Answers are grounded in this paper and cite their sources.
Background: Tracked inspection robots in large-diameter pipelines are vulnerable to unequal track adhesion, drive mismatch, slip, speed variation, and payload-induced disturbances, which accumulate into yaw and lateral deviation. Methods: A dual-loop architecture was developed in which an outer path-attraction loop generates a bounded desired heading and an inner heading loop uses a fuzzy adaptive proportional–integral–derivative (PID) controller calibrated offline by particle swarm optimization (PSO). The calibration jointly searches baseline PID gains, fuzzy scaling factors, and small-error freezing thresholds. Derivative filtering, differential-speed saturation, and back-calculation anti-windup were incorporated. Validation comprised random multi-scenario simulations, encoder and inertial measurement unit (IMU) calibration, and 72 physical trials spanning two robot configurations, three speeds, and four control methods. Results: The proposed method achieved a mean post-disturbance root-mean-square lateral-error of 0.00586 m, an engineering settling time of 2.35 s, and a control total variation of 0.663, improving these metrics by 17.2%, 26.9%, and 87.4%, respectively, relative to engineering PID. Relative to fixed PSO–PID, it reduced integral of time-weighted absolute error (ITAE) and quadratic differential-speed command cost but accepted small increases in peak error, engineering settling time, and command variation. In descriptive within-protocol physical comparisons, the reconstructed maximum lateral deviation was reduced by 76.3–91.2% relative to uncompensated motion. Conclusions: Offline global calibration and online local adaptation provide a balanced error–command–cost–smoothness compromise and improve deviation suppression over engineering PID across the tested simulation conditions. The physical evidence remains descriptive because lateral deviation was reconstructed rather than directly measured by an independent positioning system.
No takes yet. Share an insight, caveat, or question.
Luo et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: