Skypuzzler Bets Math Beats AI for Drone Airspace Safety

Jesper Skou founded Skypuzzler about 15 years ago alongside another air traffic controller and a computer scientist. As a former air traffic controller responsible for all air traffic in Denmark and Sweden, Skou brought domain expertise to the problem of coordinating multiple aircraft in shared airspace—a challenge that now applies to drones and manned aviation operating together at low altitude.

The company's platform combines strategic deconfliction (pre-departure conflict prevention) and tactical deconfliction (real-time conflict management during flight), both based on mathematical algorithms. In Europe, Skypuzzler has partnered with aerospace and defense company Thales Group and DSV, the world's largest logistics company, which operates an extensive commercial drone fleet. The system currently manages operations at the Port of Rotterdam in the Netherlands, where nearly 100 different drone operators perform diverse missions—cargo transport, survey work to detect oil leaks, and other tasks—all in the same airspace.

According to Skou, Skypuzzler's solution requires no additional hardware on drones; it relies on data already transmitted to UTM systems or infrastructure. The platform can operate fully autonomously or under human supervision. When operated with human oversight (as DSV prefers), controllers receive alerts flagging potential conflicts and are presented with conflict-resolution suggestions they can accept or reject—similar to air traffic control instructions on manned flights. This hybrid model allows a single drone pilot to supervise the operation of many UAVs, dramatically improving efficiency. Skou cites DSV's example: instead of hiring ten pilots for ten drones, the company can deploy one pilot supervising five or ten drones delivering spare parts to offshore wind farms within minutes to an hour, replacing slower vessel and helicopter deliveries.

United Airlines Ventures recently invested in Skypuzzler, signaling the company's intention to enter the U.S. market, particularly Dallas-Fort Worth, which Skou describes as one of the most active regions for drone delivery and UTM development. The FAA has already approved deliveries by Wing, Zipline, and Flytrex, and last year Flytrex and Wing announced the first U.S. implementation of the shared UTM Strategic Coordination standard. However, Skou argues that DFW's current model—coordinating only a small number of operators before departure—will not survive as manned and unmanned aviation multiply. His vision is to integrate Skypuzzler's software into existing U.S. UTM platforms such as OneSky or ANRA Technologies on a "plug and play" basis.

Crucially, Skypuzzler's competitive pitch rests on rejecting AI for aviation safety. The platform uses a deterministic mathematical model: the same input always produces the same output, and any error can be traced to its source. Competitors, Skou says, have built similar solutions using AI and machine learning, which produce unpredictable outputs—a "black box" approach incompatible with aviation safety requirements. Skou believes AI may eventually be suitable for flight safety in five to ten years but has not yet matured sufficiently. This stance creates a marketing challenge: Skou reports that some investors dismiss non-AI solutions outright, forcing him to explain why mathematical determinism, not algorithmic learning, is the right tool for airspace deconfliction.

Skypuzzler enters the U.S. market at a pivotal moment. The FAA is actively managing the transition to dense, multi-operator low-altitude airspace, having already approved drone delivery operations by companies such as Wing, Zipline, and Flytrex in regions like Dallas-Fort Worth. However, Skou identifies a critical scaling problem: the current model in places like DFW coordinates only a handful of operators before they launch, which works for early deployments but will break down as more drones, helicopters, and manned aircraft compete for the same airspace.

Skypuzzler's European track record—managing nearly 100 concurrent drone operators at Rotterdam and supporting logistics giant DSV's operations—suggests the system can handle complexity that pre-departure coordination alone cannot. The company's core claim rests on a technical distinction: deterministic mathematical algorithms provide provable, repeatable safety guarantees, whereas AI models operate as "black boxes" and can yield inconsistent results. Skou acknowledges AI has potential in aviation safety, but argues it needs five to ten more years of development. This position puts Skypuzzler against the current investment enthusiasm for AI, a challenge Skou admits makes fundraising and customer acquisition harder, even as United Airlines Ventures' backing signals that at least one major carrier sees merit in the algorithmic approach.

No comments yet. Be the first to share your thoughts!