Autonomy in military systems is often framed as a technical problem to be solved once and for all. In practice it is an ongoing negotiation between sensors, algorithms, and an adversary who can choose to erase or distort the signals those systems depend upon. Electronic warfare makes that negotiation explicit. It exposes a simple truth: autonomy is only as robust as the information it can reliably trust.
Recent campaigns have offered a practical curriculum in that truth. Adversaries have deployed GPS and datalink jammers at scale, and the effects have been real and measurable. High-power tactical systems, typified in open sources by variants of the R-330Zh family, generate interference across the PNT bands with sufficient power to degrade the accuracy of GPS-reliant munitions and unmanned systems. The result is not always binary failure. Often it is a loss of fidelity that forces a weapon or robot to revert to degraded navigation, to accept larger error bounds, or to abort a mission entirely. These are failure modes that autonomous decision logic must be designed to recognise and tolerate.
But the anecdote of jammers reducing precision tells only part of the story. Electronic warfare is messy in the field. Large jamming installations can be found, targeted, and destroyed. Smaller systems can be limited in range and capability. Forces under pressure adapt quickly. In Ukraine practitioners have both deployed jammers and fielded countermeasures; insurgent and state actors alike have learned that jamming works, but rarely works everywhere and at all times. That operational friction creates windows for autonomy to exploit, and also windows in which autonomy will fail if it has no fallback.
Where autonomy has continued to perform, we see layered engineering rather than magical software. Resilience to jamming is often obtained through a mix of approaches: robust, wideband datalinks that can hop and filter interference; direction-finding and nulling antennas; mission-level routing that uses multiple relay nodes; and graceful degraded modes that let a vehicle complete useful tasks without full connectivity. In the Ukrainian theatre, small teams and companies have pushed software defined and mesh radios into service precisely for these reasons. Those radios do not render jammers impotent. Rather, they force the contest back to physics and spectrum management, buying time and connectivity through multiplicity and agility.
A second axis of resilience is sensor diversity. Navigation and targeting systems that rely exclusively on civil GNSS will predictably fail when the GNSS environment is contested. Modern autonomy must therefore fuse inertial sensors, visual or LiDAR-based odometry, radar, and when available, authenticated military PNT signals. Visual-inertial odometry and tightly-coupled VIO/INS approaches reduce drift and extend operational time in GNSS-denied settings, but they impose constraints. They require sufficient scene texture and lighting, they accrue drift over time, and they add computational and SWaP burdens that small systems may not bear. In other words, autonomy can be designed to survive an hour of GNSS denial, or a day, but not indefinitely without either alternative PNT or periodic absolute fixes.
Practical programs and procurements reflect these trade-offs. Research and small-vendor development into assured PNT and signals-of-opportunity navigation seek alternatives to a single satellite constellation. These projects show how autonomy is becoming less a matter of single-sensor perfection and more an exercise in graceful degradation and cross-checking. The existence of active development for PNT resilience underlines the point that autonomy without a credible PNT fallback is risky in a contested electromagnetic environment.
Operationally the implications are stark. Autonomous behaviours that assume uninterrupted comms or absolute centimeter-level positioning produce brittle systems. A remotely supervised loiter, a precision-guided glide, or cooperative swarm behaviours can all be foiled by timely jamming or spoofing. Engineers must therefore bake in explicit degraded behaviors: return-to-base with conservative safety buffers; mission abort with transparent human notification; conservative use of lethal effectors when target integrity cannot be independently validated. From an ethical and command perspective, these degradations reintroduce the human question. Autonomy should reduce workload, not erase responsibility. Designing systems that fail loudly and safely is a moral as well as technical choice.
There is also a strategic feedback loop to consider. Widespread jamming forces a shift in tactics, which forces technical innovation, which in turn drives adversaries to iterate. The result is an ecology of offense and defense in the electromagnetic spectrum. That ecology favours modularity, rapid software updates, and open experimentation more than monolithic, slow-moving procurement. The conflict in which jammers have been used at scale is therefore instructive not because it proves autonomy is broken, but because it shows how resilience is won in layers: physics, hardware, algorithms, mission design, and human oversight.
Conclusions for architects of robotic warfare are straightforward if uncomfortable. First, do not over-automate assumptions about perfect signals. Second, design for degraded operations and for clear, accountable human roles when sensor integrity is compromised. Third, invest in sensor fusion and alternate PNT early rather than treating them as bolt-ons. And finally, accept that contested electromagnetics reshapes the moral calculus of autonomy: a system that cannot explain its uncertainty is a system that should not be permitted to take irreversible action.
Electronic warfare does not negate autonomy. It disciplines it. The next generation of fielded autonomous systems will not be those that promise total independence from human operators. They will be the ones that admit what they do not know, that fall back gracefully when the spectrum is violent, and that leave decisive moral choices where they belong: with humans who can see the full fog of war.