A Mars rover cannot wait for instructions from Earth before negotiating every rock. Autonomous navigation lets it inspect terrain, identify obstacles and choose a local route toward an assigned destination. People set goals and boundaries while the machine handles some decisions along the way.
The operating cycle connects sensing, state estimation, planning and control. Cameras or other sensors provide data, software estimates the surroundings, a planner selects an action and control systems operate the hardware. Subsequent observations allow corrections. AI may contribute, but autonomy does not require a language model.
Such systems exist in robotics and space exploration at different levels of independence. Autonomous does not mean capable in every situation: operating conditions and human intervention rules matter. The topic connects software agents with sensing and uncrewed vehicles. Uncrewed and autonomous are different properties, since a vehicle without an onboard crew may still be remotely controlled.