Decentralised sliding mode control for safety-critical speed tracking of under-actuated high-speed trains under uncertainties

This paper proposes a decentralised sliding mode control (DSMC) framework to address robust speed tracking in large-scale interconnected under-actuated high-speed train systems, which operate as ground-based unmanned autonomous systems (UAS) under internal uncertainties (e.g., parameter variations) and external disturbances (e.g., aerodynamic resistance). By decomposing the under-actuated train model into interconnected subsystems via coordinate transformations, we formulate a DSMC scheme in which each motor-carriage controller uses only its own local speed tracking error and locally available inter-carriage states, without any explicit communication or consensus protocol among subsystems. Unlike conventional distributed control, the inter-carriage couplings and disturbances are treated as matched uncertainties and explicitly embedded into a new large-scale sliding-mode stability condition that guarantees uniform ultimate boundedness of both the sliding motion and the full closed-loop system. To validate the performance of multi-agent objective tracking in under-actuated large-scale dynamic systems, experiments on an eight-carriage train model with DSMC demonstrate a relative error of about 0.48%, compared with 1.08% for MPC and 9.78% for the PID controller; rigorous boundedness guarantees for both tracking errors and states, providing strong disturbance rejection and resilience to bounded parameter and measurement variations, which is conceptually related to obstacle-avoidance behaviour in UAS but is not intended to cover catastrophic faults under extreme operating conditions. The DSMC’s adaptability to heterogeneous sensor data and scalability to multi-agent systems (e.g., drone swarms) highlight its potential for enhancing safety-critical control in broader UAS applications.