Abstract
Path tracking is a fundamental capability in autonomous mobile robotics, particularly under disturbances. Regulated Pure Pursuit (RPP), available within the Robot Operating System 2 (ROS 2) Navigation2 framework, extends conventional Pure Pursuit through velocity-scaled lookahead and curvature-dependent velocity regulation. However, limited experimental evidence exists on whether RPP configurations that perform favorably under nominal conditions maintain their performance when subjected to external deviations. This work presents a real-world evaluation of three RPP parameterizations: a baseline configuration (R0), a short-lookahead configuration with more conservative velocity regulation (R1), and a long-lookahead configuration with more anticipative behavior (R2). The evaluation comprised 18 real-robot trials over an approximately 18.2 m reference trajectory, including 27 perturbation events. Performance was assessed using conventional tracking metrics and a recovery-oriented analysis based on configuration-specific nominal error profiles and an explicit recovery criterion. Under nominal conditions, R1 achieved the lowest root mean square error (RMSE) (0.054 ± 0.004 m), approximately 59% and 65% lower than R0 and R2, respectively. Under controlled perturbations, R1 also maintained the lowest whole-trial RMSE (0.151 ± 0.019 m) and recovered in all 9 perturbation events, compared with 4 of 9 for R0 and 5 of 9 for R2. Thus, among the evaluated configurations, the RPP parameterization providing the best nominal tracking performance also exhibited the most favorable behavior under controlled perturbations. These results demonstrate that controlled perturbation and recovery analysis provide complementary information beyond nominal tracking metrics for the experimental evaluation and selection of RPP parameterizations.