Caterpillar joins the Army’s autonomous breaching contest

The US Army has selected Caterpillar, Forterra, IDV USA, and Overland AI to develop competing autonomous systems for clearing minefields and complex battlefield obstacles.

The Engineer Autonomous Breaching Capability programme is intended to reduce soldier exposure during a mission that normally places personnel and crewed engineering vehicles close to mines, direct fire, artillery, and prepared defensive positions.

The selected approaches range from autonomous adaptations of commercial heavy equipment to purpose-built robotic vehicles with interchangeable mission tools. Prototype development is expected to lead to an assessment with a Transformation in Contact unit in early 2027.

Breaching places unusual demands on autonomy because the terrain is deliberately difficult. Vehicles may encounter buried and surface-laid mines, wire, ditches, rubble, craters, anti-vehicle obstacles, damaged infrastructure, smoke, dust, and routes that change as the operation develops.

A breacher also carries equipment whose behaviour affects mobility. Ploughs, rollers, line charges, excavators, marking systems, and other tools change weight distribution, steering response, visibility, and the forces acting on the vehicle.

The Army’s interest in beyond-line-of-sight control moves the programme beyond simple tele-operation. A remotely driven platform can separate its operator from immediate danger, yet it still depends on continuous video and control links that may degrade behind terrain or under electronic attack.

Greater onboard autonomy would allow an operator to assign a route or task while the vehicle controls local steering, speed, obstacle avoidance, and tool operation. Human supervision could then concentrate on exceptions and tactical decisions rather than every movement of the controls.

Commercial machinery offers an established industrial base. Caterpillar’s construction and mining equipment already uses automation, remote operation, mature drivetrains, hydraulic systems, and worldwide parts support, providing a comparatively rapid route into prototype work.

Military use introduces requirements that commercial machines do not normally face. Blast, fragments, electronic warfare, rapid transport, signature control, degraded fuels, battlefield recovery, and operation without established maintenance facilities may require structural and systems changes.

Forterra and Overland AI bring autonomous ground-vehicle software and military experimentation, while IDV USA contributes vehicle integration, defence manufacturing, and through-life support experience. The mix allows the Army to compare different industrial models rather than placing the complete requirement within one traditional armoured-vehicle programme.

Purpose-built robotic platforms can be optimised around transportability, protection, sensors, and mission equipment, although they begin with smaller production volumes and less mature support networks than commercial machinery.

Autonomous conversions can reach trials sooner, but the original vehicle may contain excess mass, dimensions, emissions, or components unsuited to military service. The assessment will need to measure performance and support burden rather than comparing headline autonomy functions alone.

Modular tools could spread production volume across several engineering missions. A common carrier fitted with breaching, earthmoving, route-clearance, marking, recovery, or material-handling equipment would support a larger fleet and reduce the number of unique vehicles.

Interfaces have to withstand substantial mechanical loads while providing power, hydraulics, data, and safe attachment. Poorly controlled modules can alter stability or steering enough to invalidate the autonomy models developed for the base vehicle.

Manufacturing variation also affects robotic behaviour. Differences in braking, steering, hydraulic response, sensor alignment, tyre or track condition, and payload movement influence how accurately the control software predicts each vehicle.

A system tuned on one prototype may behave differently across a production batch unless those characteristics are measured and kept within defined limits. End-of-line calibration and automated functional tests should therefore form part of the production system.

Operational use will be harsher than most autonomy trials. Mud, dust, smoke, blast damage, broken tracks, obstructed cameras, damaged antennas, and fouled lidar can gradually reduce capability rather than causing a clean system failure.

Vehicles need to recognise that degradation and either continue at reduced performance, request assistance, or stop without blocking the breach. Replaceable sensor modules, accessible electronics, and clear diagnostic tools would allow engineers to restore systems without returning them to a specialist robotics facility.

Overland AI has already moved autonomous ground systems towards funded operational use through work supporting mobile air defence, covered in the examination of US Marine autonomous vehicles entering production tempo. Breaching adds heavier equipment and a less predictable physical environment.

Communications remain a critical dependency. Beyond-line-of-sight operation may use radio relays, satellite links, other vehicles, or airborne nodes, yet every path can be jammed, intercepted, or lost behind terrain.

The control architecture should authenticate commands, resist takeover, and define safe behaviour after a link failure. It must also avoid requiring so much bandwidth that several vehicles cannot operate together.

Operator workload will be a decisive measure during the 2027 assessment. A system that needs one person continuously driving each vehicle has improved crew protection but has not delivered the larger benefits of autonomous teaming.

Useful autonomy should allow a small team to supervise several machines, intervene where necessary, and understand each vehicle’s status without monitoring multiple uninterrupted video feeds.

Cost will shape how aggressively the systems can be used. A breacher needs enough protection and reliability to complete its task, but an excessively expensive platform may be withheld from the minefield it was designed to enter.

Commercial components can reduce cost and increase availability, provided they survive military conditions. Purpose-built parts may improve performance while narrowing the supplier base and raising replacement times.

The Army’s four-team structure creates a comparison among commercial scale, military integration, autonomy maturity, and dedicated robotic design. The resulting production choice will reveal whether autonomous breaching is best treated as a conversion of established machinery or as a new vehicle class built around the absence of a crew.