http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Hongliang Yuan,심태현 한국정밀공학회 2013 International Journal of Precision Engineering and Vol. No.
In this paper, by explicitly considering a dynamic model of the robots, the coefficients of trajectories are determined by boundary conditions, optimal performance index and collision avoidance conditions. The planned trajectory is feasible and has a closed loop expression, which is efficient for real-time updating. There are two main improvements compared with existing parametric approaches. Firstly, most of existing methods use the kinematic models of the robots, which could cause curvature discontinuities when trajectories are updated in real-time. Secondly, in some existing parametric methods, the initial position and ending waypoints cannot be aligned vertically due to singularities. The approach proposed in this paper overcomes this limitation. Computer simulations verified the effectiveness of the proposed method.
Cyclic Error Correction based Q-learning for Mobile Robots Navigation
Rongkuan Tang,Hongliang Yuan 제어·로봇·시스템학회 2017 International Journal of Control, Automation, and Vol.15 No.4
Similar to control systems, reinforcement learning can capture notions of optimal behavior using naturalinteraction experience. In the context of reinforcement learning, the temporal difference error of the generatedexperience measures how well the learner responds to the system. Specially sequential difference of accumulatedtemporal difference error can indicate the learning performance. In this paper, we fully utilize the error correction inclosed-loop peculiarity by mapping a representation error to the step-size component. The proposed cyclic step-sizecould better control how new estimates are iteratively blended together over time, and the new estimates guide theaction selection process which in turn influence the value distribution. To guide more promising action decision, anensemble action selector is proposed which incorporates the idea of ensemble wisdom of the weak. Experimentalresults conducted under gridworld mobile robot navigation task demonstrate the validity, capacity of fast learningand easy-plugged implementation of the derived algorithm, leading to increasing applicability to real-life problems.
Qi Shi,Yongpeng Lei,Yingde Wang,Huaping Wang,Lehua Jiang,Hongliang Yuan,Dong Fang,Bing Wang,Nan Wu,Yanzi Gou 한국물리학회 2015 Current Applied Physics Vol.15 No.12
B, N-codoped carbon nanofibers were massively prepared by heat treatment of electrospun carbon nanofibers with the mixture of boric acid/urea in N2 (BNCf-N) and subsequently activated in NH3 (BNCf- NA). The directly electrospun self-standing 3D non woven structure with void spaces between each fiber facilitates the mass transport of reactant and resulted molecules. Further NH3 activation gives BNCf-NA a high surface area of 306.3 m2 g-1 with micro/mesoporous structure, providing abundant passageway for proton transfer. Simultaneously, NH3 activation also realizes the optimization of surface functionalities, such as more BeNeC and pyridinic-N. These intriguing features render BNCf-NA excellent catalytic behavior with nearly four-electron oxygen reduction reaction (ORR) process in alkaline media, especially much better stability and methanol tolerance than the commercial Pt/C catalyst. Our work provides a large-scale preparation method for efficient metal-free catalysts toward ORR, thus further intensifying the commercial application of fuel cells.