Before the recent revolution in artificial intelligence (AI), robots were highly limited by their basic hard-coded algorithms. Autonomous robots could not reliably navigate in any complicated environment without significant human intervention. Intelligence was always the primary limiting factor for robots.
But today, you can now find videos of robots doing break dancing, back flips, kick boxing, running, folding clothes, packing boxes and many other impressive tasks — for a robot, that is. The new limiting factor for robots is now hardware.
Source: Gorodenkoff/Adobe Stock
The mechanics — as they are designed today — are no longer able to keep up with the new demands of AI software. Arguably, the most limiting factor for robots today is power source.
Commercial drones typically have a battery life of 20 to 30 minutes. Ground robots, even expensive military types, have only a few active runtime hours at best.
While small combustion engines using fossil fuels offer the highest energy density, they are too loud and polluting for indoor operation, and suffer a range of reliability and maintenance issues. Chemical batteries are much more reliable, have no exhaust, are quiet and require nearly no maintenance. However, they have poor energy density, are heavy and take a long time to recharge.
Powered by fuel cells?
Due to the scaling laws of physics, combined with the economics of the competing technologies, the rule of thumb is that electric systems work better for anything smaller than a motorcycle, whereas combustion performs better for anything larger. Robots, however, sit squarely in the awkward middle — they need electric quietness, zero exhaust and high reliability, yet still require fossil-fuel-like energy density for useful lightweight long endurance.
This is exactly where fuel cells excel. There are many types of fuel cell technologies, but perhaps the most common and safest (temperature wise) would be the hydrogen-fueled proton exchange membrane (PEM). These PEMs draw in oxygen from the environment, combine it with hydrogen from a fuel tank and output electrons, water and waste heat. For the rest of this article, we will only consider hydrogen PEMs.
A typical small motorcycle internal combustion engine achieves 20% to 30% real-world efficiency. PEM fuel cells routinely reach 50%, with peaks near 60% in optimized conditions.
At industrial scale, hydrogen is produced from hydrocarbons (fossil fuels) via steam methane reforming. Because the carbon is stripped away, liquid hydrogen delivers 2.6 times more energy per unit mass than gasoline, giving PEM systems a significant endurance edge.
Factoring in all considerations including fuel tank, battery weight, fuel weight, engine efficiency, engine weight and all other major supporting components, calculated for the same total kilowatt-hour power output, a PEM fuel cell power source weighs approximately 64% of an equivalent internal combustion engine, and approximately 8% of an equivalent battery powered system. The weight savings is significant!
Always tomorrow's technology
As the joke goes, "fuel cells have been the technology of the future — for the last 60-plus years."
The issues holding fuel cells back can be placed into the below three categories. Most notably is the high cost, primarily due to the expensive and exotic materials, such as platinum, iridium, ruthenium and palladium. In the 2010s there was a serious push to develop fuel cell cars. However, the fuel cell engine alone ended up costing as much as an entire conventional car. Since then, there has been much development to avoid and use less of these expensive materials within fuel cells.
Fuel cell reliability is mediocre and they are susceptible to chemical corrosion, catalyst degradation and the formation of pinholes in the membrane. In this sense, they behave similarly to aging batteries, where gradual material breakdown eventually leads to failure.
In some automotive and transit applications, modern PEM fuel cell systems have demonstrated operational durability approaching that of conventional internal combustion engines and lithium battery systems, as evidenced by comparable warranty periods and real-world fleet data.
However, at the smaller scales required for most robots and drones, manufacturers currently warrant fuel cell stacks for only 1,000 to 2,000 operating hours before significant degradation or replacement is required. For robots expected to run near-continuously, such as 24/7 factory automation, this translates to major service or full stack replacement every few months.
Also, hydrogen infrastructure is nearly non-existent and this is effectively a chicken and egg problem. Few suppliers want to spend millions on infrastructure for a questionable market. This hydrogen supply chain issue will likely not resolve without sufficient pre-existing market demand.
That said, these three drawbacks are not always an issue. For example, in military applications, cost is often secondary to efficacy. Military conflicts typically only last months, and it is not unexpected for drones and other robots to suffer destruction within the first few combat missions. Therefore, multi-year reliability is sometimes a non-issue. For drones, weight is a huge factor. Sacrificing durability and cost for roughly 1/13th the weight is an easy tradeoff to make.
Fuel cells can outperform in many other niche applications, such as lunar and planetary robotic exploration, and many others too numerous to mention.
How safe is hydrogen?
We've all seen the dramatic Hindenburg video, the blimp that burned up and psychologically condemned hydrogen for years. Yet, there were no hydrogen flames visible in that video. The "black and white" flames from the famous footage came primarily from the airship's flammable fabric coating and wood structure.
Burning hydrogen produces a nearly invisible flame, emitting light primarily in the ultra-violet and infrared spectrums. At most, you may see a slight blue flame, a faint red flame at night or the rippling of air like the water mirage seen on the surface of a hot desert highway.
So, is hydrogen really that dangerous? In most robotic applications, it is no more hazardous — and in several respects safer — than gasoline or lithium batteries.
Gasoline fumes are heavier than air. They pool and spread horizontally along the ground, feeding fires with ground material and intense radiant heat. Lithium batteries can undergo thermal runaway and are notoriously difficult to extinguish, especially given water can exacerbate lithium fires.
Hydrogen, by contrast, is extremely buoyant (14 times lighter than air). It rises and disperses rapidly, escaping out through the smallest gaps of any enclosed structure. When ignited, hydrogen fires tend to burn out quickly due to fast fuel depletion and upward dispersion, all while producing relatively little radiant heat.
The safety difference is clear. Search online and you will find countless videos of dangerous gasoline and lithium battery fires. In contrast, the vast majority of online hydrogen fire videos are controlled scientific demonstrations rather than actual accidents. A well-known 2001 test comparing a hydrogen-powered car fire to a gasoline-powered car fire illustrates why hydrogen fears are usually unfounded.
As hydrogen flames rocket upwards, modern hydrogen tanks are built with upwards pointing relief valves. Should a rupture occur due to external damage, the expelled hydrogen disperses quickly with the buoyant plume.
Hydrogen is odorless, colorless, tasteless, non-toxic and non-corrosive. On one hand this is a huge advantage over combustion engines and robot workspaces won't end up smelling like a mechanics shop. But on the other hand, it's more difficult to casually detect a hydrogen leak.
Conclusion
While robots have overcome the intelligence barrier, they still lack a reliable, long-lasting portable energy source. Fuel cells have huge potential to address this power bottleneck, delivering quiet, clean, long endurance, high-density energy where batteries fall short and combustion is impractical. Costs are declining, durability is improving and real-world deployments in fuel cell-powered drones, AUVs and military systems now have proven viability.
For indoor, covert, long-endurance or space robots, fuel cells offer a very compelling technical edge. As hydrogen infrastructure matures and exotic material requirements decrease, fuel cells may soon move from "technology of the future" to a standard robotic power source — powering the next generation of autonomous machines on Earth and beyond.
About the author
John Palmisano is a Carnegie Mellon University and CMU Robotics Institute graduate with over 25 years of hands-on robotics and electronics engineering experience. He spent nearly three years as the sole engineer responsible for all fuel cell control electronics and software on the Naval Research Laboratory's record-breaking Ion Tiger drone. He later spent a full year developing fuel cell electronic control systems for a private commercial endeavor seeking to develop fuel cell engines for very large drones.
