---
title: 'Aerial+: Integrated Robotics for Maintenance'
url: https://www.emergentmind.com/topics/aerial
type: topic
---

# Aerial+: Integrated Robotics for Maintenance

In [2401.02343], AERIAL-CORE is presented as the first integrated “Aerial+” system for inspection and maintenance of electrical power infrastructures: not just a drone platform for imaging, but a heterogeneous, autonomous aerial robotics ecosystem that combines long-range BVLOS inspection, aerial manipulation, energy harvesting/charging, and direct support for human workers at height. In this usage, Aerial+ denotes an aerial robotics paradigm that spans the full inspection-and-maintenance workflow rather than a single aircraft, a single sensing pipeline, or a single manipulation capability. The concept emerges from a broader shift in aerial robotics from passive visual tasks toward physical interaction, reconfiguration, grasping, tactile sensing, and infrastructure-aware autonomy [2605.19431] [2409.14115] [2310.00142] [2304.03026].

## 1. Industrial motivation and problem setting

Aerial+ is motivated by the mismatch between conventional drone capabilities and the operational requirements of large infrastructure. Electrical power networks span tens of millions of kilometers worldwide, and inspection and maintenance are expensive, risky, and often urgent after storms or in fire-prone conditions. Conventional methods such as manned helicopters are costly—quoted at about 150 €/km for inspection—and total inspection and maintenance cost in Europe is estimated at more than 2.2 billion €/year [2401.02343].

The central problem is not inspection alone. The workflow requires long endurance and range, precise sensing and tracking of power lines, vegetation-risk mapping, safe autonomous landing and charging, manipulation of devices on live or de-energized lines, and human–robot collaboration around elevated structures where safety is critical. The paper emphasizes that work at elevated heights is a leading cause of fatal accidents, and that power-line failures can cause outages, bird electrocution, and forest fires when vegetation contacts conductors. Because climate change is increasing damage from storms and adverse weather, the need for autonomous, scalable, and safer solutions is growing [2401.02343].

This framing distinguishes Aerial+ from standard aerial robotics. Standard aerial drone capabilities are described as insufficient for the full inspection-and-maintenance workflow because those workflows require much more than flying a camera around. Aerial+ therefore refers to a mission-complete operational stack in which sensing, mobility, manipulation, charging, communications, teleoperation, and worker support are treated as coupled subsystems rather than isolated demonstrations [2401.02343].

## 2. System architecture and mission lifecycle

AERIAL-CORE is organized as a mission-driven integrated system managed by a “Chief Inspector” who receives a utility company request, initiates inspection, receives the inspection report, and then activates maintenance and co-working subsystems based on the results. The workflow proceeds in stages: first long-range inspection, then maintenance/manipulation actions such as installing bird diverters or charging stations, and finally co-working actions where aerial robots support a human worker performing manual tasks. The paper explicitly states that the system was demonstrated successfully in a real-time integrated mission on October 27, 2023 [2401.02343].

The architecture is not a single aircraft. It is a fleet and infrastructure stack that includes the robots themselves, onboard autonomy, sensing and control, mission planning, charging, teleoperation, and communication/reporting. This organizational point is fundamental: Aerial+ is defined at the level of integrated mission execution.

| Subsystem | Representative elements | Operational role |
|---|---|---|
| Long-range inspection | MARVIN-5-M, Morpho, RGB/event-based tracking, online 3D mapping, heterogeneous multi-UAV planning | Extended-range BVLOS inspection |
| Maintenance/manipulation | DAP-C, MLMP, robotic arm, charging-station installation | Physical interaction with power lines |
| Aerial co-working | Gesture recognition, voltage checking, tool delivery, safety-monitoring formation | Direct support for human workers at height |

A common misconception is to equate Aerial+ with long-range inspection alone. The paper’s own structure rejects that interpretation: the inspection phase is only the first stage, and later stages involve manipulation, energy autonomy, teleoperation, and human–robot collaboration. A plausible implication is that Aerial+ should be understood as a lifecycle architecture rather than as a platform category narrowly defined by airframe design.

## 3. Inspection autonomy, morphing flight, and online perception

For long-range inspection, AERIAL-CORE combines morphing platforms, perception/tracking, online 3D mapping, and heterogeneous multi-UAV planning. Two morphing aircraft are highlighted. MARVIN-5-M is a fixed-wing/rotary-wing morphing VTOL aircraft whose wingspan can be changed autonomously during flight. It can increase wingspan from 1.85 m to 2.41 m, increasing wing area from 0.527 m² to 0.709 m²; the paper notes a minimum stall speed of 13.30 m/s in fixed-wing flight and ground speed over 20 m/s. Morpho is a bioinspired quad, morphing, biplane tailsitter VTOL with four independently actuated wings; it weighs 3.5 kg, has a combined flight time of about 17 minutes, flies about 8 km/h vertically and about 60 km/h horizontally, and has been tested in wind up to 9 m/s while hovering with stable or even reduced energy consumption as wind increased [2401.02343].

Perception is likewise multi-modal. The RGB tracker extends a deep-learning object detector to detect power lines from a single RGB image and outputs endpoints of detected lines in pixel coordinates. It is trained only on synthetic data—about 30k simulated images—and transfers zero-shot to real-world images without fine-tuning. The event-based tracker targets high-speed motion and difficult illumination, exploiting low latency, high dynamic range, and robustness to motion blur, and the paper reports a mean line lifetime 10 times longer than existing approaches. Both were validated onboard a quadrotor with an Nvidia Jetson TX2 [2401.02343].

A major technical element is online 3D semantic mapping for vegetation-risk assessment and infrastructure context. The mapping system is based on FAST-LIO2, modified to incorporate GNSS for robustness when geometric features are sparse. The map is semantically segmented into Powerlines, Towers, Vegetation, and Soil using LIDAR reflectivity region growing and principal component analysis. The implementation used the LR-M robot, based on a DJI Matrice 600 platform with BVLOS capability, a Livox Horizon solid-state LIDAR, and an NVIDIA Jetson Xavier NX. The paper reports more than 80 experiments under varied conditions and vegetation, and notes that a 3.7 km flight in the final demonstration produced the online map shown in the paper [2401.02343].

The planning layer integrates heterogeneous UAVs, autonomous landing, and charging. The reported team consisted of two DJI M210 multirotors and one DeltaQuad Pro fixed-wing VTOL aircraft. The planner exploits the different capabilities of the robots, integrates terrain models for safe positioning, and uses energy-consumption models that account for aerodynamics, wind, and weather to decide when a UAV should recharge. Charging stations are explicitly included in the plan, and visual servoing is used for accurate autonomous landing. According to end-user requirements, the system could inspect more than 10 km of power lines in 10 minutes, while one of the long-range flights was performed up to 10 km away from the ground control station. This matters because it demonstrates a real BVLOS operational capability rather than a lab-scale proof of concept [2401.02343].

## 4. Aerial physical interaction, perching, and energy autonomy

Aerial+ extends beyond sensing to direct maintenance actions on power infrastructure. One manipulation platform is the DAP-C dual-arm system with a rolling base, used to install bird diverters on a real power line. The manipulation system consists of a lightweight compliant anthropomorphic dual-arm robot, LiCAS A1, weighing 3 kg with a 0.7 kg payload, plus a servo-driven rolling base moving at 0.15 m/s and magnetic grippers. A quadrotor with 4 kg payload and about 10 min of flight time deploys the robot onto the line and later retrieves it using a double-cable suspended magnetic hook. The operation has three phases—aerial deployment, device installation, and aerial retrieval—and each phase takes less than two minutes. Stable perching is achieved because the center of mass is about 5 cm below the supporting cable, making the system pendulum-like and stable [2401.02343].

The more general manipulation platform is the MLMP, described as a single UAV that can fly to the line, detect it, perch autonomously, move along the cable, and install or uninstall devices with a robotic arm. The platform can handle live lines up to 125 kV because its grounded metal frame works as a Faraday cage, protecting electronics from electromagnetic interference and absorbing arcs. It uses integrated pulleys driven by DC motors to perch and move along the line. The platform weighs 45 kg MTOW and can fly for about 17 minutes, enough for up to five perching maneuvers or a full multi-device installation task. The paper reports validation for both clip-type bird diverter installation/removal and installation of a heavier charging station, repeated multiple times on the actual power line with excellent results. The authors emphasize that MLMP is the first system of its kind capable of autonomously perching on a line, moving along it, and manipulating devices with a robotic arm [2401.02343].

The manipulator mounted on MLMP is anthropomorphic, has six degrees of freedom and a gear-based spherical wrist, and is designed for dexterity and high payload manipulation. The paper gives a payload of 5 kg while weighing 3 kg, for a payload/weight ratio of 1.67. Different end-effectors are used depending on the task: one for pushing clip-type bird diverters closed, and another for holding and releasing the lightweight charging station onto the line [2401.02343].

Energy autonomy is treated as part of the manipulation ecosystem rather than as a separate logistics issue. A lightweight charging station harvests the magnetic field generated by the power line by means of a split-core current transformer, converting AC energy to DC for charging a battery. Optimization using Transfer Window Alignment combined with Perturb and Observe control and a silicon steel core produced a reported 58.6% increase in harvested power versus traditional methods, and the system can automatically identify the maximum power point without sensing primary current. A larger charging station was also developed for heavier platforms. In Aerial+ terms, this means that field robots can recharge near the infrastructure they are inspecting [2401.02343].

## 5. Human support, teleoperation, and safety monitoring

Aerial+ includes direct support for human workers operating at dangerous elevated locations. For the MLMP, the paper describes an IMU-based teleoperation method in which the operator wears eight IMUs on the upper body. These sensors estimate the operator’s pose, and the robot arm joint velocity references are generated from that pose, with the operator’s resting posture corresponding to zero velocity. Feedback is delivered through camera images projected onto smart glasses, giving the operator both situational awareness and a first-person view from the platform [2401.02343].

The aerial co-working subsystem presents three capabilities: gesture recognition for human–UAV interaction, a contact-based aerial co-worker that can measure voltage and deliver tools, and a multi-UAV safety-monitoring formation. The gesture pipeline uses onboard RGB images, an SSD-based human detector combined with a custom LDES-ODDA tracker, then 2D skeleton extraction followed by an LSTM gesture classifier. The pipeline uses the last $N$ skeleton outputs in a FIFO buffer, updated every $k$ frames, with empirically tuned values of $N = 9$ and $k = 1$. It was pre-trained on manually annotated gesture datasets and fine-tuned on aerial images. This enables a worker to command the UAV using gestures such as “arm up” or “palms together” [2401.02343].

The voltage-check application uses a contact arm with force and potential sensors; the UAV contacts the conductor and compares the measured potential to ground to estimate line voltage, which is displayed at the ground station. The tool-delivery application uses the same gesture pipeline together with a pulley-and-clamp delivery mechanism; the UAV flies to about 2 m above and 1 m beside a lift, then delivers a tool to the worker according to the worker’s gestures [2401.02343].

The co-worker platform itself is a fully actuated hexacopter with fixedly tilted rotors, allowing independent control of position and orientation. It carries a 360° laser sensor and uses an impedance-based control architecture with an outer admittance filter, so that reference trajectories are adjusted based on sensed forces and torques while the inner motion controller still provides high-performance tracking. This is especially important for contact tasks like voltage checking, because it allows compliant interaction under disturbances such as wind. The safety-monitoring subsystem uses a team of Holybro X500-based UAVs with depth cameras, 3D LIDAR, RTK GPS, and RGB cameras. One UAV detects and follows the worker, while others maintain a formation and provide additional views. A multi-stage, MPC-based cooperative motion planning method ensures collision avoidance and continuous monitoring, and the same system can switch from safety monitoring to inspection mode [2401.02343].

## 6. Relation to the wider aerial robotics landscape

Aerial+ can be situated within a larger research trajectory in which aerial robots are moving from observation to interaction. LEGION addresses the long-standing trade-off between nimble flight and robust aerial manipulation through a self-reconfigurable modular aerial robot system. Each unit has three vectored rotors, joint-equipped docking interfaces at both ends, onboard compute, sensing, and control; modules can dock in midair, assemble into a larger articulated aerial manipulator, then disassemble again and return to nimble flight. The paper reports 20/20 docking trials succeeded using onboard sensing only, average docking time with onboard sensing of 12.9 s, and manipulation primitives including pushing, pulling, rotating, grasping, and carrying. This suggests a modular path toward aerial systems that can switch morphology when force transmission matters [2605.19431].

Soft aerial grasping research addresses another component of the Aerial+ agenda: stable interaction under changing payload dynamics. A Soft Aerial Vehicle with a disturbance observer-based NMPC controller estimates and compensates for disturbance forces caused by payload changes and environment effects. The vehicle weighs 1.002 kg and achieves a maximum payload of 337 g by grasping; it successfully grasped and transported three targets, with reported success rates of 27/30 for a shuttlecock tube with 46 g load, 30/30 for a spherical container with 118 g load, and 20/30 for a plastic bottle with 80 g dyed water. A plausible implication is that practical Aerial+ manipulation depends not only on end-effectors but also on disturbance-aware control under payload-induced model mismatch [2409.14115].

Tactile sensing adds local semantic perception and direct contact estimation to aerial interaction. A fully actuated UAV equipped with a GelSight sensor at the end-effector tip uses tactile feedback for real-time force tracking via a hybrid motion-force controller and for wall texture detection during contact. The paper reports approximately 16% improvement in position tracking error when using the fused force estimate compared to relying on a single sensor, 93.4% accuracy in real-time texture recognition, and 100% post-contact. In Aerial+ terms, this expands aerial inspection from remote imaging toward contact-based material and force perception [2310.00142].

Dedicated communications infrastructure provides an additional enabling layer. A hybrid aerial-user cellular network composed of traditional terrestrial BSs and dedicated aerial BSs installed on roadside furniture is modeled by a PPP + PLCP framework, and the paper shows that deployment of dedicated BSs improves aerial coverage probability in both high-density urban areas and rural areas. It also concludes that the optimal density of dedicated BSs which maximizes Max-Min SINR decreases with the increase of the road densities. A plausible implication is that Aerial+ systems operating BVLOS at infrastructure scale will depend on communications layers tuned to road geometry, interference, and mission trajectories rather than on generic ground-user connectivity assumptions [2304.03026].

## 7. Conceptual significance and boundaries

What makes AERIAL-CORE “Aerial+” rather than just aerial robotics is its full-stack integration across the mission lifecycle. The paper states that it goes beyond standard drones in four key ways: first, it operates BVLOS over very long distances under real regulatory constraints; second, it includes morphing aircraft and event/RGB perception for efficient autonomous inspection; third, it incorporates true physical interaction with power lines through perching robots, manipulators, and charging from the line itself; and fourth, it supports humans working at dangerous elevated locations through voltage checking, tool delivery, gesture-based interaction, and safety-monitoring swarms [2401.02343].

This framing also clarifies the boundaries of the term. Aerial+ is not simply an aerial manipulator, because manipulation is only one stage in a larger workflow. It is not simply a swarm, because heterogeneous platforms, charging infrastructure, teleoperation, and reporting are integral. It is not simply a BVLOS inspection system, because maintenance/manipulation and co-working are first-class objectives. Nor is it reducible to any single enabling technology, since morphing airframes, semantic mapping, autonomous landing, perching, line-powered charging, tactile interaction, disturbance-aware control, and aerial communications each address only part of the operational problem.

The broader significance lies in the shift from passive observation to active participation in the environment. In that respect, Aerial+ is aligned with developments in self-assembling aerial manipulation, soft grasping, tactile contact control, and infrastructure-aware aerial connectivity, but it is distinguished by orchestration across the complete mission chain. This suggests that the principal research question for Aerial+ is not whether a drone can inspect, grasp, dock, perch, or communicate in isolation, but how such capabilities can be composed into a reliable operational ecosystem for real infrastructure.

Source: https://www.emergentmind.com/topics/aerial