Robotic High-Throughput Phenotyping (HTP) technology has been a powerful tool for selecting high-quality crop varieties among large quantities of traits. Due to the advantages of multi-view observation and high accuracy, ground HTP robots have been widely studied in recent years. In this paper, we study an ultra-narrow wheeled robot equipped with RGB-D cameras for inter-row maize HTP. The challenges of the narrow operating space, intensive light changes, and messy cross-leaf interference in rows of maize crops are considered. An in situ and inter-row stem diameter measurement method for HTP robots is proposed. To this end, we first introduce the stem diameter measurement pipeline, in which a convolutional neural network is employed to detect stems, and the point cloud is analyzed to estimate the stem diameters. Second, we present a clustering strategy based on DBSCAN for extracting stem point clouds under the condition that the stem is shaded by dense leaves. Third, we present a point cloud filling strategy to fill the stem region with missing depth values due to the occlusion by other organs. Finally, we employ convex hull and plane projection of the point cloud to estimate the stem diameters. The results show that the R² and RMSE of stem diameter measurement are up to 0.72 and 2.95 mm, demonstrating its effectiveness.

With the advantages of high measurement accuracy and wide detection range, LiDARs have been widely used in information perception research to develop agricultural robots. However, the internal configuration of the laser transmitter layout changes with increasing sensor working duration, which makes it difficult to obtain accurate measurement with calibration files based on factory settings. To solve this problem, we investigate the intrinsic calibration of multi-beam laser sensors. Specifically, we calibrate the five intrinsic parameters of LiDAR with a nonlinear optimization strategy based on static planar models, which include measured distance, rotation angle, pitch angle, horizontal distance, and vertical distance. Firstly, we establish a mathematical model based on the physical structure of LiDAR. Secondly, we calibrate the internal parameters according to the mathematical model and evaluate the measurement accuracy after calibration. Here, we illustrate the parameter calibration with three steps: planar model estimation, objective function construction, and nonlinear optimization. We also introduce the ranging accuracy evaluation metrics, including the standard deviation of the distance from the laser scanning points to the planar models and the 3σ criterion. Finally, the experimental results show that the ranging error of calibrated sensors can be maintained within 3 cm, which verifies the effectiveness of the laser intrinsic calibration.

High-repetitive features in unstructured environments and frequent signal loss of the Global Navigation Satellite System (GNSS) severely limits the development of autonomous robot localization in orchard settings. To address this issue, we propose a LiDAR-based odometry pipeline GLIO, inspired by KISS-ICP and DLIO. GLIO is based on a nonlinear observer with strong global convergence, effectively fusing sensor data from GNSS, IMU, and LiDAR. This approach allows for many potentially interfering and inaccessible relative and absolute measurements, ensuring accurate and robust 6-degree-of-freedom motion estimation in orchard environments. In this framework, GNSS measurements are treated as absolute observation constraints. These measurements are tightly coupled in the prior optimization and scan-to-map stage. During the scan-to-map stage, a novel point-to-point ICP registration with no parameter adjustment is introduced to enhance the point cloud alignment accuracy and improve the robustness of the nonlinear observer. Furthermore, a GNSS health check mechanism, based on the robot’s moving distance, is employed to filter reliable GNSS measurements to prevent odometry crashed by sensor failure. Extensive experiments using multiple public benchmarks and self-collected datasets demonstrate that our approach is comparable to state-of-the-art algorithms and exhibits superior localization capabilities in unstructured environments, achieving an absolute translation error of 0.068 m and an absolute rotation error of 0.856°.

Tomatoes are prone to mechanical damage due to improper gripping forces during automated harvest and postharvest processes. To reduce this damage, a dynamic viscoelastic model based on long short-term memory (LSTM) is proposed to fit the dynamic compression stress relaxation characteristics of the individual fruit. Furthermore, the classical stress relaxation models involved, the triple-element Maxwell and Caputo fractional derivative models, are compared with the LSTM model to validate its performance. Meanwhile, the LSTM and classical stress relaxation models are used to predict the stress relaxation characteristics of tomato fruit with different fruit sizes and compression positions. The results for the whole test dataset show that the LSTM model achieves a RMSE of $2.829\times {10}^{-5}$ Mpa and a MAPE of 0.228%. It significantly outperforms the Caputo fractional derivative model by demonstrating a substantial enhancement with a 37% decrease in RMSE and a 36% reduction in MAPE. Further analysis of individual tomato fruit reveals the LSTM model’s performance, with the minimum RMSE recorded at the septum position being $3.438\times {10}^{-5}$ Mpa, 31% higher than the maximum RMSE at the locule position. Similarly, the lowest MAPE at the septum stands at 0.375%, outperforming the highest MAPE at the locule position by a significant margin of 90%. Moreover, the LSTM model consistently reports the smallest discrepancies between the predicted and observed values compared to classical stress relaxation models. This accuracy suggests that the LSTM model could effectively supplant classical stress relaxation models for predicting stress relaxation changes in individual tomato fruit.