Europe crossed 300 GW of installed solar capacity in 2025. To hold up its end of the revised REPowerEU targets, it needs roughly 720 GW by 2030.
Panels can be manufactured at that rate. Sites can be permitted and built at that rate, more or less. What cannot be doubled in five years is the number of people who walk the rows with a thermal camera, wash down a string of modules in August, or go out onto a floating platform to check a mooring line. Maintenance is the quiet constraint on Europe’s solar ambition, and it is a constraint made of working hours.
That arithmetic is where TALOS started three years ago, and where the project’s first policy brief starts too.Autonomous Robotics for Solar PV Maintenance: Removing Barriers to Scale is out now, and its argument is blunt: the machines and the software that can close the maintenance gap already work, and what decides how fast they reach Europe’s solar fleet is the law.
Three years, three very different places
TALOS did its work at three operating sites. A hybrid solar and wind plant at Cruz del Hierro, near Ávila in Spain, covers the utility-scale ground-mounted case. The floating array on the Alqueva reservoir at Portel, in Portugal, covers water. A pear orchard at Randwijk, in the Netherlands, covers agrivoltaics, where panels and a working crop share the same hectares.
Each site placed different demands on the same system. Each ran into a different part of the legal thicket.
What TALOS built is a way of running a plant
The project’s output is an operating method for a solar plant, and it runs as a loop of six steps: monitor, detect, recommend, validate, execute, report.
Sensor and inspection data flow into a data lake and a digital twin of the plant. Fault detection costs each problem in energy and in money, so a hairline crack worth a few kilowatt-hours is ranked against a soiled string that is losing real revenue. A recommendation engine orders the jobs by value and puts them in front of the person responsible for the site.
That person is step four, and every mission waits for their approval before it runs.
The plant manager can move a mission to a different date, send it elsewhere on the site, assign a different machine, or hold it. Once it is under way there is a live map showing battery levels and progress, with a stop control that works at any moment. Afterwards the operator fills in a short feedback form, and what they report shapes how the system ranks the next set of recommendations.
Step five is where the machines come in, and across the three sites there are eleven of them. Inspection drones fly pre-planned routes with ordinary and thermal cameras. A ground vehicle swaps itself between a cleaning rig and a mowing rig. On the reservoir, an uncrewed boat ferries a drone out to the array so it can spend its charge inspecting the panels and not commuting to them, and the cleaning robot there runs on rails built into the platform. In the orchard, a ground robot drives the lanes, checks tree health, re-routes around people and obstacles, and counts fruit for yield forecasts.
The platform is robot-agnostic. Swap the hardware and the loop still runs.
For the people who do this work today, the brief names three changes. Suggestions arrive as decisions to be made, with a cost attached. Human control is explicit and visible at every mission. Evidence of what was done, and why, is captured by default.
Where the law runs out
The barriers TALOS met were regulatory and structural. Robotic operations and maintenance can cut maintenance costs by 25 to 50 per cent and take people out of the situations where accidents happen, and deployment still lags. The brief sorts the reasons into four kinds: member states applying contradictory rules to identical equipment, outright gaps in the law, frameworks written for human-operated machinery, and missing EU definitions for whole categories of installation.
Several EU instruments already touch this ground without covering it. The AI Act may classify maintenance tools as high-risk; the Machinery Regulation does not explicitly address mobile field robots; the UAS Regulation was drafted with leisure flying and logistics in mind.
On land, the brief asks for an “Autonomous Systems in Energy Infrastructure” classification inside the Machinery Regulation, an EU liability rule settling who carries responsibility between manufacturer, integrator and operator, and a harmonised drone category for energy infrastructure work that allows risk-based night operations at enclosed sites.
Agrivoltaics has a stranger problem. It could generate up to 944 TWh a year across 14 million hectares of suitable EU land while the farming continues underneath, and yet it has no legal identity. Is the field agricultural or is it energy? Does the soil still qualify for CAP support?
The brief asks for a legally operative EU definition covering the taxonomy, CAP and RED III, national licensing frameworks within 24 months, ground-vehicle standards for field conditions, and a financing instrument aimed at farms under 50 hectares. TALOS’s social acceptance research found that what convinces a farmer is seeing a neighbour’s results, which is why the recommendation leans on demonstration sites and farmer-to-farmer training.
Floating PV sits under four bodies of law at once: energy, water management, conservation and aviation. Drone permits over protected waters are tight and carry no floating-specific derogation. Here the brief calls for a fast-track “FPV O&M Permit” category, an integrated framework drawn up jointly by DG ENER, DG ENV and EASA, and an environmental assessment standard covering non-harmful ways of keeping birds off the panels.
Five things that apply everywhere
Above the site-specific findings, the brief sets out five systemic gaps.
Regulate smarter, with coherent EU guidance on autonomous systems in renewable energy infrastructure, so an SME is spared reconciling four partial frameworks on its own. Enable access, so high upfront costs in agrivoltaics and floating PV do not leave a market split between large operators and everyone else. Build trust, through workforce upskilling pathways, data governance aligned with the Data Act and GDPR, and community engagement that starts early. Protect and co-exist, with environmental guidelines written for each deployment type. Standardise, so maintenance data stays readable across the single market.
Each one carries a timeframe running from 2026 to 2031 and a named owner, from DG GROW and DG ENER through DG AGRI, DG ENV, DG CNECT and DG EMPL to EASA, ENISA, the EIB and CEN/CENELEC.
Read the brief
The brief is written for the people drafting the next UAS Regulation revision, the CAP review and the Machinery Regulation’s delegated acts. It was written from the experience of the plant managers, farmers and technicians who spent three years approving missions and saying afterwards what they thought.
Autonomous Robotics for Solar PV Maintenance: Removing Barriers to Scale is available from the resources section of the TALOS website under a Creative Commons BY-NC-ND 4.0 licence. The consortium invites the European Commission and member state regulators to take the recommendations forward directly, and can be reached at info@talosproject.eu.
