Military autonomy has a practical test close to home. If a base needs soldiers to drive other soldiers between routine appointments, a working transport service could return time to training. Our assessment is that this is a sensible proving ground for autonomous vehicles, provided the service is judged on the work it removes from people as well as the miles it drives.
Forterra and May Mobility announced on November 5 that they had advanced to Phase III of DIU's ROADS challenge. May Mobility is to provide its autonomous driving system and passenger vehicles. Forterra is to operate the fleet and test supervision through OneLink. The release describes mapping in November and a planned month-long Fort Benning pilot. Selection advances prototyping; it does not establish a fleet-wide purchase or a completed driverless service. Read the source account.
What the evidence supports
The division of responsibilities deserves attention. Vehicle motion and fleet operation solve different problems. A vehicle can negotiate a road yet still arrive too late, serve the wrong demand pattern or require an operator to intervene repeatedly. Combining the two suppliers creates an opportunity to test the whole service. It also gives evaluators a reason to track which part of the system causes a missed journey.
The near-term test
In the short term, the credible benefit is better use of an existing transport task. The companies describe onboard safety operators, remote assistance and safe-stop behavior during the challenge. Those safeguards are part of the operating model, and their staffing requirements belong in any calculation of time saved. A ride completed with substantial human support can still be useful, but it has a different cost structure from an unattended service.
The longer-term case
Over time, a reliable base service could let installations organize transport around demand instead of assigning a driver to every vehicle. That outcome depends on availability, maintenance, accessible service and a workable response when a vehicle stops. The strongest counterargument is that a supervised pilot can look efficient while hiding dispatch and recovery effort. Evaluators should count those hours.
The most informative comparison would start with the existing service. If a conventional shuttle already carries passengers efficiently on a fixed route, replacing its driver may deliver a smaller gain than improving dispatch. If journeys are irregular and vehicles spend long periods waiting, a demand-responsive service might address a different problem. The pilot should make that baseline explicit. Otherwise, an improvement caused by better scheduling could be attributed to autonomous driving, while a genuinely useful driving capability could be dismissed because the service was poorly organized. The procurement decision needs to separate those effects before scaling.
Recovery is another part of the service that deserves its own accounting. A stopped vehicle can require remote assistance, a replacement ride and a technician. Each response uses people whose time may be drawn from elsewhere on the installation. That does not make the service unsuccessful. It means a cost comparison should include the operating team behind the vehicle, rather than count only the person who would otherwise sit in the driver's seat. Recording the reason for each intervention would also show whether improvements depend on software, route design, maintenance or a change in local procedures.
There is a sensible limit to what this experiment should be asked to prove. A military base gives access to a real customer and recurring journeys, which can make service quality measurable. It does not reproduce the full demands of a tactical deployment. A positive result would justify a transport decision within the evaluated operating conditions. Expansion should then follow the evidence to other suitable routes and installations. Treating a successful base service as a general verdict on battlefield autonomy would weaken an otherwise useful and specific result.
Our assessment
We would judge the trial by completed requested journeys, passenger waiting time and total personnel hours per journey. A good result would show that the service releases useful time while meeting the installation's safety requirements. That would give defense autonomy a measurable operational case before anyone tries to extrapolate it to a contested battlefield.
Sources and reporting dates
Underlying development date: 2026-11-05. Analysis published November 5, 2026. Company performance claims are attributed to their source; assessments and evaluation criteria are ours.
This analysis draws on the public sources linked in the text. Send corrections to [email protected].
