The Robotic Combat Vehicle effort is, at its core, an institutional experiment in how a large bureaucracy listens. The Army has moved deliberately to put early prototypes into Soldier hands rather than deferring human factors questions until after a system is fully funded. This soldier centered design approach is visible across Project Origin experiments, Soldier Operational Experiments, and the formal Soldier touchpoints organized by the Next Generation Combat Vehicles Cross-Functional Team. These events are not publicity stunts. They are intended as the mechanism by which tacit battlefield knowledge is translated into technical requirements for platforms, autonomy software, user interfaces, and tactics, techniques and procedures.
That translation process is the feedback loop. In engineering terms it is simple: prototypes exposed to operational conditions generate qualitative and quantitative data; analysts and engineers consume that feedback and iterate designs; revised prototypes are retested; the loop closes when engineering changes are reflected in doctrine, procurement documents, or both. The Army has institutionalized elements of this cycle by running repeated SOEs, multinational rotations, and touchpoints that intentionally place RCV surrogates into messy, contested training environments so that Soldiers and leaders can evaluate utility, limitations and failure modes. The Joint Readiness Training Center and Joint Multinational Readiness Center rotations have been particularly fertile for rapid learning because the tempo and variety of conditions accelerate discovery of edge cases that smaller trials cannot reveal.
There are three virtues to the current approach. First, early exposure reduces the risk of delivering technically polished systems that are tactically irrelevant. Soldiers using Project Origin surrogates repeatedly told evaluators that robots were “game changing” in some missions yet insufficiently valuable in others. That kind of blunt, context-specific feedback is precisely the information engineers need.
Second, repeated touchpoints create a shared vocabulary between developers and operators. When user interfaces such as the Warfighter Machine Interface are exercised in the field, developers stop arguing in abstract about latency or autonomy levels and instead talk about the same observable problems as the Soldier on patrol: mission sequencing, overwatch priorities, communications fragility, and maintenance overhead. That alignment is indispensable if complexity is to be tamed.
Third, the campaign-of-learning model allows incremental technical fixes to be distributed across a portfolio of robotic programs rather than baked into a single, monolithic vehicle. The Army has used modular open systems and common autonomy kernels to propagate lessons across RAS efforts, which economizes learning.
Yet the loop has failure modes that must be acknowledged if RCVs are to mature responsibly. The first risk is representativeness. Soldier touchpoints often rely on limited units, a handful of training rotations, and surrogates that do not perfectly mirror eventual production vehicles. Lessons learned in a Fort Benning or Hohenfels rotation are valuable, but they are not statistically representative of the mosaic of units, terrains, climates, and threat doctrines the Army might face. This selection bias can skew requirements toward the needs of the units tested rather than toward the broader force.
Second, there is a cognitive and organizational latency between feedback capture and institutional change. Engineers can update software between iterations, but changing doctrine, logistics chains, and training pipelines is slow. When soldiers complain that a control scheme imposes excessive cognitive load or that a maintenance cycle is too brittle, the technical fixes may be straightforward while the doctrinal adoption is not. Closing that part of the loop requires sustained leadership attention and cross-domain investment.
Third, feedback is only as useful as the methods used to record and analyze it. Anecdotal soldier impressions are necessary but not sufficient. We need rigorous human factors metrics: time-on-task, error rates under stress, workload indices, physiological indicators of operator strain, and maintenance fault trees. Without systematic mixed-methods data collection, lessons risk becoming colorful but non-actionable testimony that is shelved after a post-exercise briefing.
Finally, trust and ethics are underappreciated loop elements. Soldiers will only employ unmanned systems aggressively if they trust their behavior in life-critical situations and if commanders can legally and morally justify delegation of certain functions. Testing must therefore include scenarios that stress autonomy boundaries, degrade communications, and simulate adversary deception. The feedback loop must not only report on whether a vehicle performs a function but how its performance affects human decision-making, moral responsibility, and unit cohesion. These are not peripheral concerns; they shape design choices from control handover ergonomics to data logging for post-engagement accountability.
What practical changes would tighten the loop? First, broaden and systematize sampling. Expand touchpoints across a wider set of units and environments, and rotate representatives from light, Stryker, and heavy brigades through the same experiment to expose cross-force variance. Second, embed human factors protocols into every SOE by default. Require pre-registered metrics, baseline cognitive assessments, and longitudinal follow-ups so that subjective impressions are anchored in reproducible data. Third, shorten institutional latency by committing to pre-authorized rapid updates of software and training materials based on defined thresholds of operational learning. That means leaders must be willing to accept provisional capabilities and to fund iterative sustainment rather than episodic retrofit. Fourth, integrate red team and ethical stress testing into the campaign of learning so that survivors lessons include adversarial manipulations and moral hazard assessments. These steps align with the Army’s existing campaign-of-learning ethos but push it toward methodological rigor and organizational responsiveness.
If the Army’s RCV program is to be more than an engineering project it must be a social experiment in how a military listens and adapts. The soldier feedback loop is not an optional addendum to procurement. It is the primary technology by which a human institution will domesticate machine systems for morally fraught and mission-essential tasks. Done well, iterative Soldier touchpoints will produce vehicles that make formations safer and more effective. Done poorly, they will produce elegant machines that sit idle because the institution never aligned training, logistics, doctrine, and legal frameworks with the promises of autonomy. The test of success is not whether a vehicle clears a mobility course. The test is whether the Army can translate messy, on-the-ground soldier wisdom into enduring changes in design, doctrine, and culture.