Track skid and slip often occur, when the tracking chassis work in soft soil and uneven field. The accuracy of curved path tracking can inevitably deteriorate in automatic navigation. In this work, path tracking was designed to consider the track skid and slip in field conditions. Firstly, the coefficient of turning radius correction was derived to describe the skid and slip characteristics using the kinematic model of the tracked chassis. Then, the lateral and heading deviations were determined by the look-ahead model. Finally, a three-input fuzzy controller was established with the controller inputs of lateral deviation, heading deviation and turning radius correction coefficient. A dual-motor tracked chassis was designed to validate the model. The control hardware included two brushless DC motors, a motor driver, an STM32 microcontroller, and two photoelectric rotary encoders. In navigation, the position, heading and angular velocity of the chassis were measured by a high-precision RTK-GNSS and an MTi-30 inertial sensor. The measurement values were sent to the upper-level computer using a serial port. The path tracking was carried out to calculate the control signal, and then sent into the lower-level microcontroller. The tracking of the reference path was realized by the motor driver of the lower-level microcontroller on both sides. A field experiment was conducted on automatic navigation to verify the performance of a three-input fuzzy controller. The field was selected with a lot of weeds, where the soil was wet and sticky with a moisture content of 22.4%. There was the low flatness with some pits, bulges and hardened soil blocks that scattered throughout the whole field. The speed of the chassis was set to 0.6 m/s in navigation. The look-ahead distance proportional coefficient of the look-ahead model was set to 1.5. An initial lateral deviation of about 1000 cm was set for the correction. The results showed that the average response time of the control system was 6.79 s, while the average distance was 4.08 m. The tracking test of the U-shaped path was performed on the three-input fuzzy controller. The maximum absolute deviation, average absolute deviation and standard deviation of the straight-line and curve tracking section were 12.0 and 21.3 cm, 3.6 and 8.6 cm, as well as 4.4 and 8.4 cm, respectively. The curve tracking with the three-input fuzzy controller was compared with the conventional one, in order to further determine the path tracking accuracy under track skid and slip. The maximum absolute deviation, average absolute deviation and standard deviation of the conventional and three-input fuzzy controller were 31.1 and 22.4 cm, 8.3 and 7.4 cm, as well as 10.2 and 9.1 cm, respectively. Compared with the conventional, the three-input fuzzy controller was improved by 27.95%, 10.84 % and 10.78 %, respectively, in the maximum absolute deviation, average absolute deviation and standard deviation. The better control performance of the navigation system was achieved in the three-input fuzzy controller, indicating the reduced influence of track skid and slip. This finding can also provide a strong reference for the high-precision navigation of tracked chassis in soft soil environments in the field.

Currently, self-propelled trimming machines for tea gardens have problems of low automation level and poor maneuverability. In this work, a contact-style automatic ridge-following navigation system and its corresponding control algorithm for a small-scale, wheeled trimming machine in tea gardens were proposed considering the structural features of tea ridge boundaries. A contact-style automatic ridge detection mechanism based on low-cost laser ranging sensors was designed.The working principle of the detection mechanism was introduced and analyzed. The position and heading information of the trimming machine with respect to the tea ridge was derived through the opening and closing of two side baffles of the detection mechanism caused by the changing profiles of the tea ridge boundaries. The navigation system hardware included a power supply unit, remote control unit, sensing unit, steering unit, driving unit , and control unit. The sensing unit consisted of two low-cost laser ranging sensors for measuring the changing distances of the baffles of the ridge following the detection mechanism and an absolute optical encoder for determining the real-time steering angle of the steering wheel. The steering unit replaced the original manual steering mechanism of the trimming machine with a stepper motor to realize electrical control. By analyzing the four pose states of the trimming machine with respect to the tea ridge, the kinematic navigation model of the trimming machine was derived, and the lateral and heading deviations of the trimming machine with respect to the tea ridge centerline were obtained. A navigation controller was designed based on an LQR (linear quadratic regulator) algorithm derived from the kinematic model of the trimming machine. The position and heading information of the machine were taken as control inputs. After calculation by the control algorithm, proper steering signals were sent to the stepper motor for deviation correction. To validate the feasibility of the contact-style navigation system and evaluate the performance of the control algorithm, field experiments in a tea garden were conducted. The trimming machine was controlled to automatically follow a tea ridge at three different speeds, 0.2, 0.4 , and 0.6 m/s. The experiment for each speed was repeated three times. Results showed that under three speeds, the maximum absolute deviation of the trimming machine was no larger than 0.134 m, the absolute deviation maximum was no larger than 0.107 m, and the standard deviation maximum was no larger than 0.105 m. This navigation performance was comparable to mainstream navigation systems developed based on RTK-GNSS, vision and laser sensors. The mean absolute deviation and standard deviation increased slightly as the running speed of the trimming machine increased. However, they were kept in the 0.15 m limit. This indicated that the navigation method based on the contact-style automatic ridge-following mechanism was feasible, and the LQR algorithm derived based on the kinematic model of the trimming machine could meet the accuracy requirement of tea ridge following under a practical tea garden environment. The low-cost, contact-style navigation system and its corresponding algorithm proposed in this work can provide a reference for the development of automatic navigation systems for self-propelled operation machines in tea gardens.