Planning a reasonable driving path for trucks in mining areas is a key point to improve mining efficiency. In this paper, a path planning method based on Rapidly-exploring Random Tree Star (RRT*) is proposed, and several optimizations are carried out in the algorithm. Firstly, the selection process of growth target points is optimized. Secondly, the process of selecting the parent node is optimized and a Dubins curve is used to constraint it. Then, the expansion process from tree node to random point is optimized by the gravitational repulsion field method and dynamic step method. In the obstacle detection process, Dubins curve constraint is used, and the bidirectional RRT* algorithm is adopted to speed up the iteration of the algorithm. After that, the obtained paths are smoothed by using the greedy algorithm and cubic B-spline interpolation. In addition, to verify the superiority and correctness of the algorithm, an unmanned mining vehicle kinematic model in the form of front-wheel steering is developed based on the Ackermann steering principle and simulated for CoppeliaSim. In the simulation, the Stanley algorithm is used for path tracking and Reeds-Shepp curve to adjust the final parking attitude of the truck. Finally, the experimental comparison shows that the improved bidirectional RRT* algorithm performs well in the simulation experiment, and outperforms the common RRT* algorithm in various aspects.

Aiming at Articulated Road Roller (ARR) path planning near the construction area, a novel path planning method using Support Vector Machine (SVM) and Longest Accessible Path with Course Correction (LAPCC) is proposed. First, the feasible path selected by Goal-Directed RRT (GDRRT) will be repeated 20 times to determine which category the obstacle belongs to, and then the error penalty factor and kernel parameter of SVM are selected by the grid search and cross-validation parameter optimization algorithm. Then, A set of different virtual obstacles are used on both sides of ARR to control the start and end points of the zero-potential decision boundary. Next, Longest Accessible Path (LAP) is presented to search critical turning points on the decision boundary. Finally, Course Correction (CC) corrects ARR course at these critical turning points. Simulation experiments show that the novel path planning method for ARR using SVM and LAP-CC has the advantages of simplicity, feasibility, low computational cost and good repeatability.

Sign language recognition (SLR) is a bridge linking the hearing impaired and the general public. Some SLR methods using wearable data gloves are not portable enough to provide daily sign language translation service, while visual SLR is more flexible to work with in most scenes. This paper introduces a monocular vision-based approach to SLR. Human skeleton action recognition is proposed to express semantic information, including the representation of signs’ gestures, using the regularization of body joint features and a deep-forest-based semantic classifier with a voting strategy. We test our approach on the public American Sign Language Lexicon Video Dataset (ASLLVD) and a private testing set. It proves to achieve a promising performance and shows a high generalization capability on the testing set.

It requires terrain classification for unmanned Mars Rover to identify the safe areas. The current deep learning-based semantic segmentation and object recognition suffer from a large number of parameters and long training time. In this paper, a lightweight segmentation framework called Mobile-DeepRFB is proposed for the Martian terrain classification. It improves from the DeepLabV3 $+$ by taking the MobileNetV3 as the backbone module to decrease the parameters and the Receptive Field Block (RFB) module to strengthen the feature extraction capability as well as to enlarge the receptive field. Experimental results on the NASA Mars terrain dataset AI4MARS show that the presented method reduces 94% about the parameter number and improves the mean pixel accuracy by 2% compared to the existing ResNet101 and Xception backbone networks. The deployment of this framework on a low-computing power embedded platform (NVIDIA Jetson Xavier) demonstrates its great potential to apply to Mars rovers. Note to Practitioners —This paper was motivated by the problem of terrain classification of planetary rovers. Existing methods are typically based on semantic segmentation technology to recognize various terrains while suffering from the drawback of a large number of parameters. We propose a lightweight segmentation framework to address this issue. In particular, the lightweight backbone network is applied to significantly reduce the number of parameters. The receptive field module is substantially improved to enhance the feature extraction capability. Eventually, we deploy the framework on a low-computing platform. Experimental tests show that the framework can significantly reduce the number of network parameters and it can be used for planetary rovers with limited computational resources.

The special steering characteristics and task complexity of autonomous articulated vehicle (AAV) make it often require multiple forward and backward movements during autonomous driving. In this paper, we present a simple yet effective method, named head correction with fixed wheel position (HC-FWP), for the demand of multiple forward and backward movements. The goal-directed rapid-exploring random tree (GDRRT) algorithm is first used to search for a feasible path in the obstacle map, and then, the farthest node search (FNS) algorithm is applied to obtain a series of key nodes, on which HC-FWP is used to correct AAV heading angles. Simulation experiments with Dynapac CC6200 articulated road roller parameters show that the proposed improved goal-directed rapid-exploring random tree (IGDRRT), consisting of GDRRT, FNS, and HC-FWP, can search a feasible path on maps that require the AAV to move back and forth.