ALFRED: Requirement-driven development of an open-source mobile manipulator for long-term plant monitoring

Ciarán Miceal Johnson; Christopher Quail; Garry Ellard; Alistair McConnell; Steve Tonneau; Fernando Auat Cheein

cs.ROOct 1, 2026

ALFRED: Requirement-driven development of an open-source mobile manipulator for long-term plant monitoring

Authors: Ciarán Miceal Johnson, Christopher Quail, Garry Ellard, Alistair McConnell, Steve Tonneau, Fernando Auat Cheein

Abstract

Tracking seasonal change in crops and forests requires observing the same plants repeatedly. Ground robots can do this at close range, and a manipulator gives their sensors more viewpoints. Yet the robots behind long-term field datasets are rarely released with their design files, and how a robot's own structure limits arm reach and occludes its sensors is seldom compared between builds. We present ALFRED, an open-source mobile manipulator built from commercially available components. It carries a six-degree-of-freedom arm, LiDAR, RGB-D cameras, RTK GNSS and an IMU on an Ackermann-steered base, all mounted on a reconfigurable aluminium strut frame, and runs containerised ROS software. It was developed through four builds against six requirements for repeated outdoor deployment: durability, modularity, repairability, sensing reach, endurance and reproducibility. Model-based analysis of the last three builds shows the usable share of the arm's reachable poses rising from 34.0% to 60.0% and then 66.1%, and ray casting shows that only the final build keeps the frame-mounted LiDAR's horizontal view clear both forwards and backwards. ALFRED completed a year of monthly forest surveys (528 traversals) without missing a scheduled collection. This was despite battery degradation, reconfiguration for another researcher's study, and the parallel development of ALFRED 2.0 for autonomous crop-row operation, with each switch between builds taking about six hours. The deployment also showed that mechanical modularity is only as dependable as the robot description that tracks it.

Figures & tables

Panel of Figure 1 — Figure 1 : ALFRED as configured for the ForestYear3D campaign, showing the aluminium strut frame, hinged access panels, manipulator and sensor suite.

Panel of Figure 2 — Figure 2 : Point clouds of the same forest corridor from ForestYear3D, showing the contribution of each LiDAR and the change between June and February.

Panel of Figure 3 — Figure 3 : ALFRED 2.0 as configured for the ROWCAST experiments, showing the second frame layer, the second laptop and the removed LiDAR.

Panel of Figure 4 — Figure 4 : RGB-D point clouds of two artificial plants reconstructed by ALFRED 2.0 during the ROWCAST trials.

Joint	Minimum [rad]	Maximum [rad]
1 (waist)	$-2.618$	$2.618$
2 (shoulder)	$0$	$2.967$
3 (elbow)	$-2.880$	$0$
4 (wrist 1)	$-1.518$	$1.518$
5 (wrist 2)	$-1.344$	$1.344$
6 (wrist 3)	$-2.793$	$2.793$

Table 1 : Manipulator joint ranges, as specified in the vendor robot description 2 .

Panel of Figure 5 — Figure 5 : Simplified power distribution within the electronics enclosures. Each enclosure has an independent $48\text{\,}\mathrm{V}$ $30\text{\,}\mathrm{A}\text{\,}\mathrm{h}$ battery and steps down to the voltages its loads require.

Figure 6 : Network topology of ALFRED 2.0. Colour distinguishes computers, network hubs, control hardware, and sensors used in the reported experiments from those carried but not used. Line style distinguishes Ethernet, USB, Wi-Fi and Bluetooth connections. The wireless link carries RTK correction messages from the base station to the rover.

Panel of Figure 7 — Figure 7 : The four builds of the platform. Prototype 0 established component integration, Prototype 1 was the first complete assembly, ALFRED collected the ForestYear3D dataset, and ALFRED 2.0 added the computing, networking and sensing required for autonomous crop-row operation.

Figure 8 : Each of the four builds assessed against the six design requirements as not satisfied, partially satisfied or satisfied.

Figure 9 : ALFRED’s aluminium profile strut frame before panelling and components were mounted.

Panel of Figure 10 — Figure 10 : Interchangeable sensor mounts, each built on the common manipulator interface.

Figure 11 : ALFRED with a hinged door open, giving field access to an electronics enclosure.

Figure 12 : ALFRED reconfigured for the concurrent payload study, with the manipulator replaced by a weight cage and an alternative GNSS receiver fitted.

Panel of Figure 13 — Figure 13 : RGB-D mounts redesigned with ventilation after the cameras overheated during extended operation.

Panel of Figure 14 — Figure 14 : CAD designs for Prototype 1’s 3D-printed cage housing the electronics enclosures.

Panel of Figure 15 — Figure 15 : Robot descriptions of the three complete builds, with the manipulator in the zero posture, used for the visibility and reachability analyses.

Figure 16 : ALFRED in the drive posture, also used as the angle-scan posture in ForestYear3D.

Figure 17 : Ray termination for the $-22.5\text{\,}\mathrm{\SIUnitSymbolDegree}$ , $-0.7258\text{\,}\mathrm{\SIUnitSymbolDegree}$ and $22.5\text{\,}\mathrm{\SIUnitSymbolDegree}$ beams. Green rays reach the $5\text{\,}\mathrm{m}$ cap, ochre rays meet the ground first and red rays meet the platform. White marks the region beyond a first intersection, not absent coverage.

	Beam	Clear aperture [deg]		Scan share [%]		Largest arc [deg]
Configuration	[deg]	Long. F/R	Trans. L/R	Clear	Plat.	Plat.	Non-clear
Prototype 1 † , zero	$-$ 22.50 ‡	0.00/0.00	0.00/0.00	7.32	18.65	54.84	278.79
	$-$ 0.73	0.00/0.00	90.70/90.70	50.39	7.91	28.48	150.12
	+22.50	360.00/360.00	360.00/360.00	100.00	0.00	0.00	0.00
Prototype 1 † , drive	$-$ 22.50 ‡	0.00/0.00	0.00/0.00	14.45	11.43	41.13	266.84
	$-$ 0.73	0.00/0.00	90.70/90.70	50.39	7.91	28.48	150.12

Table 2 : Self-occlusion of the frame-mounted LiDAR for each build and arm posture. Clear aperture is the clear arc about the forward/rear (longitudinal) and left/right (transverse) directions; scan share is the percentage of azimuths reaching the

5\text{\,}\mathrm{m}

cap (clear) or terminating on the platform; largest arc is the widest contiguous platform-occluded and non-clear sector. † Prototype 1’s sensor is pitched. ‡ Beam elevations are as emitted in the sensor frame.

Figure 18 : Reachability lattice, with each translation coloured by the number of orientations with a valid inverse kinematics solution.

Build	Pairs	Reachable	Collision	Clearance	Usable	Usable [%]
Prototype 1	128	25,096	12,751	3,821	8,524	34.0
ALFRED	102	24,950	4,398	5,578	14,974	60.0
ALFRED 2.0	84	24,950	3,310	5,146	16,494	66.1

Table 3 : Usable manipulator workspace by build, from

3\,110\,400

target poses each. Pairs is the number of collision pairs checked; reachable counts poses at which inverse kinematics converged within the inset joint limits, of which collision and clearance rejections are mutually exclusive subsets. Percentages are of reachable poses.

Figure 19 : Usable workspace by end-effector position relative to the manipulator base; the lower row repeats the upper at a rotated azimuth. Each marker is a translation cell retaining at least one of the 1,728 orientations, and empty cells are omitted. Marker area and colour scale with the number of retained orientations on a common 0–290 range, so the darkest green is the observed maximum rather than a full set.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Hardware Type	Company	Device Name	Quantity	Cost (£)
Mobile Base	Agile X	Hunter 2.0	1	9,524.44
Robot Arm	Unitree	Z1	1	11,812.52
Laptop	Dell	Latitude 7330 Rugged Extreme	1	3,992.08
Laptop	MSI	Stealth GS77	1	2,455.00
Laptop	Alienware	m16 r1	1	2,613.44
LiDAR	Ouster	OS1-32	2	10,622.10

Appendix

Table 4 : Hardware components of the ALFRED platform. Costs are estimates for the quantity listed, in GBP as purchased between 2023 and 2026, and will vary with supplier and date. The Alienware laptop is the operator’s field-edge machine and is not carried on the platform.

Hardware Type	Amount	Cost (£)
Waterproof Box	2	54.62
Cable Glands	6	9.99
48V 30mAh Battery	2	158.06
48V to 24V 25A step-down converter	1	41.75
48V to 24V 5A step-down converter	2	84.46
48V to 5V 2A step-down converter	1	9.34

Appendix

Table 5 : Main components and estimated costs for the two electronics enclosures, which power the arm, sensors and network switches. Costs are estimates for the quantity listed, in GBP as purchased between 2023 and 2026, and will vary with supplier and date. Fuses, wiring, DC power connectors, threaded connectors and mounting plates are excluded and should be sourced to suit the build.

Hardware Type	Amount	Cost
40x40mm 8mm Groove	10,000 mm	330.25
RS Pro Profile Strut
40x40x40mm 8mm Groove	72	662.40
RS Pro Angle Brackets
3mm Acrylic	270x996 mm	14.06
6mm Acrylic	195x1836 mm	51.16

Appendix

Table 6 : Main components and estimated costs for the aluminium strut frame and its mounting hardware, covering both the ALFRED frame and the second layer added for ALFRED 2.0. Costs are estimates for the quantity listed, in GBP as purchased between 2023 and 2026, and will vary with supplier and date. Filament (PLA, PETG, TPU), common fasteners and access to a 3D printer and laser cutter are assumed and excluded.

Explore similar work

Aug 11, 2026cs.RO

Deployment Is Not Destiny: Robot Recomposition in the Field with Unseen Software, Hardware, and Compute Payloads

The tight coupling of subsystems in most robots, though a natural consequence of their complexity, leads to monolithic designs that are time-consuming and difficult to adapt after initial deployment. To address this challenge, we present a framework and supporting abstractions for recomposition during runtime that enable robots to quickly integrate previously unseen modular software, hardware, and compute payloads. Our approach allows non-expert users to quickly add new capabilities in the field through a true plug-and-play process. Crucially, new resources are not only immediately available to a host robot but are also shared with distributed peers, enabling compute-constrained systems to access powerful new remote capabilities. Our framework reduces reconfiguration time to a matter of minutes with no developer intervention, in stark contrast to the hours of expert effort often required for traditional manual integration. We demonstrate our method in two disaster response scenarios, including radioactive source localization at an operational nuclear reactor facility and a thermal-guided search for people in dark, difficult-to-reach spaces. These demonstrations show how in-field recomposition provides timely, flexible, and accessible adaptation to dynamic requirements, representing a critical step toward creating robots that can quickly evolve alongside the tasks, technologies, and environments they support.

Steven Swanbeck, Jonathan Salfity, Jeffery Gunawan +3

Texas Robotics and the Walker Department of Mechanical Engineering, The University of Texas at Austin, Austin, TX 78712, USA

Reconfiguration Robot Systems Modular Framework

Ask PDF

Aug 7, 2026cs.RO

Representation Handoffs for OpenArm-Based Laboratory Mobile Manipulation

Open-source robotics and foundation models have lowered the barrier to embodied AI, yet language-guided laboratory automation still requires reliable alignment from instructions and observations to safe actions. This field report presents an OpenArm-based mobile manipulation prototype for laboratory-style tasks, built by integrating dual OpenArm manipulators with a mobile base, vertical slide, RGB-D sensing, lidar-based mapping, ROS2/MoveIt execution, and profile-defined skill interfaces. The system is organized around representation handoffs: natural language requests are constrained into registered skill calls, sensor observations are grounded into maps and object poses, object priors provide role and skill constraints, and runtime bindings compile validated skills into executable motion goals. We use dry-run traces and startup checks to evaluate this integration path, showing how the prototype exposes missing calibration, incomplete object assets, and unfinished real-scene visual grounding as explicit deployment blockers. These intermediate representations serve as practical debugging interfaces for integrating language, perception, planning, and robot safety in embodied systems.

Yang Shen, Chonghao Cheng, Ziyi Zhao +6

University of Technology Sydney · Southern University of Science and Technology

Handoff Skills

Ask PDF

Jul 21, 2026cs.RO

ModPack: An Extensible Teleoperation Interface for Bimanual Mobile Manipulation

Existing teleoperation systems are often tailored to specific robot hardware and task domains, limiting their scalability and adaptability. We present ModPack, a modular and extensible teleoperation system designed to support diverse robot embodiments and task requirements within a unified framework. At the core of ModPack is a self-contained wearable "backpack" that integrates onboard computation, power, communication, and data storage. Built on top of this shared interface, the system supports plug-and-play capability modules including joint-level teleoperation with haptic feedback, mobile manipulation, and active perception. Experiments across two distinct robot platforms and real-world mobile manipulation tasks demonstrate that ModPack provides a flexible and reusable framework for data collection and policy learning. To support future research, we open-source the complete hardware design and software stack. Project website: https://modpack-robotics.github.io/

Joshua Citron, Renee Zbizika, Zeyi Liu +1

Stanford University

Teleoperation Robotic Systems Real-World Manipulation Tasks Bimanual Manipulation Plug-And-Play

Ask PDF