SkyPlanter: Aerial Reforestation With an Ultralight, Seedling–Planting Drone

IEEE/ASME TRANSACTIONS ON MECHATRONICS 1 SkyPlanter: Aerial Reforestation With an Ultralight, Seedling–Planting Drone Steffan Lloyd , Member, IEEE, Thor Kamp Opstrup, Emad Samuel Malki Ebeid , Senior Member, IEEE, and Rasmus Astrup Abstract—This article presents SkyPlanter, the first drone-based system for aerial reforestation with tree seedlings. Traditional tree planting is labor-intensive, physically demanding, and expensive—making it ideal for automation. Current mechanized solutions depend on large, heavy, ground-based excavator-based solutions best suited for extensive clear-cuts, but which struggle on steep or uneven terrain, and carry prohibitive relocation costs for smaller operations. SkyPlanter is a drone-mounted seedling-planting system that enables it to easily traverse rugged or steep terrain while remaining inexpensive, easily transported, and highly scalable. It uses an ultra-lightweight compressed air planting mechanism that inserts seedlings and compacts the surrounding soil. Its innovative double-telescoping design reduces vehicle weight to 15.2 kg (without batteries) or 16.4–20.8 kg (with batteries, depending on flight duration). This article details the system’s novel planting and ground compression mechanisms, its unique high-pressure pneumatic power systems, and its custom quadrotor carrier drone. We demonstrate its feasibility in the first-ever aerial seedling-planting tests in a forest environment. The system is proposed as a costeffective, scalable reforestation solution with high automation potential. Index Terms—Mechanism design, pneumatics, reforestation, authonomous aerial vehicle (AAV). I. INTRODUCTION TREE planting is essential for sustainable forest management, regenerating forests after harvest. Today, about 7% Received 30 June 2025; accepted 20 August 2025. Recommended by Technical Editor W. He and Senior Editor Y.-J. Pan. This work was supported by the European Union’s Horizon Europe Research and Innovation programme through project SPADE under Grant 101060778. An earlier version of this paper was presented in part at the 2025 IEEE International Conference on Robotics and Automation (ICRA) [DOI: 10.1109/ICRA55743.2025.11128303]. (Corresponding author: Steffan Lloyd.) Steffan Lloyd and Rasmus Astrup are with the Department of Forest and Forest Resources, Norwegian Institute for Bioeconomy Research, 1431 Steinkjer, Norway (e-mail: [email protected]; [email protected]). Thor Kamp Opstrup and Emad Samuel Malki Ebeid are with the Institute of Mechanical and Electrical Engineering, University of Southern Denmark, 5230 Odense, Denmark (e-mail: [email protected]; [email protected]). This article has supplementary material provided by the authors and color versions of one or more figures available at https://doi.org/10.1109/TMECH.2025.3606620. Digital Object Identifier 10.1109/TMECH.2025.3606620 of global forests (≈290 Mha) are planted forests [1]. Meanwhile, deforestation, habitat loss, and climate change drive extinctions and ecosystem degradation. Forest loss releases CO2 and reduces sequestration capacity. Land-use changes, primarily driven by deforestation, account for ≈25% of anthropogenic emissions [2], and exacerbate erosion, disrupt water cycles, and harm local livelihoods. Reforestation directly combats this crisis by restoring habitats and sequestering CO2. Governments worldwide are implementing large-scale planting initiatives—e.g., Canada’s two-billiontree pledge by 2030 [3] and similar programs in China and India [4],[5]. Rapid planting is also vital for postwildfire recovery, as fires grow more frequent and intense with warming temperatures [6]. Tree planting also indirectly supports a shift from carbon-intensive materials (steel and concrete) to sustainable wood products [7],[8]. Improving reforestation techniques is therefore critical for planetary health, climate change mitigation, and a sustainable economy. Currently, tree planting is primarily performed manually by workers carrying seedlings and planting them individually with a spade, dibber, or other specialized device [17]. However, this method is labor-intensive and costly. It also faces challenges, such as workforce shortages and high turnover, driven in part by the physically demanding tasks and difficult working conditions [13],[17]. Some automated tree planting solutions exist, but their adoption remains limited [18],[19]. Excavator-based planters, such as those developed by Risutec [9] or Bracke [20] [see Fig. 1(a)], use heavy, high-powered hydraulics mounted on large excavators to plant trees. Although effective, these systems are often not cost-competitive due to the high relocation costs of the large machines, as well as the associated manual operations required to reload the planting device. Due to these costs and their large size, these planters are best suited to operations in large, flat, and open areas (e.g., large clear-cut sites). However, extensive areas of the world’s forests are in challenging terrain that is inaccessible to these machines, and managed and reforested in small-scale operations, where the significant relocation costs cannot be justified. Moreover, their significant weight can cause environmental damage and ground disturbance. Aerial seeding, or “seed-bombing,” involves dropping fertilized seed pellets from an authonomous aerial vehicle (AAV), shown in Fig. 1(b). While this approach is scalable and straightforward with modern drones, it yields poor survival outcomes, with few pellets developing into healthy trees [21]. This can be attributed in part © 2025 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 2IEEE/ASME TRANSACTIONS ON MECHATRONICS Fig. 1. Current reforestation solutions. (a) Excavator-mounted planter by Risutec [9]. (b) “Seed bombing” drone by AirSeed [10].(c)AutoPlant robot [11],[12],[13].(d)RoboFoR planting robot [14],[15],[16]. to the fact that seeds or seedlings should be placed in appropriate substrates and locations. More recently, researchers have explored custom robotic platforms for tree planting. The Swedish AutoPlant system [see Fig. 1(c)] on the wheeled arctic off-road robotics (AORO) platform [11],[12],[13] employs a hydraulic/electric planting head from Bracke Forest mounted on a custom autonomous forest machine, demonstrating fully autonomous site preparation and planting cycles with reduced soil disturbance compared to traditional excavator-mounted planters. Similarly, the Polish RoboFoR platform [see Fig. 1(d)][14],[15],[16] is a custom wheeled robot that integrates a hydraulic scarification/planting device with a seedling-feeding system, and has undergone preliminary testing. Industrial initiatives, such as Södra’s BraSatt robot [22] and the PlantmaX project [23], are also developing similar robots, and other concepts have been proposed in research— albeit without experimental validation [24],[25]. All these systems remain under development and are not commercially available. Furthermore, they still rely on large ground-based tractorstyle machines for mobility, which may be limiting in large parts of the forest areas with challenging terrain and obstacles. A key challenge in forestry robotics and automated tree planting is the paradoxical design goals of a platform that can handle rugged forest terrain, while remaining small enough to navigate narrow spaces between trees and minimize soil disturbance. The forest floor is littered with ditches, boulders, fallen deadwood, and other obstacles, making wheeled locomotion exceedingly difficult [12],[18]. Moreover, navigation can also be challenging, as GPS systems have trouble penetrating the forest canopy, and common SLAM solutions can struggle in the dynamic environment presented by foliage and undergrowth [26],[27]. This article introduces a novel solution to these challenges: SkyPlanter, the first drone-based seedling-planting mechanism mounted on a quadrotor AAV. The pneumatically actuated system is capable of penetrating the soil with a dibber, depositing a seedling, and compacting the soil afterward. The design is enabled by the following novel contributions. 1) A double-telescoping mechanism is developed to align the load paths of the planting and compression actions through a single-central tube, reducing structural requirements and weight significantly compared to existing planting mechanisms. 2) A unique high-pressure pneumatic power system is developed to enable explosive, high-power motions with minimal mass. 3) A lightweight passive pincer mechanism is created to compress the soil inward around the tree, introducing a mechanical advantage and effectively compacting the soil without heavy hydraulics. 4) A custom quadrotor aligns the central tube of the mechanism with the drone’s center of mass, resulting in a compact design that minimizes rotational inertia and prevents tipping during planting. These innovations combine into a fully functional aerial planter weighing just 15.2 kg (16.4–20.8 kg with batteries), far lighter than any existing ground-based system. To the best of the authors’ knowledge, this system also is the first to use a high-pressure compressed air (100–300 bar) power source in mobile robotics generally, and demonstrates its unique ability to generate high-power motions with relatively compact and lightweight actuators. This novel aerial solution avoids the locomotion challenges faced by wheeled robots navigating the obstacle-laden forest floor and steep or uneven terrain while with a wingspan of only 2.1 m, SkyPlanter is small enough to navigate in many open forest areas. It can be used in both large clear-cut reforestation and operations with smaller canopy gaps, and with its small size and light weight, it can be easily transported to the reforestation site or even fly itself to more remote locations. The preliminary design of the lightweight planting mechanism, a key component of the SkyPlanter system, was presented at ICRA 2025 [28]. This conference article detailed the initial design of a lightweight, double-telescoping planting mechanism, prior to its integration with the drone system, and validated its performance in planting while attached to a fixed pedestal in a laboratory setting. However, significant questions remained regarding the mechanism’s real-world performance and its suitability for drone-mounted aerial planting. The current article bridges the gap between the preliminary standalone planting mechanism from [28] and the fully integrated, drone-mounted SkyPlanter aerial planting system. The planter mechanism is refined with an updated compression module that, thanks to a novel passive pincer mechanism, redirects the downward compression forces inward to secure the seedling more firmly and introduces a 57% mechanical force advantage for firmer compaction—all while avoiding unbalanced lateral forces that could tip the drone. We introduce the high-powered carrier drone, custom-made to carry the planting mechanism, reloading system, and power systems while remaining compact and maneuverable. Finally, the system is validated through aerial planting trials with spruce seedlings in a forest environment. We further perform kinematic characterization of the planter motions and productivity assessments of the SkyPlanter system. The rest of the article is organized as follows. Section II discusses current planting methods and associated constraints and challenges. Section III presents our proposed SkyPlanter This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. LLOYD et al.: SKYPLANTER: AERIAL REFORESTATION WITH AN ULTRALIGHT, SEEDLING–PLANTING DRONE 3 Fig. 2. Simplified schematic of the current operating concept for most existing hydraulic, excavator-mounted planting devices (adapted from [29]). system, including the planter design, compression mechanism, and carrier drone design. Section IV validates the design with real-world planting tests in a forest environment. Finally, Section Vconcludes this article. II. BACKGROUND AND PREVIOUS WORK This article examines the feasibility of planting seedlings from an aerial platform, such as a quadrotor. Within the European Union Aviation Safety Agency drone flight regulations, aerial vehicles can be flown with lesser regulatory requirements when they weigh less than 25 kg. Thus, this is the target weight for our system. After accounting for the drone and its batteries, about 10 kg remains for the planting mechanism, seedlings, and power systems. We plan to plant standard M95 Norwegian spruce seedlings, which have a cylindrical plug (root package) approximately 3 cm in diameter and 10 cm in length. The plant itself ranges from 30 to 40 cm in height. The system should be capable of reloading and planting multiple seedlings autonomously and must be robust to rocks or other obstacles in the soil. Planting a seedling involves three key steps, although scarification (exposing loose mineral soil beneath the topsoil to improve survival rates) is performed before planting, in some cases. First, a hole is created for the seedling. Second, the seedling is deposited into the hole. Finally, the soil is compacted around the tree to secure it in the ground and ensure that the roots can expand into the surrounding material. Without proper compaction, freeze–thaw cycles in colder climates can cause soil to expand, effectively pumping the seedling out of its hole and causing it to topple. Current excavator-mounted solutions use a series of hydraulic pistons to automate the planting process, as illustrated in the simplified schematic in Fig. 2. In these systems, a seedling is placed in a central planting tube with a planting dibber and hatch at its lower end. The tube is hydraulically driven into the ground, after which a second hydraulic actuator opens the dibber hatch to create the planting hole. Finally, two opposing angled cylinders, attached to the planter’s steel frame, extend to compress the soil downward and inward around the plant. Although effective, the resulting design relies on a large and powerful hydraulic powerpack, cylinders, and heavy steel framing that far exceeds the weight limits for aerial usage. Designing a device capable of performing these same steps, while adhering to the strict weight Fig. 3. SkyPlanter aerial planting system. limits and centralized load path needed for an aerial solution, presents a challenging design problem with conflicting goals. Our proposed solution is presented in the following section. III. SKYPLANTER PLANTING SYSTEM The proposed aerial planting system, as shown in Figs. 3– 5, consists of three main components: the pneumatic doubletelescoping planting mechanism, the seedling reloading mechanism, and the quadrotor carrier drone. The drone enables the system to reach and land at the desired planting location. The seedling reloading mechanism moves the seedling tray side-toside and drops seedlings down the central tube of the planting mechanism to reload. Once deposited, the seedling rests inside the dibber hatch. When planting, the mechanism explosively drives the dibber into the ground, creating a hole for the seedling plug. The dibber hatch then opens, releasing the seedling into the hole. Finally, the compression mechanism presses the soil downward and inward around the seedling, ensuring good survivability. Compared to existing hydraulic systems, as in Fig. 2, this mechanism achieves large weight reductions through an innovative, double-telescoping design that minimizes structural elements and directs all loads through a single-carbon fiber tube. Simultaneously, the use of high-pressure compressed air as a power source enables high-powered motions with simple and lightweight components, further reducing weight. The following sections detail the key SkyPlanter subsystems. The planter mechanism is described in Section III-A, the seedling reloading system in Section III-B, and the custom quadrotor carrier drone in Section III-C. A. Planting Mechanism Design At the core of SkyPlanter is an ultra-lightweight planting mechanism, as illustrated in Fig. 4. The device uses highpressure air to drive three pneumatic cylinders to perform the four-part planting process, as depicted in Fig. 5: landing, planter insertion, hatch opening, and ground compression. The following sections describe the design in further detail. Section III-A1 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 4IEEE/ASME TRANSACTIONS ON MECHATRONICS Fig. 4. Key parts of the planting mechanism (drone system omitted). Left: Cross-sectional view. Center: Side view. Right: Front view. Fig. 5. Four steps of the planting sequence. (a) Landing. (b) Planter insertion. (c) Hatch opening and seedling release. (d) Ground compression. discusses the pneumatic power system, Section III-A2 explains the planting mechanism and motions, Section III-A3 details the novel pincer-style ground compression mechanism, and Section III-A4 describes the onboard sensors that control the planting process. 1) Power Systems: Electrical power is often used in aerial vehicles due to its cleanliness, convenience, and high energy density. However, generating rapid, explosive motion with electricity typically requires large, heavy actuators, or additional mechanical energy storage elements, such as springs, which require extra reinforcement and structural components. Instead, to achieve high-power, high-force actions while minimizing weight, we chose pneumatics. Pneumatic systems are commonly used in industrial settings, but rarely used in mobile robotics. Nonetheless, they are a viable power source when air is stored at very high pressures. Air tanks used in shooting sports or paintball can store pressures up to 300 bar. At this pressure, the energy density of air can be approximated, assuming isobaric expansion, as Um=W/m ≈PV/m=P/ρ =84.1kJ/kg (1) where Wis total work, mis mass, Vis volume, W=PV, P=300 ·105Pa, and the density of the compressed air is calculated from the ideal gas law as ρ=356 kg/m3at 20 ◦C. Although this energy density is lower than the Um= 396–1080 kJ/kg achievable by modern LiPo batteries [30],[31], This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. LLOYD et al.: SKYPLANTER: AERIAL REFORESTATION WITH AN ULTRALIGHT, SEEDLING–PLANTING DRONE 5 Fig. 6. Pneumatic systems of the planting mechanism. and despite the potentially optimistic assumption of isobaric expansion, the amount of stored energy remains substantial. Given that this energy can be discharged rapidly into lightweight aluminum pneumatic cylinders, producing powerful, explosive motions that are not easily achieved with electrical power, highpressure compressed air is chosen over other power sources. While there are some examples of low-pressure pneumatics (<10 bar) used in literature for hopping robots and similar [32], [33], SkyPlanter is, to the best of the authors’ knowledge, the first mobile robotic platform to employ a high-pressure (>100 bar) compressed air power system. We believe it to be an ideal demonstration of the potential for high-pressure pneumatic power systems to generate explosive motions with minimal mass in mobile applications. SkyPlanter’s pneumatic system is depicted in Fig. 6, and includes the specific models of the components. Air is stored in a small 0.25 L, 300 bar carbon fiber reinforced paintball tank and is regulated to 7 bar in two stages. A high-pressure Ninja Pro V2 paintball tank regulator reduces the air pressure to 38 bar, while serving as a tank valve and integrating burst disks for safety. Then, a custom low-pressure regulator adapts the paintball tank connection to a standard national pipe thread (NPT) thread and further regulates the pressure to 7 bar. A 0.1 L air reservoir maintains consistent system pressure during cylinder movements, compensating for the regulators’ limited response speed. Pilot-actuated electric solenoid control valves direct the air to control the cylinder movements: two 5/3 highflow solenoids control both double-acting cylinders, while a 3/2 low-flow solenoid controls the hatch motion. A complete planting sequence consumes approximately 0.6 L of air at 7 bar, depending on the exact timing and sequence of motions. The expanded air volume from the tank is given by V2=V1P1/P2=0.25 ·300/7=10.7 L. Thus, a single, small 0.25 L tank can plant roughly 17 trees. The system can also accommodate a larger 1.1 L high-pressure storage tank, which can theoretically plant 78 trees, at the cost of 0.70 kg of additional mass. 2) Planting Motion: The proposed planting mechanism, as depicted in Fig. 4, comprises three main subassemblies: an inner telescoping tube assembly rigidly attached to the drone, an outer telescoping tube assembly that slides on a custom plastic sleeve bearing on the outside of the inner telescoping tube, and a compression assembly that slides linearly on the outside of the outer telescoping tube assembly on another plastic sleeve bearing. Three 1.5-mm carbon fiber rods are epoxied longitudinally, at 90◦intervals, on the outer surface of both telescoping tubes. These rods align with slots cut into the inner surface of each of the sleeve bearings, acting as keys to prevent rotational motion between the tubes. The inner telescoping tube assembly is composed of a carbon fiber tube 1epoxied to an aluminum flange 13at the top and a mechanical stop at the bottom 4. The flange 13has two bolt patterns that connect to each of the drone’s carbon fiber body plates. The outer telescoping tube assembly comprises a carbon fiber tube  15 , with a custom aluminum flange 20 epoxied at the top. This flange provides mounting points for the upper brackets for the hatch and compression cylinders  16 and 22 . The planting dibber 9is epoxied to the base of the outer telescoping tube and includes mounting points on both sides. The left mount connects to the lower planting cylinder bracket  29 , where the planting cylinder propels the dibber into the ground. This bracket also integrates the ground detection assembly  30 , discussed further in Section III-A4. The double-acting, ∅20 mm, 250 mm stroke planting cylinder 7, capable of delivering a theoretical force of 220 N at 7 bar, provides the explosive force to push the planting dibber into the ground, sliding the outer telescoping tube  15 along the inner tube 1. This cylinder size was chosen to maximize thrust without exceeding the system’s weight and lifting the drone. However, the actual ground penetration force will nonetheless momentarily exceed the cylinder force on impact due to the added force from the momentum of the lower mechanism components on impact. The right mount on the planting dibber 9bolts to the hatch bracket, which supports the hatch lid  18 and moment arm 17 via a clevis pin joint. The single-acting, ∅20 mm, 100 mm stroke hatch cylinder  16 applies a lateral force of 185 N to the 263 mm moment arm  17 , opening the hatch with approximately 49 Nm of torque to create the planting hole and release the seedling. When pressure is released, a light spring integrated into the cylinder closes the hatch. Most components are made of aluminum sheet and woven carbon fiber tubes, with some CNC-machined parts for flanges. The hatch lid  18 is made of stainless steel for superior toughness during planting—particularly if the planter hits a rock during planting instead of soft ground. Aluminum parts that are epoxied to the carbon fiber are anodized first to provide an insulating barrier between the two materials, preventing galvanic corrosion. These anodized parts are depicted in red in Fig. 4. Nonstructural components are 3D-printed from ASA and are shown in yellow. 3) Ground Compression: The final ground compression step is performed by forcefully driving the compression assembly into the ground, compacting the soil around the tree. The compression assembly consists of a welded aluminum flange  24 with a central aluminum tube 23 . This assembly slides along the outer surface of the outer telescoping tube  15 on a custom This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 6IEEE/ASME TRANSACTIONS ON MECHATRONICS Fig. 7. Illustration of the passive pincer motion for soil compression. plastic bearing 3. The compression flange 24 is also welded to the lower compression bracket  28 that connects to the ∅20 mm compression cylinder  22 , capable of delivering 220 N of downward force. The lower compression bracket  28 extends downward beneath the flange  24 , accommodating the full length of the compression cylinder  22 without the need to extend the planting mechanism. In the previous presentation of the planting mechanism in [28], the compression assembly used static compression arms with shaped pads designed to direct the soil downward and inward upon impact. However, this method often failed to adequately compact the soil around the tree, as it lacked sufficient inward force and was highly sensitive to soil hardness. In the large, hydraulic excavator-mounted designs, two opposing cylinders press the soil inward; however, this solution is not feasible in our design due to its heavy structural requirements. In this article, we replace our previous static compression module with a novel passive pincer design  25 and 27 ,asshown in Fig. 7. Opposing pincer arms  25 , fabricated from 3 mm water-jet cut and bent aluminum sheet, are rigidly bolted to the compression flange  24 and extended outward on opposite sides. At the end of each arm is a revolute pin joint, onto which a compression pincer shovel  27 is mounted. These shovels are also made of 3 mm water-jet cut and bent aluminum sheet, and are angled inward toward the tree. Both the arms and shovels feature cutouts in low-stress areas to reduce weight. As the compression cylinder is extended, the planting cylinder is simultaneously retracted. The pincer arms  25 apply a vertical downward force on the planting shovels  27 at their joints. This force is counteracted by a vertical ground reaction force at the shovel base, and because the reaction force is offset inward relative to the applied force at the top of the shovel, a moment is created. This moment simultaneously forces the shovels inward and downward, creating the precise motion required to optimally pack the soil around the tree. A break-out on the right pincer arm in Fig. 7shows the mechanical stops cut into the design of the compression shovel, restricting its motion between 58◦ (open) and 40◦(compressed) relative to horizontal. These angles correspond to a horizontal distance between the shovels of 143 (open) and 66 mm (compressed). Dual-torsion springs in the revolute joint ensure the shovels  27 return and remain in their outward, open position after the compression motion. Fig. 8. Force and mechanical advantage analysis of the compression shovels. The pincer mechanism also provides a mechanical advantage, multiplying the cylinder force to achieve a stronger compression using the same cylinder size. This mechanical advantage is visualized in Fig. 8. During compression, the ground force Fgis split into a horizontal component Fg,x and a vertical component Fg,y. The reaction force on the pincer arm Fais similarly divided into Fa,x and Fa,y, which are respectively equal and opposite to the ground forces. During the apex of the compression, the shovel angle is roughly 40◦, and the vertical forces can be assumed to be approximately half the cylinder force Fc,as Fa,y =Fg,y =Fc/2=110 N. The rotational moment from the coupling of Fa,y and Fg,y is counteracted by the equal and opposite moment from the horizontal forces Fa,x and Fg,x, such that the resultant forces Faand Fgare exactly aligned with the line of action of the shovel, and can be calculated as Fg=Fa=Fg,y/sin θ=172.2N(2) yielding an effective mechanical advantage MA of MA = 1/sin θ=1.57. This compression mechanism provides several advantages over conventional designs. By using the outer surface of the outer telescoping tube as the bearing surface for the compression motion in a double-telescoping construction, we obviate the need for external framing extending downward to the device’s base, as in existing excavator-mounted designs [29]. The passive pincer mechanism ensures the correct downward and inward motion to compress the soil, while relying on a single pneumatic cylinder and adding minimal weight. Its inherent mechanical advantage increases the applied force by 57% relative to the cylinder force. Finally, the design ensures the resultant compression force is aligned through the central planting tube, avoiding lateral forces that could tip the drone. During planting, the ground compression assembly can be actuated multiple times to improve soil compaction. For our tests, the compression mechanism was triggered three times in succession. However, this may be excessive; further study is needed to determine the optimal balance between planting time, air usage, soil compression, and seedling survival rates. 4) Sensory Feedback: The SkyPlanter system includes sensors to inform the planting process, as given in the following. 1) Ground detection assembly: The ground detection assembly  30 , bolted to the side of the planting dibber, consists This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. LLOYD et al.: SKYPLANTER: AERIAL REFORESTATION WITH AN ULTRALIGHT, SEEDLING–PLANTING DRONE 7 Fig. 9. Key components of the tree reloading system. of an aluminum block with an adjustable plate that sets the planting depth via two bolts. The entire block slides a short distance on two shoulder bolts. When pressed upward, it activates a miniature limit switch embedded in the lower planting bracket  29 , indicating that the dibber has successfully embedded in the ground and signaling the planter to continue or abort the planting sequence. 2) Ground distance sensor: ASICK UM18 ultrasonic distance sensor  26 is mounted to the planting cylinder 7, facing downward to accurately measure the ground distance—a crucial measurement for successful planting. Future enhancements for the device include the addition of sensors, such as cameras or light detection and ranging (LiDAR), enabling the planter to intelligently identify suitable planting sites and simultaneously localize itself while mapping the environment. B. Seedling Reloading System The seedling reloading system, as shown in Fig. 9, feeds new seedlings into the central planting tube, allowing the system to plant multiple times. It is designed to be lightweight and allows for simple replacement of the seedling tray between runs. The system consists of three main components: a base plate  37 , a carriage 34 , and a seedling tray 33 . The base plate  37 is made from light foam-core carbon fiber sandwich panel, and has a central opening beneath which the main planting tube is mounted. When a seedling is positioned over the hole, it drops into the planting tube. Attached to the base plate are aluminum guide rails  35 fitted with plastic bearings, and an aluminum flange that secures a NEMA11 stepper motor-lead screw assembly  38 and 40 with a limit switch for homing. The seedling carriage, made from 2 mm cut and bent aluminum sheet, is mounted on the guide bearings and incorporates a matching plastic lead nut  39 onto the lead screw. This design allows the stepper motor to move the carriage linearly along the slide bearings. The carriage has two 3D-printed ASA wedges that form a dovetail slot along its length. The seedling tray has a matching design, allowing it to be slid into the slot and locked in place with a spring-loaded release pin  32 . The tray has a 3D-printed base with seven holes along its length, each fitted with a thin-walled plastic tube to secure the seedlings. A plastic handling plate  41 keeps the seedlings in place during tray handling but is removed once the tray is loaded onto the carriage, Fig. 10. Key components of the SkyPlanter aerial quadrotor vehicle. causing the seedlings to rest on the base plate 37 , ready to be dropped into the planting tube. In normal operations, the carriage is moved to the center position before reloading the seedling tray. The operator swaps the seedling tray and removes the handling plate, causing the center tree (position 4) to drop into the planting tube. The seedlings are planted in the following order: 5, 3, 6, 2, 7, and 1, alternating sides outward to minimize unbalanced drone loads. C. Aerial Vehicle Design The carrier drone is custom-built to fulfill the unique requirements of the SkyPlanter system. It is designed for high payload capacity, with light, but strong body components that can withstand the planting forces. The design aligns the planting forces with the drone’s center of mass to prevent tipping, and its flight time supports multiple planting cycles without a battery change. The proposed vehicle is shown in Fig. 10. The drone uses an X-configuration with a diagonal motor distance of 1.32 m. Its central body comprises two custom water-jet-cut 4 mm carbon fiber sandwich panels  19 , spaced 37 mm apart with lightweight aluminum spacers. Each panel features a central hole and bolt pattern that interfaces with the drone attachment flange 13on the planting mechanism. This configuration aligns the resultant planting force with the drone’s center of mass, maximizing the reaction force that the drone provides to the planting mechanism and preventing tipping. It also allows the seedling reloading mechanism  33 to be mounted on the top plate, keeping the mass of the planter and reloader close to the drone’s center of mass, minimizing rotational inertia and maximizing maneuverability. Two adjustable telescoping carbon fiber tube legs  48 attached to the base plate allow easy adjustment of the drone height, as the optimal planting height was not known beforehand. The planting control computer  42 is mounted to the top plate, while the pneumatic power system components, such as air tank/regulators  47 and solenoids 46 , are affixed to the bottom sandwich plate. With both external plate surfaces occupied by the planting mechanism components, the drone electronics are housed between the sandwich plates  19 . The RX module, GPS This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 8IEEE/ASME TRANSACTIONS ON MECHATRONICS TABLE I WEIGHT BREAKDOWN OF THE SKYPLANTER SYSTEM antenna, telemetry module, and flight controller are positioned on top to maintain reliable connections. The drone arms are hinged to the body for easier transportation and deployment. The vehicle uses four T-Motor P80-X 100KV Alpha 80 A motor and electronic speed controller (ESC) propulsion systems  44 , paired with ∅775 mm T-Motor MF3016 propellers  43 , each providing a maximum theoretical thrust of 14.3 kg at 50.4 A, 47.1 V. Maximum takeoff mass is 40–50 kg (depending on needed drone agility), well above the current drone weight. The propulsion systems are powered through a MAUCH 2x 200 A power distribution board, allowing them to be safely and remotely disconnected from power during handling, without shutting down the flight controller and other electronics. Power is supplied by four 6 s, 10 000 mAh, 30 C LiPo batteries  45 , connected pairwise in series, resulting in a two-battery 12 s equivalent system. The batteries are housed in custom, 3D-printed polylactic acid (PLA) mounts attached to the side of the drone body between each of the arms. During hover flight with the planter, the system consumes an estimated hover power of 2.2 kW. The batteries provide a theoretical energy of 888 Wh, yielding an estimated flight time of roughly 23.8 min. However, battery capacity and weight can be adjusted based on desired flight time, down to 1.25 kg for 4×2 500 mAh batteries giving approximately 6 min of flight. The flight controller is a Nora+ running the latest firmware and updates. Its parameters have been adjusted to accommodate the slower response time of the larger SkyPlanter drone during takeoff and stable flight, while still preserving responsive flight behavior. Table Idetails the system’s weight breakdown. The planting mechanism only contributes 4 kg to the system weight. The compressed air power system adds 2.4 kg and the seedling reloading system (and seedlings) 1.3 kg. On the carrier drone, the batteries represent the most significant weight; the current setup uses four 1.39 kg batteries, totaling 5.57 kg. As mentioned, Fig. 11. Image of the SkyPlanter system hovering above the sample planting plot, after planting four trees in succession. Full video is available in the Supplementary Material. lighter batteries could be used at the expense of flight time, although reduced drone weight may affect planting force. As the drone’s weight remained below 25 kg, two 1 kg steel weights were attached to the drone body to increase the weight closer to the 25 kg limit, giving the planter more weight and improving the feasibility of the planting process. D. Electronics and Control The SkyPlanter and carrier drone electronics are fully independent and modular. The planting mechanism is controlled via a Raspberry Pi 5 running ROS2 Humble  42 . A custom electronics board, mounted as a shield on the Raspberry Pi, handles sensor signal acquisition and conditioning, triggers solenoids, drives the stepper motor, and conditions power from the drone batteries. Motions and sensor information are exposed via a custom ROS2 C++ node. A base station computer running Ubuntu 22.04 uses a high-power Wi-Fi access point to maintain a stable wireless link with the drone. Drone status and parameters are monitored and updated remotely via QGroundControl over the telemetry link. IV. EXPERIMENTAL VALIDATION The SkyPlanter system was validated in a privately owned forest in ˚ As, Norway. For the initial tests, the drone was manually piloted, and the planting mechanism was triggered manually via the Wi-Fi connection. The seedlings used were 2-year-old M95 Norwegian spruce seedlings. Preliminary experiments were conducted in a standard garden planter for repeatability, while later tests were carried out in a forest environment. In the garden planter, lightly packed gardening soil was used, while in the forest test, the soil was loose clay. Overall, the tests were very successful. Fig. 11 shows an image from the garden planter test, where the SkyPlanter system effectively planted four trees in succession, with the drone taking off and landing between each planting until the available space was exhausted. The forest test was similarly successful. A series of frames in Fig. 12 illustrates the full planting sequence and timing. In this test, the drone lands, extends its dibber explosively This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. LLOYD et al.: SKYPLANTER: AERIAL REFORESTATION WITH AN ULTRALIGHT, SEEDLING–PLANTING DRONE 9 Fig. 12. Frames of the SkyPlanter aerial planting process. Full video is available in the Supplementary Material. Fig. 13. Planted spruce seedling from the forest experiment. into the ground, and opens the hatch to release the seedling. The compression pincers are then activated three times to ensure that the seedling is firmly planted before takeoff. The entire planting process—from landing to takeoff—took about 9 s. Fig. 13 shows a close-up of the planted spruce seedling, with two symmetric compression marks on either side holding it securely. A full video of both tests is available in the Supplementary Material. The current tests used spruce seedlings, but the system can accommodate other species as long as the plants fit within the ∅37 mm planting tube, and the plugs are 10–15 cm in size. This size is appropriate for a variety of species, including spruce, pine, birch, maple, and more. Planting depth is adjustable via the ground detection assembly (see Section III-A4) to suit different plug sizes. For significantly larger or smaller seedlings, the planting tube and shovel could be modified accordingly. 1) Kinematic Analysis: The planting experiment shown in Fig. 12 was recorded from separate angles at 30 fps. This video was used to track the mechanism components during planting, resulting in the data in Fig. 14 and enabling kinematic analysis of the mechanism. Note that the compression shovel angle was less easily measured due to a wider field of view in that camera angle, resulting in noisier data. The data are roughly as expected: the planting dibber embeds itself 130 mm into the ground. The hatch mechanism is opened, rotating from its nominal position approximately 17.5◦inward. The compression mechanism is activated three times, each time embedding itself into the ground and compressing the shovels approximately 10◦inward. During planting, the planter reaches a maximum downward velocity of 1.65 m/s upon soil contact, with a maximum acceleration of 18.0 m/s2(above ground) and a maximum deceleration of 19.2 m/s2(into the ground). During compression, the maximum acceleration is 25.7 m/s2(above ground) and the maximum deceleration is 25.8 m/s2(into the ground). These values are approximate, given the limited framerate and resolution: in particular, the accelerations lack a sufficiently high sampling rate, and the true accelerations are likely higher. 2) Seedling Capacity: This prototype has a relatively small capacity of seven seedlings; however, ideally, the number of seedlings would be maximized to match the drone’s flight time and air tank capacity—reducing the reloading frequency and This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.