Refine
Language
- English (2)
Has Fulltext
- yes (2)
Keywords
- Cluster computing (1)
- Compiler (1)
- Compilers (1)
- Concurrency (1)
- Massive Parallelität (1)
- Mehrkernprozessor (1)
- Multicore (1)
- Nebenläufigkeit (1)
- Parallelism (1)
- Programming Languages (1)
Institute
Irregular data-parallelism in a parallel object-oriented language by means of Collective Replication
(2010)
In parallel object-oriented languages it is hard to elegantly express efficient data-parallel operations on objects of an irregularly-shaped object structure that is spread across the parallel computing environment. This paper presents a new programming model that smoothly integrates both task and data parallelism in a distributed object-oriented context. So called Collective Replication combines enhanced data locality for parallel tasks with data-parallel computations on irregular data structures. Collectively replicated objects exploit the bulk- synchronous data-parallel pattern in an object-oriented language and relieve the programmer from explicitly coding the communication step even for irregular data structures. To make the consistency protocol efficient, only modified fields are shipped instead of whole objects. Moreover, a novel graph coloring approach is used to broadcast updates to all replicas; this technique avoids bottlenecks and is more efficient than known approaches. We suggest Java language extensions that can be han- dled by a pre-processor and achieve good performance on a set of benchmark applications.
Pedestrian Dead Reckoning (PDR) uses inertial measurement units (IMUs) and combines velocity and orientation estimates to determine a position. The estimation of the velocity is still challenging, as the integration of noisy acceleration and angular speed signals over a long period of time causes large drifts. Classic approaches to estimate the velocity optimize for specific applications, sensor positions, and types of movement and require extensive parameter tuning. Our novel hybrid filter combines a convolutional neural network (CNN) and a bidirectional recurrent neural network (BLSTM) (that extract spatial features from the sensor signals and track their temporal relationships) with a linear Kalman filter (LKF) that improves the velocity estimates. Our experiments show the robustness against different movement states and changes in orientation, even in highly dynamic situations. We compare the new architecture with conventional, machine, and deep learning methods and show that from a single non-calibrated IMU, our novel architecture outperforms the state-of-the-art in terms of velocity (≤0.16 m/s) and traveled distance (≤3 m/km). It also generalizes well to different and varying movement speeds and provides accurate and precise velocity estimates.