This paper describes a multi-agent mobile system that walks. In particular, the gate of the system can be considered as an expansion of the ordinary wave gate, since the class of system configuration is not restricted...This paper describes a multi-agent mobile system that walks. In particular, the gate of the system can be considered as an expansion of the ordinary wave gate, since the class of system configuration is not restricted in a line shape. The system consists of a number of identical units with distributed controllers. The homogeneous units are mechanically connected to construct a mobile platform. Every unit has its local controller that communicates only with its adjacent units. This basic configuration of supervisor-less structure affirmatively confines the dependence of each unit to a local area, and therefore any unit can be removed from/add into any part of a system regardless of the timing without disturbing the performance of the whole system. This flexibility of configuration significantly contributes to easy maintenance of units, such as battery charging or hot-replacing for faulty units. Utilizing the flexibility as well, the system is capable of adapting to a variety of tasks including transportation application and to target objects having various kinds of shape and/or a wide range of mass. A proposed example unit contains a Gough-Stewart Platform, a symmetrical type of parallel link manipulator, as its leg. The whole mobile system is aimed at transportation platform, with high system flexibility, i.e., the system is able to adapt to wide range of target objects. The "digital actuation (D-actuation)" concept is applied to the local unit controller. D-actuation is a concept to drive a mechatronic system with numbers of "digital actuator (D-actuator)" that has only discrete stable states, such as pneumatic cylinders or solenoids. D-actuation yields great benefits: high repeatability, system simplicity, and low cost. Because of the simplicity of the communication data, the control strategy, and the concept of D-actuation, the controlling framework can be implemented as distributed and localized one on every unit. The Schlafli symbol is applied to denote the system configurations. For example, two-dimensional honeycomb like connection of the units is denoted as {3,6}. A simple, but effective, coordinate system, HC/P (HoneyComb by Projection), is introduced to denote the connecting relations among the units in the {3,6} system. In short, HC/P utilizes three-axes to describe 2D system, and the redundant triplet notation enables direct and clear computation regarding unit coordinates. The basic architecture of the unit mechanism and gait controller are justified with simulation results. The performed simulation shows the feasibility of the whole mobile system.展开更多
毫米波雷达具有分辨率高、抗干扰能力强和对人体隐私侵犯少等优点,在身份识别领域中具有较好的应用前景.其中,基于毫米波雷达点云的步态识别已成为热门的研究方向之一.但这类方法大多基于点模型进行全局信息处理,对局部信息感知不足,从...毫米波雷达具有分辨率高、抗干扰能力强和对人体隐私侵犯少等优点,在身份识别领域中具有较好的应用前景.其中,基于毫米波雷达点云的步态识别已成为热门的研究方向之一.但这类方法大多基于点模型进行全局信息处理,对局部信息感知不足,从而导致算法的准确性不够.针对上述问题,该文提出了一种基于点体素交叉注意力机制的步态识别方法(gait recognition based on Point-Voxel fusion and Cross-attention,gaitPVC).该方法对数据采用了多帧融合的处理,利用双分支网络分别从点数据和体素数据协作提取并融合全局与局部特征,然后利用时序网络提取时序特征,以更好地提取人体步态信息.仿真结果表明,该文方法具有较好的鲁棒性和准确率.展开更多
文摘This paper describes a multi-agent mobile system that walks. In particular, the gate of the system can be considered as an expansion of the ordinary wave gate, since the class of system configuration is not restricted in a line shape. The system consists of a number of identical units with distributed controllers. The homogeneous units are mechanically connected to construct a mobile platform. Every unit has its local controller that communicates only with its adjacent units. This basic configuration of supervisor-less structure affirmatively confines the dependence of each unit to a local area, and therefore any unit can be removed from/add into any part of a system regardless of the timing without disturbing the performance of the whole system. This flexibility of configuration significantly contributes to easy maintenance of units, such as battery charging or hot-replacing for faulty units. Utilizing the flexibility as well, the system is capable of adapting to a variety of tasks including transportation application and to target objects having various kinds of shape and/or a wide range of mass. A proposed example unit contains a Gough-Stewart Platform, a symmetrical type of parallel link manipulator, as its leg. The whole mobile system is aimed at transportation platform, with high system flexibility, i.e., the system is able to adapt to wide range of target objects. The "digital actuation (D-actuation)" concept is applied to the local unit controller. D-actuation is a concept to drive a mechatronic system with numbers of "digital actuator (D-actuator)" that has only discrete stable states, such as pneumatic cylinders or solenoids. D-actuation yields great benefits: high repeatability, system simplicity, and low cost. Because of the simplicity of the communication data, the control strategy, and the concept of D-actuation, the controlling framework can be implemented as distributed and localized one on every unit. The Schlafli symbol is applied to denote the system configurations. For example, two-dimensional honeycomb like connection of the units is denoted as {3,6}. A simple, but effective, coordinate system, HC/P (HoneyComb by Projection), is introduced to denote the connecting relations among the units in the {3,6} system. In short, HC/P utilizes three-axes to describe 2D system, and the redundant triplet notation enables direct and clear computation regarding unit coordinates. The basic architecture of the unit mechanism and gait controller are justified with simulation results. The performed simulation shows the feasibility of the whole mobile system.
文摘毫米波雷达具有分辨率高、抗干扰能力强和对人体隐私侵犯少等优点,在身份识别领域中具有较好的应用前景.其中,基于毫米波雷达点云的步态识别已成为热门的研究方向之一.但这类方法大多基于点模型进行全局信息处理,对局部信息感知不足,从而导致算法的准确性不够.针对上述问题,该文提出了一种基于点体素交叉注意力机制的步态识别方法(gait recognition based on Point-Voxel fusion and Cross-attention,gaitPVC).该方法对数据采用了多帧融合的处理,利用双分支网络分别从点数据和体素数据协作提取并融合全局与局部特征,然后利用时序网络提取时序特征,以更好地提取人体步态信息.仿真结果表明,该文方法具有较好的鲁棒性和准确率.
