Motion capture data is crucial but creating a large dataset can be challenging due to complexities in acquisition. Generative Adversarial Network (GAN)-based motion data augmentation offers a potential solution to thi...
Motion capture data is crucial but creating a large dataset can be challenging due to complexities in acquisition. Generative Adversarial Network (GAN)-based motion data augmentation offers a potential solution to this issue. However, GANs often struggle with learning from limiteddata, resulting in poor quality output. In this study, we propose a dynamic Time Warping (dTW) filtering method that filters out generateddata significantly deviating from real-world examples. Through this approach, we have achieved an improvement in the fidelity of the generateddata, even with dataset size constraints, as evidenced by an increase in action recognition accuracy.
This study proposes an optimization method for sensor placement in lower-limb exoskeleton robots, aiming to improve the accuracy of the Center of Pressure (CoP) estimation during various human activities. Utilizing th...
This study proposes an optimization method for sensor placement in lower-limb exoskeleton robots, aiming to improve the accuracy of the Center of Pressure (CoP) estimation during various human activities. Utilizing the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm, the proposed method identifies an optimal sensor count, as the root Mean Square Error (rMSE) decelerates when the sensor count exceeds 6. Application of this methodresulted in a reduction of the Mean Absolute Error (MAE) to 4.13 mm and 8.92 mm on the mediolateral and anteroposterior axes respectively, a 22.8% improvement in CoP estimation accuracy compared to traditional anatomical methods. Further analysis revealed that weight parameters influence the CoP estimation accuracy, suggesting an enhancement in sensor placement efficiency through the adjustment of individual objectives' significance. The proposed sensor placement optimization method is expected to significantly enhance the performance of lower-limb exoskeleton robots and increase user satisfaction. Moreover, it could substantially contribute to the enhancement of human-computer interaction in robot control by providing a more accurate reflection of the user's intentions. These findings highlight the importance of continuing research in the field of lower-limb exoskeleton robots.
Autonomous navigation technology forunmanned Surface Vehicles (USVs) has seen many recent advancements. In order for a USV to operate autonomously, the path to move from its current location to its destination must b...
详细信息
Crowd simulation technologies andsystems can show and tell a lot of insights on massive crowds' movement behavior. Benefiting from such characteristics, they have found themselves useful in many application field...
详细信息
Crowd simulation technologies andsystems can show and tell a lot of insights on massive crowds' movement behavior. Benefiting from such characteristics, they have found themselves useful in many application fields. On the other hand, a typical educational institution, such as university, has a large body of student population, whose course schedules dominate theirdaily movement patterns. How to arrange courses according to the school resource of building vacancy androad availab.lity is thus an important problem to be solved by administrators. Traditional ways of arranging course tables are solely based on the consideration of faculty and students schedule availab.lity, often causing road congestions on campus. Furthermore, such arrangements may not be optimized for emergency situations, under which student crowds need to be evacuated. Enlightened by these observations, we designed and implemented a system dubbed as daddy Planner (data driven dynamic Planner) which allows school administrators to compare multiple planning solutions of course table visually. Our system makes use of crow simulation and shows advantage in usability as well as efficiency.
暂无评论