No abstract

Benefiting from high mobility and robust mechanical structure, ground mobile robots are widely adopted in the outdoor environment. The mobility of skid-steered mobile robots highly depends on the nonlinear and uncertain interaction between the tire and terrain. This paper introduces an approach to estimate the position, orientation, velocity, and wheel slip for the skid-steered mobile robots navigating on off-road terrains. More specifically, a Multi-Innovation Unscented Kalman Filter (MI-UKF) is developed to fusing different sensors' data. Historical innovations generated along the time sequence are merged into the update process of standard UKF to improve the accuracy of motion estimation. In the proposed estimator, an asymmetric ICR kinematic indicating wheel slip is taken into localization process. A four-wheeled prototype is introduced and three challenging test scenarios are designed. The improvements in orientation and velocity estimation are achieved according to results comparison. In the turning maneuver, the ICRs-based model operates more steady than the traditional wheel slip/skid model.

With robust structure and high maneuverability, skidsteered vehicles have been widely used in terrain exploration, construction, rescue and relief fields. Inevitable slipping and sliding of the tire that makes the vehicle status difficult to obtain. In this paper, an equivalent differential driven kinematic model is proposed. Additionally, the mapping relationship of slippage coefficient defined in the original 4-wheel unmanned skid-steered vehicle and the wheel location in the equivalent model are discussed. The proposed equivalent wheel location obtains explicit physical significance. The slippage of the tire has also been calculated by estimating the position of instantaneous center of rotation (ICR) of the wheels from multi-sensors. An unscented Kalman filter (UKF) based fusion method is adapted to obtain the vehicle status. A prototype vehicle named DUBHE is adapted to verify the reliability of the informed method. Preliminary experimental results are compared to demonstrate the effectiveness of the method in different scenarios.

No abstract

Neural machine translation (NMT) has shown promising progress in recent years. However, for reducing the computational complexity, NMT typically needs to limit its vocabulary scale to a fixed or relatively acceptable size, which leads to the problem of rare word and out-of-vocabulary (OOV). In this paper, we present that the semantic concept information of word can help NMT learn better semantic representation of word and improve the translation accuracy. The key idea is to utilize the external semantic knowledge base WordNet to replace rare words and OOVs with their semantic concepts of WordNet synsets. More specifically, we propose two semantic similarity models to obtain the most similar concepts of rare words and OOVs. Experimental results on 4 translation tasks show that our method outperforms the baseline RNNSearch by 2.38~2.88 BLEU points. Furthermore, the proposed hybrid method by combining BPE and our proposed method can also gain 0.39~0.97 BLEU points improvement over BPE. Experiments and analysis presented in this study also demonstrate that the proposed method can significantly improve translation quality of OOVs in NMT.