C2RL: Convolutional-Contrastive Learning for Reinforcement Learning Based on Self-Pretraining for Strong Augmentation.

Reinforcement learning agents that have not been seen during training must be robust in test environments. However, the generalization problem is challenging to solve in reinforcement learning using high-dimensional images as the input. The addition of a self-supervised learning framework with data augmentation in the reinforcement learning architecture can promote generalization to a certain extent. However, excessively large changes in the input images may disturb reinforcement learning. Therefore, we propose a contrastive learning method that can help manage the trade-off relationship between the performance of reinforcement learning and auxiliary tasks against the data augmentation strength. In this framework, strong augmentation does not disturb reinforcement learning and instead maximizes the auxiliary effect for generalization. Results of experiments on the DeepMind Control suite demonstrate that the proposed method effectively uses strong data augmentation and achieves a higher generalization than the existing methods.

A Simultaneous Pipe-Attribute and PIG-Pose Estimation (SPPE) Using 3-D Point Cloud in Compressible Gas Pipelines.

An accurate estimation of pipe attributes, pose of pipeline inspection gauge (PIG), and downstream pipeline topology is essential for successful in-line inspection (ILI) of underground compressible gas pipelines. Taking a 3D point cloud of light detection and ranging (LiDAR) or time-of-flight (ToF) camera as the input, in this paper, we present the simultaneous pipe-attribute and PIG-pose estimation (SPPE) approach that estimates the optimal pipe-attribute and PIG-pose parameters to transform a 3D point cloud onto the inner pipe wall surface: major- and minor-axis lengths, roll, pitch, and yaw angles, and 2D deviation from the center of the pipe. Since the 3D point cloud has all spatial information of the inner pipe wall measurements, this estimation problem can be modeled by an optimal transformation matrix estimation problem from a PIG sensor frame to the global pipe frame. The basic idea of our SPPE approach is to decompose this transformation into two sub-transformations: The first transformation is formulated as a non-linear optimization problem whose solution is iteratively updated by the Levenberg-Marquardt algorithm (LMA). The second transformation utilizes the gravity vector to calculate the ovality angle between the geometric and navigation pipe frames. The extensive simulation results from our PIG simulator based on the robot operating system (ROS) platform demonstrate that the proposed SPPE can estimate the pipe attributes and PIG pose with excellent accuracy and is also applicable to real-time and post-processing non-destructive testing (NDT) applications thanks to its high computational efficiency.

An Area Partitioning and Subgraph Growing (APSG) Approach to the Conflation of Road Networks.

A road network represents a set of road objects in a geographic area and their interconnections, and it is an essential component of intelligent transportation systems (ITS) enabling emerging new applications such as dynamic route guidance, driving assistance systems, and autonomous driving. As the digitization of geospatial information becomes prevalent, a number of road networks with a wide variety of characteristics may coexist. In this paper, we present an area partitioning and subgraph growing (APSG) approach to the conflation of two road networks with a large difference in the level of details and representation rules. Our area partitioning (AP) scheme partitions the geographic area using the Network Voronoi Area Diagram (NVAD) of the low-detailed road network. Next, a subgraph of the high-detailed road network corresponding to a complex intersection is extracted and aggregated into a supernode so that high precision can be achieved via 1:1 road object matching. For the unmatched road objects due to missing road objects and different representation rules, we also propose a subgraph growing (SG) scheme that sequentially inserts a new road object while keeping the consistency of its connectivity to the matched road objects by the AP scheme. From the numerical results at Yeouido, Seoul, Korea, we show that our APSG scheme can achieve an outstanding matching performance in terms of the precision, recall, and F1-score.

Spatiotemporal Local-Remote Senor Fusion (ST-LRSF) for Cooperative Vehicle Positioning.

Vehicle positioning plays an important role in the design of protocols, algorithms, and applications in the intelligent transport systems. In this paper, we present a new framework of spatiotemporal local-remote sensor fusion (ST-LRSF) that cooperatively improves the accuracy of absolute vehicle positioning based on two state estimates of a vehicle in the vicinity: a local sensing estimate, measured by the on-board exteroceptive sensors, and a remote sensing estimate, received from neighbor vehicles via vehicle-to-everything communications. Given both estimates of vehicle state, the ST-LRSF scheme identifies the set of vehicles in the vicinity, determines the reference vehicle state, proposes a spatiotemporal dissimilarity metric between two reference vehicle states, and presents a greedy algorithm to compute a minimal weighted matching (MWM) between them. Given the outcome of MWM, the theoretical position uncertainty of the proposed refinement algorithm is proven to be inversely proportional to the square root of matching size. To further reduce the positioning uncertainty, we also develop an extended Kalman filter model with the refined position of ST-LRSF as one of the measurement inputs. The numerical results demonstrate that the proposed ST-LRSF framework can achieve high positioning accuracy for many different scenarios of cooperative vehicle positioning.

Wavelength-division-multiplexing optical packet switch with heterogeneous wavelength conversion capabilities.

Among the wavelength-division-multiplexing (WDM) optical packet switches (OPSs) using wavelength converters (WCs), a shared-per-node switch architecture has been considered as a way to utilize WCs efficiently. We propose a new switch control algorithm for the architecture. The proposed algorithm, different from previous algorithms, focuses on using the heterogeneous WC blocks (HeWCBs), where a HeWCB consists of WCs with different wavelength conversion degrees (WCDs). The results show that the WDM OPS architecture using HeWCBs reduces the number of WCs with a higher WCD, while minimizing the packet loss from wavelength contention at outbound links.