Self-driving ground vehicles are now in the Ukraine war

The vehicles, built by a company called Forterra, are designed to handle the heavy lifting that smaller, battery-powered drones cannot. While a typical small drone might carry a few pounds, these gas-powered robotic carts can carry up to 750 kilograms of cargo. They have completed over 1,100 missions and 52 evacuations, racking up 2,500 miles on the ground. However, these machines are not fully autonomous in the way a self-driving car might be. For now, Ukrainian soldiers primarily steer these vehicles remotely rather than letting the machine navigate unsupervised. This is partly for safety; the vehicles are expensive, and losing one to a mine or an attack is a major loss for the unit.

Modern autonomy relies on two different approaches operating in tandem. First, there is classical robotics, which functions like a very sophisticated, rigid rulebook. This is how the vehicle knows how to keep its balance on hills or follow a GPS path through a forest. It is excellent at mechanical movement but terrible at understanding surprise. If the vehicle encounters an enemy soldier or a new, sudden hazard, the classical system does not know how to interpret that as a threat because it hasn't been programmed for that specific scenario. To solve this, companies are trying to integrate newer artificial intelligence. The goal is to build software that can look at a scene and analyze it generally, the way a person does, rather than needing an exact, pre-coded instruction for every possibility. Currently, companies are struggling to find the right data to teach these systems, as no amount of training on civilian self-driving car footage can prepare a machine for the chaotic, unpredictable environment of a minefield or an active combat zone.

The deployment of these vehicles reveals that we have reached a point where moving robots across rough terrain is no longer the hardest part; the real challenge is teaching them how to act when they enter a war zone. For soldiers on the ground, these tools are already valuable for simple, dangerous logistics tasks, like carrying wounded soldiers out of reach of artillery. However, until these machines can perceive and react to changing threats on their own, they will remain remote-controlled tools rather than truly autonomous teammates. This shift is turning the front lines into an intense testing ground, where the feedback loop—the difference between a piece of equipment that works in a lab and one that survives in the mud—is moving faster than ever before.