International Journal of Semantic Computing

Generating good and human-like locomotion or other legged motions for bipedal robots has always been challenging. One of the emerging solutions to this challenge is to use imitation learning. The sources for imitation are mostly state-only demonstrations, so using state-of-the-art Generative Adversarial Imitation Learning (GAIL) with Imitation from Observation (IfO) ability will be an ideal framework to use in solving this problem. However, it is often difficult to allow new or complicated movements as the common sources for these frameworks are either expensive to set up or hard to produce satisfactory results without computationally expensive preprocessing, due to accuracy problems. Inspired by how people learn advanced knowledge after acquiring a basic understanding of specific subjects, this paper proposes a Motion capture-aided Video Imitation (MoVI) learning framework based on Adversarial Motion Priors (AMP) by combining motion capture data of primary actions like walking with video clips of target motion like running, aiming to create smooth and natural imitation results of the target motion. This framework is able to produce various human-like locomotion by taking the most common and abundant motion capture data with any video clips of motion without the need for expensive datasets or sophisticated preprocessing.

This paper presents and analyzes different parallelization strategies in VVenC, an open and optimized software encoder implementation of the Versatile Video Coding (VVC) standard. VVC has been developed to address the increasing demand for higher compression of digital video data, and it reduces the bitrate by around 50% for the same perceived quality compared to its predecessor, the High-Efficiency Video Coding (HEVC) standard. However, this increase in compression efficiency comes with an increase in computational complexity, particularly on the encoder side. VVenC integrates algorithmic optimizations for each coding tool in VVC and defines a set of five presets from faster to slower that provide Pareto-optimal tradeoffs between runtime and efficiency. Multithreading is employed to further reduce runtime while preserving most of the compression efficiency of each preset. With a hierarchical combination of pre-processing, picture-level and in-picture parallelization, VVenC achieves a 4× speedup for four threads. Further speedup using a higher number of threads depends on the video resolution and used encoder preset. For 16 threads, it ranges from 6 to 9 for high definition to 10-12 for ultra-high-definition video. Compared to previous work, the usage of temporal prediction for the adaptive loop filter reduces the associated coding efficiency loss from 0.4% to almost zero. A better scaling for higher numbers of threads can be achieved at the cost of a higher coding efficiency loss. In the presented framework, increased speedup by smaller block Coding Tree Unit (CTU) sizes, a combination of wavefront parallel processing and various tiles picture partitioning configurations is examined. Furthermore, results on a 20-core ARM-based Apple M1 computer indicate a better scaling for multithreading compared to × 86-based architectures. The analysis is complemented by profiling, which exhibits the overhead by idle threads and identifies the mode estimation as the main bottleneck of the presented framework.

In this paper, an analysis of the precision of range sensors is presented. Light Detection and Ranging (LIDAR) data from three different sensors (HLS-LFCD-LDS, SICK TIM561, and Kinect V2), stereo data from the Realsense D435, and structured light data from both Kinect V1 for Xbox and the Xtion PRO Live were analyzed regarding the influence of range, incident angle to the surface, and surface material. In addition, a comparison with standard deviation models based on vendor-provided specifications was performed. We found that the vendor specifications are in general not specific enough to facilitate accurate sensor simulations. Therefore, we developed a data-driven model for range precision. Our model can be used to create realistic sensor simulations and to develop robot navigation algorithms weighing sensor range readings based on the precision.

Incorporating semantics in modern projects can be time-consuming, rely on limited re-sources, and require expert knowledge to implement in a future-proof manner. Across many working groups and industrial fields, semantics continues to help solve data consolidation challenges; however, there are still emerging situations such as where modeling and ontology engineering efforts overlap, ongoing siloed data, or even continuing to create tech-debt for the future. Tools have been developed and researched that offer users assistance with the automation of select expert tasks to help streamline semantic artifact creation, understanding of design decisions, and domain data alignments. The Semantic Knowledge Graph Generator and Recommendation Framework encourage semantic projects that are automated from the start to follow industry standards, enhanced with recommendations, and use components which make semantic artifacts more accessible for non-semantic experts. Based on that tool, this extension article presents several industry use cases and related discussion where semantic recommendations and deployment automation have assisted with project development, teaching about ontology structures, and gaining accessibility of semantic artifacts.

This paper presents an open-source MATLAB/Simulink physics simulator for rigid-body articulated systems, including manipulators and floating-base robots. Thanks to MATLAB/Simulink features like MATLAB system classes and Simulink Function blocks, the presented simulator combines a programmatic and block-based approach, resulting in a flexible design in the sense that different parts, including its physics engine, robot-ground interaction model, and state evolution algorithm are simply accessible and editable. Moreover, through the use of Simulink dynamic mask blocks, the proposed simulator supports robot models integrating open-chain and closed-chain kinematics with any desired number of links interacting with the ground. This simulator can also integrate second-order actuator dynamics. Furthermore, the simulator benefits from a one-line installation and an easy-to-use Simulink interface.

The popularity of Unmanned Aerial Vehicles (UAVs), aka drones, has increased rapidly in recent years. UAVs are becoming easily accessible to more users. Malicious intentions can erode public safety when least expected. Current methods used for UAV detection systems include computer vision, radar, radio frequency and audio approaches. We choose the audio method for its high accuracy, low computational requirement and low cost. However, the lack of publicly available datasets is one of the main bottlenecks for developing an audio-based UAV detection and classification system. To fill this gap, we select 15 different UAVs, ranging from toy hand drones to Class I drones and record a total of 8120 s length of audio data generated from the flying UAVs. To the best of our knowledge, the proposed dataset is the largest audio dataset for UAVs so far. We further implement a Convolutional Neural Network (CNN) model for 15-class UAV classification and trained the model with the collected data. The average test accuracy of the trained model is 98.7% and the test loss is 0.076.

Energy efficiency of surface curvature determination in range images, a core component of a variety of robotics/multimedia tasks, is studied on a popular computing architecture (Intel x86-64). Eight methods for that type of curvature determination are part of the study reported here. Their base energy usage and their reduced energy usage when energy consumption-conscious strategies are applied to them are considered within the study. Comparative evaluation of energy consumption is included in the study. The focus here on core computing component power reduction aims at enabling robotics/multimedia task computation to exhibit reduced energy usage. Some general notes on methods for potentially improving energy efficiency are also provided.

Landing is the most challenging and risky aspect of basic multirotor drone flight, and only simple landing methods exist for autonomous drones. We explore methods for autonomous drone landing in two scenarios. In the first scenario, we examine methods for landing on known landing pads using fiducial markers and a gimbal-mounted monocular camera. This method has potential in drone applications where a drone must land more accurately than global positioning system (GPS) can provide (e.g. package delivery in an urban canyon). We expand on previous methods by actuating the drone’s camera to track the marker over time, and we address the complexities of pose estimation caused by fiducial marker orientation ambiguity. In the second scenario, and in collaboration with the Rover-Aerial Vehicle Exploration Network (RAVEN) project, we explore methods for landing on solidified lava flows in Iceland, which serves as an analog environment for Mars and provides insight into the effectiveness of drone-rover exploration teams. Our drone uses a depth camera to visualize the terrain, and we are developing methods to analyze the terrain data for viable landing sites in real time with minimal sensors and external infrastructure requirements, so that the solution does not heavily influence the drone’s behavior, mission structure, or operational environments.