Friction‐Aware Force‐Realizable Safety Control for Four‐Wheel Independently Steered and Driven Mobile Robots

ABSTRACT Safety‐critical control remains a major challenge for four‐wheel independently steered and driven (4ISD) mobile robots, where high‐accuracy path tracking and obstacle avoidance must be achieved under over‐actuated steering–driving coordination, tire–ground friction limits, and wheel‐torque constraints. This article develops an observer‐based safety‐critical control framework that links recoverability‐aware barrier regulation with force‐aligned torque allocation. A friction witness reconstructed from wheel‐force channels is processed by a generalized super‐twisting algorithm (GSTA) observer to obtain a conservative lower bound on available friction. This bound is used by a filtered sampled‐data control barrier function governor that adapts the safety radius according to friction‐limited recoverability and command‐filter realization error. The resulting virtual acceleration is then converted into wheel torques by a backstepping dynamic layer that conditions the demanded generalized force against the friction‐limited force set before over‐actuated longitudinal‐force allocation. The analysis proves cascade practical input‐to‐state stability for the observer–governor–dynamic‐control interconnection and sampled‐data practical safety for the filtered barrier recursion. In a two‐factor ablation on a complex low‐friction obstacle‐course task, the full controller increases the minimum safety margin from to , reduces the peak wheel torque from to , and lowers the RMS realization error from to relative to a CLF–CBF–QP and traction‐allocation baseline. Under nominal pose noise and bounded model mismatch, the same controller reaches the goal in all tested cases while exhibiting positive measured safety margins; the adaptive friction lower bound reduces conservative tracking loss without producing a safety‐margin violation.