The automatic definition of the ground from 3D point clouds has been a common process for the last two decades,with many different approaches and applications that can be found in a vast literature.This paper presents...The automatic definition of the ground from 3D point clouds has been a common process for the last two decades,with many different approaches and applications that can be found in a vast literature.This paper presents a comparison of three different methodological concepts for ground classification,in order to establish the advantages and drawbacks of each method.First,a heuristic method,based on previous knowledge of the geometry and context of the 3D data.Secondly,a Deep Convolutional Network based on SegNet that classifies 2D images generated from the 3D point cloud.Finally,the third method applies a Deep Learning classification based on PointNet,which takes 3D points directly as inputs.To validate each method and compare them,public and labelled point clouds from the Actueel Hoogtebestand Nederland dataset are employed.Furthermore,the three methods are validated against the ISPRS 3D Semantic Labeling Contest benchmark.The results obtained show that the deep learning-based approaches outperform the heuristic method,with F-scores above 96%.The best results were obtained using a shallower version of SegNet,with F-score above 97%.展开更多
Adaptive locomotion in different types of surfaces is of critical importance for legged robots.The knowledge of various ground substrates,especially some geological properties,plays an essential role in ensuring the l...Adaptive locomotion in different types of surfaces is of critical importance for legged robots.The knowledge of various ground substrates,especially some geological properties,plays an essential role in ensuring the legged robots'safety.In this paper,the interaction between the robots and the environments is investigated through interaction dynamics with the closed-loop system model,the compliant contact model,and the friction model,which unveil the influence of environment's geological characteristics for legged robots'locomotion.The proposed method to classify substrates is based on the interaction dynamics and the sensory-motor coordination.The foot contact forces,joint position errors,and joint motor currents,which reflect body dynamics,are measured as the sensing variables.We train and classify the features extracted from the raw data with a multilevel weighted k-Nearest Neighbor(kNN) algorithm.According to the interaction dynamics,the strategy of adaptive walking is developed by adjusting the touchdown angles and foot trajectories while lifting up and dropping down the foot.Experiments are conducted on five different substrates with quadruped robot FROG-I.The comparison with other classification methods and adaptive walking between different substrates demonstrate the effectiveness of our approach.展开更多
基金the Spanish Ministry of Economy and Competitiveness through the Human Resources program FPI[grant number BES-2014-067736]Xunta de Galicia through grant number ED431C2016-038This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No.769255.
文摘The automatic definition of the ground from 3D point clouds has been a common process for the last two decades,with many different approaches and applications that can be found in a vast literature.This paper presents a comparison of three different methodological concepts for ground classification,in order to establish the advantages and drawbacks of each method.First,a heuristic method,based on previous knowledge of the geometry and context of the 3D data.Secondly,a Deep Convolutional Network based on SegNet that classifies 2D images generated from the 3D point cloud.Finally,the third method applies a Deep Learning classification based on PointNet,which takes 3D points directly as inputs.To validate each method and compare them,public and labelled point clouds from the Actueel Hoogtebestand Nederland dataset are employed.Furthermore,the three methods are validated against the ISPRS 3D Semantic Labeling Contest benchmark.The results obtained show that the deep learning-based approaches outperform the heuristic method,with F-scores above 96%.The best results were obtained using a shallower version of SegNet,with F-score above 97%.
文摘Adaptive locomotion in different types of surfaces is of critical importance for legged robots.The knowledge of various ground substrates,especially some geological properties,plays an essential role in ensuring the legged robots'safety.In this paper,the interaction between the robots and the environments is investigated through interaction dynamics with the closed-loop system model,the compliant contact model,and the friction model,which unveil the influence of environment's geological characteristics for legged robots'locomotion.The proposed method to classify substrates is based on the interaction dynamics and the sensory-motor coordination.The foot contact forces,joint position errors,and joint motor currents,which reflect body dynamics,are measured as the sensing variables.We train and classify the features extracted from the raw data with a multilevel weighted k-Nearest Neighbor(kNN) algorithm.According to the interaction dynamics,the strategy of adaptive walking is developed by adjusting the touchdown angles and foot trajectories while lifting up and dropping down the foot.Experiments are conducted on five different substrates with quadruped robot FROG-I.The comparison with other classification methods and adaptive walking between different substrates demonstrate the effectiveness of our approach.
