Stochastic optimization of a chain sliding mode controller for the mobile robot maneuvering

Terekhov, Alexander V.; Mouret, Jean-Baptiste; Grand, Christophe · 2011 · OpenAlex-citations

DOI: 10.1109/iros.2011.6094956

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of controlling a four-wheeled autonomous mobile robot during aggressive turning maneuvers on loose surfaces, a task where traditional path-tracking algorithms often fail due to vehicle drift and non-holonomic constraints. The authors aim to develop a controller capable of performance comparable to professional rally drivers, specifically targeting a 90-degree turn. The motivation stems from the limitations of existing virtual vehicle algorithms, which typically exclude drift, and the difficulty of designing feedback stabilization for highly non-linear, locally non-controllable lateral movements. To solve this, the authors propose a "chain sliding mode controller," which consists of a sequence of local second-order sliding mode controllers. These local controllers are activated sequentially as the robot's state crosses specific switching hyperplanes in the state space. The parameters defining these controllers and switching surfaces are optimized using multiobjective stochastic optimization, specifically the NSGA-II algorithm. The optimization process balances two conflicting objectives: maximizing the average speed of the maneuver and minimizing the deviation from the desired trajectory. To prevent overfitting to a specific simulation, the optimization includes various model disturbances, such as changes in friction coefficients, maximum velocity, and vehicle mass/inertia. The robot's dynamics are modeled using a brush model for wheel-road interaction, accounting for weight redistribution and slip. The results demonstrate that the optimized controller successfully guides the robot through the aggressive turn. In simulations, the controller achieved high precision, with a maximum trajectory deviation of less than 6 cm. The controller induces significant understeering, causing the robot to decelerate and brake near the turn apex while slipping laterally. When tested on the physical "fastBot 2" robot, the controller successfully executed the maneuver, although performance was lower than in simulations due to discrepancies in initial speed and surface friction. Crucially, the experimental results showed that the steering radius achieved with this controller was two times smaller than the minimal steering radius admitted by the robot's mechanical constraints, validating the controller's ability to exploit drift for tighter maneuvering. The significance of this work lies in demonstrating that stochastic optimization can effectively tune complex, non-linear sliding mode controllers for aggressive robotic maneuvers. By combining feedforward optimal trajectory execution with feedback stabilization via a chain of sliding modes, the approach overcomes the limitations of linear feedback systems in highly dynamic, non-holonomic scenarios. This method offers a viable path for improving the maneuverability and speed of autonomous vehicles operating in unstructured environments where drift is inevitable.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• steering pattern