Trajectory tracking control method for a crawler-type fruit harvesting robot based on fuzzy immune PID

Autonomous navigation is increasingly required in smart agriculture. However, the unstructured orchard ground has resulted in the low accuracy of path tracking for crawler-type picking robots. This study aims to propose an improved tracking navigation using fuzzy-immune PID control. (1) Dual-antenna real-time kinematic (RTK) device of Beidou satellite was used to calculate the heading angle and the current position of the chassis; (2) An improved pure tracking algorithm was applied to evaluate the current heading deviation of the chassis, which was used as the control quantity for path tracking; (3) The immune feedback mechanism and fuzzy control were combined to adaptively adjust the proportional parameters of the PID, since the classical PID control failed to eliminate steady-state error and slow response, according to the classical PID controller. (4) Fuzzy control was used to correct the integral and differential parameters of the PID. The stability and response speed of the navigation system were improved after optimization. Linear simulation showed that the online distance and time of the proposed algorithm were 70 cm and 2.4 s, respectively, at an initial angle of 30°, an initial lateral deviation of 0, and a forward-looking distance of 50 cm, with an overshoot of 9.8 cm; in immune PID control, the online distance and time were 303 cm and 10.3 s, respectively, with an overshoot of 11.48 cm; in improved PD control, the online distance and time were 252 cm and 8.5 s, respectively, with an overshoot of 11.21 cm; in sliding mode control, the online distance and time were 127 cm and 4.3 s, respectively, with an overshoot of 9.75 cm. Furthermore, the online distance and time of the proposed algorithm were 81 cm and 3.2 s, respectively, at an initial angle of 0°, an initial lateral deviation of 50 cm, and a forward-looking distance of 50 cm; in immune PID control, the online distance and time were 488 cm and 17.4 s, respectively; in improved PD control, the online distance and time were 235 cm and 8 s, respectively; in sliding mode control, the online distance and time were 119 cm and 4.3 s, respectively. In addition, the online distance and time of the proposed algorithm were 107 cm and 5.1 s, respectively, at an initial angle of 0°, an initial lateral deviation of 100 cm, and a forward-looking distance of 80 cm; in immune PID control the online distance and time were 563 cm and 22.2 s, respectively; improved PD control was not online after traveling 10 m; in sliding mode control, the online distance and time were 159 cm and 6.4 s, respectively. Curve simulation showed that the mean lateral deviation of the improved experiments was 2.12, 3.28 and 4.45 cm, respectively, at three speeds, which were 50.2%, 41.8% and 38.5% lower than those of the classic pure pursuit algorithm; the standard deviations were 2, 2.88 and 3.71 cm, respectively, which were 56.3%, 51.4% and 49.7% lower than those of the classic pure pursuit algorithm; the root mean square errors were 2.91, 4.37 and 5.79 cm, respectively, which were 31.7%, 46.6% and 43.9% lower than those of the classic pure pursuit algorithm. Field experiments showed that the mean lateral deviations of the three experiments were 2.95, 3.04 and 3.44 cm, respectively, at the speed of 0.3 m/s, while the root mean square errors were 3.59, 3.71 and 4.34 cm, respectively, and the standard deviations were 2.04, 2.13 and 2.64 cm, respectively; the mean lateral deviations of the three experiments were 3.80, 4.75 and 3.53 cm, respectively, at the speed of 0.4 m/s, while the root mean square errors were 4.56, 5.80 and 6.35 cm, respectively, and the standard deviations were 2.52, 3.33 and 4.15 cm, respectively; the mean lateral deviations of the three experiments were 4.68, 4.97, and 5.67 cm, respectively, at the speed of 0.5 m/s, while the root mean square errors were 5.98, 6.16 and 6.98 cm, respectively, and the standard deviations were 3.73, 3.65 and 4.07 cm, respectively.