MAYS IBRAHIM (ABU DHABI)
Inside a custom-built arena in Abu Dhabi, autonomous drones will dodge rivals, predict crashes and make split-second decisions at breakneck speed, all without human hands on the controls – just lines of code battling for the fastest, cleanest run.
The A2RL Drone Championship returns for its second season at UMEX Abu Dhabi 2026 on Wednesday, bringing six of the world’s most advanced autonomous AI teams face-to-face with elite human FPV pilots – and with each other.
The inaugural edition in 2025 made global headlines when an autonomous drone defeated a world-champion human FPV pilot. This year, the championship has evolved beyond raw speed.
“It’s about agility, perception, and decision-making,” Stéphane Timpano, CEO of ASPIRE, the ATRC entity driving A2RL, told Aletihad in an interview on Tuesday.
Season 2 introduces a shorter but more technically demanding course, packed with tighter gates, higher-density layouts, and mission-style tasks that test how systems plan, adapt, and respond under pressure.
The drones are capable of exceeding 140 kilometres per hour, reaching peak speeds on straight tracks.
“Collisions can happen in milliseconds,” Timpano said. “Avoiding them becomes just as important as finding the fastest line.”
The two-day championship carries a $600,000 prize pool, split across three race formats that test different capabilities.
The AI Speed Challenge sees single drones complete two laps of the track as quickly as possible. The best teams are expected to finish in under 15 seconds, according to Timpano.
The AI vs AI Multi-Drone Race places three or four autonomous drones on track at the same time, forcing them to balance optimal trajectories with real-time collision avoidance.
The event culminates in the Human vs AI Showdown, pitting the top four AI teams against four world-class FPV pilots.
“This year, the gap is even smaller,” Timpano said. “We’re talking milliseconds.”
To keep the competition focused on software rather than hardware, all teams race identical drones using a deliberately minimalist sensor setup: a single RGB camera, an inertial measurement unit and fully onboard computing.
There is no GPS, no LiDAR and no external processing.
“What separates winning systems is how they process information,” Timpano explained. “They all know where the gates are. The challenge starts when another drone enters the picture.”
Teams must continuously interpret visual data, predict the behaviour of competitors and make instant decisions at speeds that push the limits of perception. Timpano identified lighting sensitivity as one of the main technical bottlenecks for fully autonomous drones right now.
Changing or excessive lighting can confuse the camera-based vision system, making it harder for the AI to interpret gates and surroundings accurately, he explained. As a result, the race environment maintains stable, controlled lighting conditions.
More than 150 teams from around the world entered the qualification process, with just six making it to the final grid.
Among the finalists is TII Racing, the in-house team of Abu Dhabi’s Technology Innovation Institute, competing independently alongside international rivals from the Netherlands, South Korea, Türkiye and the UAE.
The league is part of a broader push by the Abu Dhabi Autonomous Racing League, created under the Advanced Technology Research Council, to accelerate real-world autonomy.
Prior to the race, teams were given access to advanced simulators, allowing them to repeatedly test, fail and refine their systems.
“They had the chance to try, fail, try again, and improve their software,” Timpano said.
“Simulation is critical. You can break a virtual drone hundreds of times without consequences, and that’s how you build the right behaviours for the real world.”
By placing AI systems under extreme, transparent conditions, A2RL generates performance benchmarks that are difficult to replicate in laboratories alone.
The insights feed into applications ranging from autonomous inspection and logistics to emergency response and future air mobility.