Fractional-Order Memory-Enhanced Fixed-Time Anti-Disturbance Tracking Control for Heavy-Duty Cutter-Changing Manipulators Under Payload Jumps and Symmetry Constraints

Heavy-duty cutter-changing manipulators in shield tunneling operate under abrupt payload attachment and release, extraction impacts, load-dependent friction, model uncertainty, actuator saturation, and measurement noise. These coupled effects can produce large transient tracking errors and excessive actuator demand. This paper develops a fractional-order memory-enhanced fixed-time symmetry-constrained tracking method. Payload-induced inertia variation, gravity mismatch, nonlinear friction, impacts, and unmodeled dynamics are aggregated into a heavy-load lumped disturbance. An acceleration-free momentum residual is combined with a Caputo fractional-order disturbance observer. For each joint, the tracking error is restricted to a zero-centered time-varying symmetric set, and an odd invertible transformation maps the constrained error into an unconstrained coordinate. A fixed-time terminal residual is then constructed in the transformed coordinates. Instead of placing a fractional integral directly in the sliding manifold, the proposed method introduces a nonnegative fractional memory term into the robust reaching gain, which preserves a standard fixed-time Lyapunov certificate while retaining historical residual information. A load-aware gain schedule is generated from the estimated disturbance magnitude. The fractional observer satisfies a componentwise Mittag–Leffler practical-error bound for bounded Caputo derivatives of the filtered input. The ideal sign-based reaching and terminal dynamics admit fixed-time bounds provided that the combined estimation and saturation residual remains below an explicit robustness margin; the smooth finite-memory implementation is evaluated numerically. Numerical studies on a two-link heavy-duty manipulator compare four controllers and examine ablations, fractional-order sensitivity, payload jumps, impacts, saturation and noisy model mismatch. Randomized paired trials quantify the loss of constraint feasibility under uncertain loading and saturation; computational-cost measurements characterize the offline implementation.