The deployment of autonomous robots for close-proximity applications involving physical interactions with other robots or external infrastructure has shifted the focus of control design toward safety during interaction. In these applications, safety violations may lead to undesired collisions and catastrophic failure. Ensuring safety becomes particularly challenging in the presence of parametric uncertainties in the system dynamics. Moreover, when an externally planned safe trajectory is unavailable, the controller must generate the required motion toward the target while inherently enforcing the safety requirements. Accordingly, this dissertation proposes two control approaches to address two fundamental challenges: enforcing strict constraints during trajectory tracking under parametric uncertainties, and formulating state constraints that enable safe physical interaction through reactive control.

The first part of this dissertation presents a tracking controller that tracks a preplanned trajectory for physical interaction while accounting for parametric uncertainties without requiring a priori knowledge of the system dynamics or the bounds of the uncertainties. An adaptive controller is formulated using a Barrier Lyapunov function to enforce strict constraints on the position, orientation, and velocities of the system. The stability of the controller is analyzed using a Lyapunov-like method, and its performance is validated using a space-docking hardware.

The second part presents a control barrier function (CBF)-based safety-critical controller that enables safe close-proximity operations without requiring an externally planned trajectory. Task-specific CBF constraints are formulated for aerial physical interaction, space docking, and UAV landing, and imposed over a nominal go-to-goal controller. The validity of the CBFs is analyzed theoretically, and the resulting controllers are demonstrated through simulation and hardware experiments. The UAV landing framework is further extended to multi-robot and large-scale heterogeneous aerial-ground task execution using centralized and edge-based hybrid centralized-decentralized control architectures.

The third part extends the CBF framework toward two real-world deployment scenarios involving perception limitations and implementation uncertainties. The first scenario considers UAV navigation between two surface vehicles and landing in an outdoor environment with limited visual localization. The control framework incorporates perception constraints for navigation and an adaptive controller to compensate for external disturbances and unmodeled uncertainties. The second scenario considers navigation through cluttered environments with unknown obstacle geometry, where a composite CBF formulation is constructed directly from onboard point-cloud measurements for autonomous infrastructure inspection.

Together, the thesis provides control frameworks for safe close-proximity robotic operations by addressing uncertain dynamics, the absence of reliable externally planned trajectories, multi-robot coordination, and real-world deployment challenges.