The quadruped/biped reconfigurable walking robot with parallel leg mechanism can realize not only the quadruped walking, but also the biped walking. The converting process from the quadruped to the biped includes lock...
详细信息
The network file system (NFS) protocol, as the de facto standard for sharing files in a distributed environment, has deployed Infiniband as the underlying transport of sunRPC, namely NFS over RDMA. In the current Read...
详细信息
The network file system (NFS) protocol, as the de facto standard for sharing files in a distributed environment, has deployed Infiniband as the underlying transport of sunRPC, namely NFS over RDMA. In the current Read-Write design of NFS over RDMA, NFS write performance is limited for not fully utilizing the features of Infiniband. In this paper, we take on the challenge of enhancing the write performance of NFS. We propose and evaluate a new design of sunRPC over RDMA, namely Write-Write design. To guarantee the security of our design, we propose an HCA-based memory protection extension of Infiniband. Evaluations show that our Write-Write design increases the kernel-to-kernel RPC bandwidth by 15~27%. In real disk test, our Write-Write design gains 15%~22% in multi-client benchmarks compared with the Read-Write design.
Codelet model is a fine-grained, event-driven hybrid parallel model inspired by dataflow, whose performance depends on the scheduling policy. How to design optimal codelet scheduling policy based on the features of ta...
详细信息
ISBN:
(纸本)9781509032068
Codelet model is a fine-grained, event-driven hybrid parallel model inspired by dataflow, whose performance depends on the scheduling policy. How to design optimal codelet scheduling policy based on the features of tasks is important to the codelet-based system performance. In this paper, we propose an adaptive codelet scheduling policy by combing "pure" genetic algorithm for tasks with complex dependencies. It is verified that the policy is effective based on bunches of experimental results.
Sink Scheduling, in the form of scheduling multiple sinks among sink sites to leverage traffic burden, is an effective mechanism for the energy-efficiency of wireless sensor networks (WSNs). Due to the inherent diffic...
详细信息
Decision making for autonomous driving is a safety-critical control problem. Prior works of safe reinforcement learning either tackle the problem with reward shaping or with modifying the reinforcement learning explor...
详细信息
ISBN:
(数字)9781728185262
ISBN:
(纸本)9781728185279
Decision making for autonomous driving is a safety-critical control problem. Prior works of safe reinforcement learning either tackle the problem with reward shaping or with modifying the reinforcement learning exploration process. However, the former cannot guarantee the safety during the learning process, while the latter relies heavily on expertise to design exquisite exploration policy. Currently, only short-term decision makings for low-speed driving were achieved in road scenes with basic geometries. In this paper, we propose a two-stage safe reinforcement learning algorithm to automatically learn a long-term policy for high-speed driving that guarantees safety during the entire training. In the first learning stage, model-free reinforcement learning is followed by a rule-based safeguard module to avoid danger at low speed without expert ne-tuning. In the second learning stage, the rule-based module is replaced with a data-driven counterpart to develop a closed-form analytical safety solution for high-speed driving. Moreover, an adaptive reward function is designed to match the different objectives of the two learning stages for faster convergence to an optimal policy. Experiments are conducted on a racing simulator TORCS which has complex racing tracks (e.g. sharp turns, hills). Compared with the state-of-the-art baselines, the results show that our method achieves zero safety violation and quickly converges to a more efficient and stable policy with an average speed of 127 km/h (3.3% higher than the best result of baselines) and an average swing of 3.96 degrees.
In this paper,volume models are obtained from closed surface models by an accurate voxelization method which can handle the hidden cavities. This kind of 3D binary images is then converted to gray-level images by a fa...
详细信息
In this paper,volume models are obtained from closed surface models by an accurate voxelization method which can handle the hidden cavities. This kind of 3D binary images is then converted to gray-level images by a fast Euclidean distance transform (EDT).Moment invariants (MIs) which are invariant shape descriptors under similarity transformations,are then computed based on the gray images. Applications in shape analysis area such as principal axis determination,skeleton and medial axis extraction,and shape retrieval can be carried out base on EDT and MIs.
MemoryIO, a sort of extended I/O in embedded systems, is presented in this paper. MemoryIO makes it powerful for embedded systems to achieve the high-performance interconnect. In view of the facts that the main memory...
详细信息
MemoryIO, a sort of extended I/O in embedded systems, is presented in this paper. MemoryIO makes it powerful for embedded systems to achieve the high-performance interconnect. In view of the facts that the main memory system is absolutely necessary in any embedded system, and not all embedded systems integrate HyperTransport (HT), PCI Express or RapidIO interface, the MemoryIO based interconnect in embedded systems has more universalities compared with that based on HT, PCI Express or RapidIO. MemoryIO can not only thoroughly compensates for the lack of high performance data transfer channel, but also efficiently utilizes the memory bus bandwidth and the direct memory access (DMA) engine to reduce the latency for data transfer in embedded systems. This paper discusses some key technologies of MemoryIO, and presents its application in DCNet and the implementation of MemoryIO IP core. The MemoryIO technology can be used in various systems, but not limited to embedded systems.
An effective interconnect network interface card (NIC) is critical to the achievement of a high-performance cluster system. An original NIC architecture based on the Intel IOP310 I/O processor chipset is presented in ...
详细信息
An effective interconnect network interface card (NIC) is critical to the achievement of a high-performance cluster system. An original NIC architecture based on the Intel IOP310 I/O processor chipset is presented in this paper. The NIC is a part of DCNet, which is the cluster interconnect network developed by NCIC. The I/O processor makes it powerful for the NIC to offload the processing of communication protocol from the host CPU. In the NIC architecture, the memory bus is extended to be a local bus for system peripheral interconnection, and a memory integrated network interface (MINI) is implemented and embedded. The testing results of DCNet show that the NIC obtains reasonable user-level communication performance, and prove that the NIC architecture, which based on I/O processor and the MINI approach, is feasible and effective to achieve the high performance.
As the size of datasets and neural network models increases, automatic parallelization methods for models have become a research hotspot in recent years. The existing auto-parallel methods based on machine learning or...
详细信息
Point matching is a problem of finding the optimum matching between two sets of key points which are extracted from the surfaces of objects. A popular approach represents the features of a set of points with a graph m...
详细信息
ISBN:
(纸本)9781479970995
Point matching is a problem of finding the optimum matching between two sets of key points which are extracted from the surfaces of objects. A popular approach represents the features of a set of points with a graph model. Traditionally, the measurement applied in the graph model is the Euclidian distance, which is not suitable for objects with non-rigid deformations. In this paper, we propose a novel graph model called the geodesic graph model (GGM) which uses a geodesic-like distance as its measurement. GGM can better tackle non-rigid deformations because the geodesic-like distance is a kind of invariant structural feature during non-rigid deformations. The building process of the GGM is justified under the assumption that all the key points are spanning on a manifold. To further handle the deviations of key point locations, we come up with a feature weighting process to increase our algorithm's robustness. We conduct several experiments on different kinds of deformations over several widely used datasets. Experimental results demonstrate the effectiveness of our algorithm.
暂无评论