This article reviews key challenges and approaches in enhancing the autonomy of robotic systems across diverse environments. We focus on three fundamental problems: autonomous navigation in GPS-denied conditions, dexterous physical manipulation of objects, and collaborative control of heterogeneous robotic groups. For each area, we analyze current methods, including semantic scene analysis, optimized gripper design, and hybrid swarm-hierarchical control algorithms. The review covers robotic platforms for ground, aerial, surface, and underwater operations, concluding with an overview of systems developed at SPC RAS and ICS RAS for the intertwined challenges of navigation, manipulation, and collaboration is crucial for advancing the operational autonomy of robotic systems in unstructured environments.