that the proposed framework permits safer department manipulation whereas sustaining process feasibility. By combining branch-constraint-aware planning with real-time drive suggestions, the robotic can adapt its movement to cut back extreme drive and reduce the danger of department injury. These findings spotlight the worth of force-aware planning for sensible robotic manipulation in agricultural environments.

Do you could have plans to additional lengthen this work?

MR: Sure, there are a number of instructions for extending this work.

One present limitation is the necessity to outline a protected drive threshold prematurely. In observe, various kinds of branches require totally different drive limits for protected manipulation. A key path for future work is to be taught or estimate protected drive thresholds routinely from department geometry or visible cues.

One other extension is to enhance grasp-point choice. As an alternative of solely replanning after greedy, the system might additionally motive about probably the most appropriate grasp level beforehand in order that the required manipulation drive is lowered from the beginning.

We’re additionally fascinated with designing a compliant gripper with built-in drive sensing that’s higher suited to manipulating delicate branches. In the long run, we plan to combine this methodology right into a multi-arm agricultural robotic, the place one arm manipulates the department and one other performs pollination, pruning, or harvesting.

General, this work advances the event of agricultural robots that may actively manipulate branches to assist duties reminiscent of harvesting, pruning, and pollination. By exposing fruits, minimize factors, and hidden flowers inside the cover, this functionality may help overcome key boundaries to the broader adoption of robot-assisted agricultural applied sciences.

References

[1] Smith, Trevor, Madhav Rijal, Christopher Tatsch, R. Michael Butts, Jared Beard, R. Tyler Prepare dinner, Andy Chu, Jason Gross, and Yu Gu. Design of Stickbug: a six-armed precision pollination robotic. In 2024 IEEE/RSJ Worldwide Convention on Clever Robots and Methods (IROS), pp. 69-75. IEEE, 2024.
[2] Rijal, Madhav, Rashik Shrestha, Trevor Smith, and Yu Gu, Power Conscious Department Manipulation To Help Agricultural Duties. In 2025 IEEE/RSJ Worldwide Convention on Clever Robots and Methods (IROS), pp. 1217-1222. IEEE, 2025.

About Madhav

Madhav Rijal is a Ph.D. candidate in Mechanical Engineering at West Virginia College working in agricultural robotics. His analysis combines movement planning, optimization, multi-agent collaboration and distributed choice making to develop robotic methods for precision pollination and different plant-interaction duties. His present work focuses on department manipulation and protected robotic operation in agricultural environments.

tags: IROS