CN108897315A - A kind of Multi-agent Team Formation - Google Patents
A kind of Multi-agent Team Formation Download PDFInfo
- Publication number
- CN108897315A CN108897315A CN201810600769.5A CN201810600769A CN108897315A CN 108897315 A CN108897315 A CN 108897315A CN 201810600769 A CN201810600769 A CN 201810600769A CN 108897315 A CN108897315 A CN 108897315A
- Authority
- CN
- China
- Prior art keywords
- channel
- robot
- neuron
- nucleus
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 230000015572 biosynthetic process Effects 0.000 title claims abstract description 48
- 238000001514 detection method Methods 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 34
- 210000004227 basal ganglia Anatomy 0.000 claims abstract description 20
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 210000002569 neuron Anatomy 0.000 claims description 57
- 230000006399 behavior Effects 0.000 claims description 39
- 210000001577 neostriatum Anatomy 0.000 claims description 32
- 210000001905 globus pallidus Anatomy 0.000 claims description 30
- 210000004281 subthalamic nucleus Anatomy 0.000 claims description 20
- 101000894525 Homo sapiens Transforming growth factor-beta-induced protein ig-h3 Proteins 0.000 claims description 12
- 102100021398 Transforming growth factor-beta-induced protein ig-h3 Human genes 0.000 claims description 12
- 210000003710 cerebral cortex Anatomy 0.000 claims description 10
- VYFYYTLLBUKUHU-UHFFFAOYSA-N dopamine Chemical compound NCCC1=CC=C(O)C(O)=C1 VYFYYTLLBUKUHU-UHFFFAOYSA-N 0.000 claims description 10
- 238000013178 mathematical model Methods 0.000 claims description 9
- 230000002401 inhibitory effect Effects 0.000 claims description 6
- 229960003638 dopamine Drugs 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 2
- 210000000495 subthalamus Anatomy 0.000 claims description 2
- 230000010354 integration Effects 0.000 claims 1
- 230000000542 thalamic effect Effects 0.000 claims 1
- 238000005755 formation reaction Methods 0.000 abstract 3
- 230000006870 function Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 4
- 238000005381 potential energy Methods 0.000 description 4
- 230000003542 behavioural effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004936 stimulating effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Manipulator (AREA)
Abstract
Description
技术领域technical field
本发明涉及机器人的编队领域，具体涉及一种具有行为选择功能的多机器人编队方法。The invention relates to the field of robot formation, in particular to a multi-robot formation method with a behavior selection function.
背景技术Background technique
机器人代替人完成在复杂、危险的环境下的重复性工作，加速了社会的发展。近年来社会市场对机器人的功能的提出了新的需求，机器人需要完成更为复杂的检测任务甚至作业任务，但是面对复杂的，高效率的、并行完成的任务时，单一的机器人无法胜任，需要多机器人协同。Robots replace humans to complete repetitive tasks in complex and dangerous environments, accelerating the development of society. In recent years, the social market has put forward new requirements for the functions of robots. Robots need to complete more complex detection tasks and even homework tasks. However, when faced with complex, efficient, and parallel tasks, a single robot is not competent. Multi-robot collaboration is required.
机器人编队控制是一种多个机器人在达到目标队形的过程中，既能组成目标队形，又可以适应具体的环境约束的一种控制方法。现有的常用的方法主要有基于领航—跟随者方法、虚拟结构法、基于行为的方法、基于图论、势能法的方法。Robot formation control is a control method in which multiple robots can form a target formation and adapt to specific environmental constraints in the process of reaching the target formation. The existing commonly used methods are mainly based on the leader-follower method, the virtual structure method, the behavior-based method, the graph theory-based method, and the potential energy method.
其中领航者—跟随者方法的基本思想是在多机器人组成的编队系统中，可以存在一个或者多个领航机器人，其他非领航者机器人即为跟随机器人，跟随机器人以相对其领航机器人的位置相对距离、相对角度中作为输入控制量，使得跟随机器人与领航机器人的相对位置无限逼近目标值。领航跟随方法的控制结构比较简单，但是由于领航机器人和跟随机器人之间没有位置反馈，并且是领航者机器人单点控制，导致容易出现机器人掉队等情况，系统的鲁棒性较差。The basic idea of the leader-follower method is that in a multi-robot formation system, there can be one or more leader robots, and other non-leader robots are follower robots, and the relative distance between the follower robot and its leader robot position is , The relative angle is used as the input control quantity, so that the relative position of the follower robot and the leader robot approaches the target value infinitely. The control structure of the pilot-following method is relatively simple, but since there is no position feedback between the pilot robot and the following robot, and the single-point control of the pilot robot, it is easy for the robot to fall behind, and the robustness of the system is poor.
虚拟结构法的基本思想是将多机器人组成的系统看成一个假想的刚体结构，各个机器人在参考坐标系下的坐标不变,即机器人之间的相对位置不变。The basic idea of the virtual structure method is to regard the system composed of multiple robots as a hypothetical rigid body structure, and the coordinates of each robot in the reference coordinate system remain unchanged, that is, the relative positions between the robots remain unchanged.
基于行为的方法的基本思想是将多机器人编队的各个环节看成由单个机器人的多个基本行为构成，只需要研究各个基本行为的控制方法即可通过基本行为的组合控制多机器人形成编队。基本行为一般包括目标跟踪、避免障碍、避免碰撞、编队生成和编队保持等。基于行为的方法由各个机器人相互感知控制，系统比较容易实现分布式控制，具有很好的鲁棒性。但是由于无法结合准确的数学模型分析，导致不能十分有效的保证系统的可靠性。The basic idea of the behavior-based method is to regard each link of the multi-robot formation as composed of multiple basic behaviors of a single robot, and only need to study the control method of each basic behavior to control the formation of multi-robots through the combination of basic behaviors. Basic behaviors generally include target tracking, obstacle avoidance, collision avoidance, formation generation and formation maintenance, etc. The behavior-based method is controlled by mutual perception of each robot. The system is relatively easy to realize distributed control and has good robustness. However, due to the inability to combine accurate mathematical model analysis, the reliability of the system cannot be guaranteed very effectively.
势能法势场法是与上而所述传统方法不巧的另一种队形形成控制方法，它将队形形成的所有步骤合为一起，通过构建合近的势函数，由势场函数来决定机器人的控制律将其他机器人对其不同的势力作用，结合期望的队形图形，使机器人朝预期的队形图运动，当各自到达编队图时，整个系统的势能玻小。在队形形成过程中，队形目标点是动态变化的，属于动态队形形成方法，机器人是通过不断调整相互之间的距离而达到一个平衡态；但队形形成的快速性和可控性无法把握，由于是按照相对距离的编队图来形成队形，那么说明机器人之问的相互相对位置其实是事先确定的，这就使机器人的目标位置可能不是最佳的，又由于整个队形目标上保持一个相对位置固定的编队图，假设一个机器人未能达到指定队形点，其他机器人将一直处于动态搜寻最小势能点的状态并一直运动。Potential energy method Potential field method is another method of formation formation control that is different from the traditional method mentioned above. It combines all the steps of formation formation. By constructing a close potential function, the potential field function can Determine the control law of the robot to act on different forces of other robots, combined with the expected formation figure, so that the robots move towards the expected formation figure. When each reaches the formation figure, the potential energy of the whole system will decrease. In the formation formation process, the formation target point is dynamically changed, which belongs to the dynamic formation formation method, and the robots reach an equilibrium state by constantly adjusting the distance between each other; but the rapidity and controllability of formation formation Unable to grasp, because the formation is formed according to the formation diagram of the relative distance, it means that the relative positions of the robots are actually determined in advance, which makes the target position of the robot may not be optimal, and because the entire formation target Maintain a relatively fixed formation diagram, assuming that one robot fails to reach the designated formation point, other robots will always be in the state of dynamically searching for the minimum potential energy point and keep moving.
针对上述问题，提出一种基于行为选择的多机器人编队方法。Aiming at the above problems, a multi-robot formation method based on behavior selection is proposed.
发明内容Contents of the invention
本发明的目的在于解决的多机器人编队过程中系统的可靠性与鲁棒性无法保证的问题。The purpose of the present invention is to solve the problem that the reliability and robustness of the system cannot be guaranteed in the multi-robot formation process.
为了解决上述技术问题，本发明提供一种基于行为选择的多机器人编队方法，通过引入一种基底神经节行为选择编队策略，增强系统的可靠性与鲁棒性。In order to solve the above technical problems, the present invention provides a multi-robot formation method based on behavior selection, which enhances the reliability and robustness of the system by introducing a basal ganglia behavior selection formation strategy.
具体技术方案为：The specific technical solutions are:
综合考虑机器人向形成编队目标位置的运动的路程之和dissum，机器人的适宜移动速度vi，机器人调整到面向目标点姿态的时间Tadjust，机器人运动到目标点的时间Tmove，机器人在下个采样周期后到达位置的检测收益，进行行为选择。Comprehensively consider the sum dis sum of the movement distance of the robot to the formation target position, the suitable moving speed v i of the robot, the time T adjust for the robot to adjust to the attitude towards the target point, and the time T move for the robot to move to the target point. The detection income of the arrival position after the sampling period is used for behavior selection.
Tcost＝Tmove+Tadjust T cost = T move + T adjust
步骤一、设定与位置相关的检测收益计算方法。Step 1, setting a detection revenue calculation method related to the location.
(1)目标探测位置收益区设定：目标探测位置收益区设定具体是如下的条件：收益区为以目标探测中心为圆形的数块扇形区域。(1) Setting of profit area of target detection position: The setting of profit area of target detection position is specifically as follows: the profit area is several fan-shaped areas with the center of target detection as a circle.
(2)检测收益计算方法：在距离收益区较近的扇形内，设定该机器人的该检测点收益为与目标点重要程度相关的固定收益；在距离收益区较远的圆环内，设定该机器人的该检测点收益为与机器人与目标点之间距离成反比的动态收益。并在检测完成后，当前作业周期的检测收益将清零。(2) Calculation method of detection income: in the sector closer to the income area, set the detection point income of the robot as a fixed income related to the importance of the target point; in the circle farther from the income area, set The detection point income of the robot is determined to be a dynamic income inversely proportional to the distance between the robot and the target point. And after the detection is completed, the detection income of the current job cycle will be cleared.
步骤二、定义机器人的具体行为：根据机器人的具体结构与功能定义机器人在行为选择完成后可能会出现的具体行为。对各个行为的激励抑制关系进行限制与设定分析，若某一机器人行为可被激发，则采用基底神经节通道模型进行验证；若被激发的机器人行为可以通过验证，则执行此行为。Step 2. Define the specific behavior of the robot: according to the specific structure and function of the robot, define the specific behavior that the robot may appear after the behavior selection is completed. The incentive-inhibition relationship of each behavior is restricted and analyzed. If a certain robot behavior can be stimulated, the basal ganglia channel model is used for verification; if the stimulated robot behavior can pass the verification, this behavior is executed.
步骤三、确定基底神经节的通道数量，建立基底神经节通道模型，初始化相关参数。Step 3: Determine the number of channels of the basal ganglia, establish a channel model of the basal ganglia, and initialize relevant parameters.
基底神经节数学模型包括纹状体、苍白球外核、底丘脑核、核团苍白球内核。The mathematical model of the basal ganglia includes the striatum, the outer nucleus of the globus pallidus, the subthalamic nucleus, and the inner nucleus of the globus pallidus.
(1)建立行为通道的数学模型：每个通道用一个漏积分神经元表示：(1) Establish a mathematical model of behavioral channels: each channel is represented by a leaky integral neuron:
其中，x是神经元的状态，u为神经元的输入，y为神经元的输出，H为阶跃函数，其余为模型参数。Among them, x is the state of the neuron, u is the input of the neuron, y is the output of the neuron, H is the step function, and the rest are model parameters.
yi C＝Si y i C = S i
大脑皮层整合、加工得到结合步骤一中各个因素的综合重要性指标Si，其中i为通道号，yi C表征大脑皮层中第i个通道的输出。The cerebral cortex is integrated and processed to obtain a comprehensive importance index S i combining all factors in step 1, where i is the channel number, and y i C represents the output of the i-th channel in the cerebral cortex.
纹状体D1的状态可以描述为:The state of striatum D1 can be described as:
其中，ui SD1是纹状体D1，i通道的神经元输入，ai SD1是该通道神经元的状态，yi SD1是该通道神经元的输出，wCSD1是大脑皮层到纹状体D1的权重。1+λ描述的是多巴胺对纹状体D1的激励作用，εSD1为纹状体D1的输出阈值。Among them, u i SD1 is the striatum D1, the neuron input of the i channel, a i SD1 is the state of the channel neuron, y i SD1 is the output of the channel neuron, w CSD1 is the cerebral cortex to the striatum D1 the weight of. 1+λ describes the stimulating effect of dopamine on striatum D1, and ε SD1 is the output threshold of striatum D1.
纹状体D2的状态可以描述为:The state of striatum D2 can be described as:
其中，ui SD2是纹状体D2，i通道的神经元输入，ai SD2是该通道神经元的状态，yi SD2是该通道神经元的输出，wCSD2是大脑皮层到纹状体D2的权重,1-λ描述的是多巴胺对纹状体D2的抑制作用，εSD2为纹状体D2的输出阈值。Among them, u i SD2 is the striatum D2, the neuron input of the i channel, a i SD2 is the state of the channel neuron, y i SD2 is the output of the channel neuron, w CSD2 is the cerebral cortex to the striatum D2 The weights, 1-λ describe the inhibitory effect of dopamine on striatal D2, and ε SD2 is the output threshold of striatal D2.
苍白球外核可以描述为：The outer nucleus of the globus pallidum can be described as:
其中，ui GPe是苍白球外核，i通道的神经元输入，ai GPe是该通道神经元的状态，yi GPe是该通道神经元的输出，wSD2GPe是纹状体D2到苍白球外核的权重,εGPe是苍白球外核的输出阈值。Among them, u i GPe is the outer nucleus of the globus pallidus, the neuron input of the i channel, a i GPe is the state of the neuron of the channel, y i GPe is the output of the neuron of the channel, w SD2GPe is the striatum D2 to the globus pallidum The weight of the outer nucleus, εGPe is the output threshold of the outer nucleus of the globus pallidum.
底丘脑核的模型可描述为:The model of the subthalamic nucleus can be described as:
其中，ui STN是底丘脑核，i通道的神经元输入，ai STN是该通道神经元的状态，yi STN是该通道神经元的输出，wGPeSTN是苍白球外核到底丘脑核的权重,εSTN是底丘脑核的输出阈值。Among them, u i STN is the subthalamic nucleus, the neuron input of the i channel, a i STN is the state of the channel neuron, y i STN is the output of the channel neuron, w GPeSTN is the subthalamic nucleus of the external nucleus of the globus pallidum Weight, ε STN is the output threshold of the subthalamic nucleus.
根据解剖学上研究，底丘脑核对于苍白球内核的投射十分的分散，因此在描述苍白球内核的数学模型时，令苍白球内核的输入包括其他左右通道的底丘脑核出。According to anatomical studies, the projection of the subthalamic nucleus to the globus pallidus nucleus is very scattered. Therefore, when describing the mathematical model of the globus pallidus nucleus, the input of the globus pallidus nucleus includes the output of the subthalamic nucleus of other left and right channels.
描述为：described as:
其中，ui GPi是苍白球内核，i通道的神经元输入，ai GPi是该通道神经元的状态，yi GPi是该通道神经元的输出，wSD1GPi是纹状体D1到苍白球内核的权重,wSTNGPi为底丘脑核到苍白球内核的输出权重，εGPi是苍白球内核的输出阈值，需要注意的是，wCSD1、wCSD2、wSTNGPi为正值表征激励性连接、wSD1GPi、wSD2GPe、wGPeSTN为负值，表征抑制性连接。Among them, u i GPi is the globus pallidus kernel, the neuron input of the i channel, a i GPi is the state of the neuron of the channel, y i GPi is the output of the neuron of the channel, w SD1GPi is the striatum D1 to the globus pallidus kernel , w STNGPi is the output weight from the subthalamus nucleus to the globus pallidus kernel, ε GPi is the output threshold of the globus pallidus kernel, it should be noted that w CSD1 , w CSD2 , and w STNGPi are positive values representing the incentive connection, w SD1GPi , w SD2GPe , w GPeSTN are negative values, representing inhibitory connections.
步骤四：校正基底神经节的通道模型参数：当机器人的整体运行情况在以用户偏好为标准的条件下有所偏差时，进行对步骤三中模型各项参数的校正。Step 4: Correct the channel model parameters of the basal ganglia: When the overall operation of the robot deviates from the user preference as the standard, correct the parameters of the model in step 3.
本发明与现有技术相比，其显著优点在于：(1)引入了一种基于行为选择的动态机器人编队方法，有效解决了现有编队检测方法中鲁棒性不强，可靠性不强等缺点；(2)通过开放参数与各项权重修改通道，使得本发明具有更加广泛的通用性；(2)引入目标检测收益变量，可以在整体上提升多机器人进行检测任务的收益，间接提升了系统运行效率。Compared with the prior art, the present invention has significant advantages in that: (1) It introduces a dynamic robot formation method based on behavior selection, which effectively solves the problems of weak robustness and low reliability in the existing formation detection methods. Disadvantages; (2) The present invention has wider versatility by opening parameters and various weight modification channels; (2) Introducing target detection income variables can improve the income of multi-robots for detection tasks as a whole, indirectly improving System operating efficiency.
附图说明Description of drawings
图1为本发明机器人检测任务检测收益范围示意图。Fig. 1 is a schematic diagram of the detection profit range of the robot detection task in the present invention.
图2为本发明基底神经节示意图。Fig. 2 is a schematic diagram of the basal ganglia of the present invention.
具体实施方式Detailed ways
本发明基于行为选择的多机器人编队方法，包括以下步骤：The multi-robot formation method based on behavior selection of the present invention comprises the following steps:
步骤一、设定与位置相关的检测收益计算方法。Step 1, setting a detection revenue calculation method related to the location.
(1)目标探测位置收益区设定：目标探测位置收益区设定具体是如下的条件：收益区为以目标探测中心为圆形的数块扇形区域。(1) Setting of profit area of target detection position: The setting of profit area of target detection position is specifically as follows: the profit area is several fan-shaped areas with the center of target detection as a circle.
(2)检测收益计算方法：在距离收益区较近的扇形内，设定该机器人的该检测点收益为与目标点重要程度相关的固定收益；在距离收益区较远的圆环内，设定该机器人的该检测点收益为与机器人与目标点之间距离成反比的动态收益。并在检测完成后，当前作业周期的检测收益将清零。(2) Calculation method of detection income: in the sector closer to the income area, set the detection point income of the robot as a fixed income related to the importance of the target point; in the circle farther from the income area, set The detection point income of the robot is determined to be a dynamic income inversely proportional to the distance between the robot and the target point. And after the detection is completed, the detection income of the current job cycle will be cleared.
其中ηprofit为当前目标点的动态收益，ηk为当前目标点的固定收益，r为机器人距离目标检测点的距离，r1为固定检测收益区与目标检测点的最短距离，r2为动态收益区与目标检测点的最长距离。Among them, η profit is the dynamic profit of the current target point, η k is the fixed profit of the current target point, r is the distance between the robot and the target detection point, r 1 is the shortest distance between the fixed detection profit area and the target detection point, and r 2 is the dynamic The longest distance between the revenue zone and the target detection point.
步骤二、定义机器人的具体行为：根据机器人的具体结构与功能定义机器人在行为选择完成后可能会出现的具体行为。对各个行为的激励抑制关系进行限制与设定分析，若某一机器人行为可被激发，则采用基底神经节通道模型进行验证；若被激发的机器人行为可以通过验证，则执行此行为。设计机器人姿态行为有顺时针转动、顺时针转动、前进、后退、搜索、回程六种。Step 2. Define the specific behavior of the robot: according to the specific structure and function of the robot, define the specific behavior that the robot may appear after the behavior selection is completed. The incentive-inhibition relationship of each behavior is restricted and analyzed. If a certain robot behavior can be stimulated, the basal ganglia channel model is used for verification; if the stimulated robot behavior can pass the verification, this behavior is executed. There are six kinds of posture behaviors for the design robot: clockwise rotation, clockwise rotation, forward, backward, search, and return.
其中si表征各个行为的重要程度，ηi k为待整定常数。Among them, s i represents the importance of each behavior, and η i k is a constant to be adjusted.
步骤三、确定基底神经节的通道数量，建立基底神经节通道模型，初始化相关参数。Step 3: Determine the number of channels of the basal ganglia, establish a channel model of the basal ganglia, and initialize relevant parameters.
基底神经节数学模型包括纹状体、苍白球外核、底丘脑核、核团苍白球内核。The mathematical model of the basal ganglia includes the striatum, the outer nucleus of the globus pallidus, the subthalamic nucleus, and the inner nucleus of the globus pallidus.
(1)建立行为通道的数学模型：每个通道用一个漏积分神经元表示：(1) Establish a mathematical model of behavioral channels: each channel is represented by a leaky integral neuron:
其中，x是神经元的状态，u为神经元的输入，y为神经元的输出，H为阶跃函数，其余为模型参数。Among them, x is the state of the neuron, u is the input of the neuron, y is the output of the neuron, H is the step function, and the rest are model parameters.
yi C＝Si y i C = S i
大脑皮层整合、加工得到结合步骤一中各个因素的综合重要性指标Si，其中i为通道号，yi C表征大脑皮层中第i个通道的输出。The cerebral cortex is integrated and processed to obtain a comprehensive importance index S i combining all factors in step 1, where i is the channel number, and y i C represents the output of the i-th channel in the cerebral cortex.
纹状体D1的状态可以描述为:The state of striatum D1 can be described as:
其中，ui SD1是纹状体D1，i通道的神经元输入，ai SD1是该通道神经元的状态，yi SD1是该通道神经元的输出，wCSD1是大脑皮层到纹状体D1的权重。1+λ描述的是多巴胺对纹状体D1的激励作用，εSD1为纹状体D1的输出阈值。Among them, u i SD1 is the striatum D1, the neuron input of the i channel, a i SD1 is the state of the channel neuron, y i SD1 is the output of the channel neuron, w CSD1 is the cerebral cortex to the striatum D1 the weight of. 1+λ describes the stimulating effect of dopamine on striatum D1, and ε SD1 is the output threshold of striatum D1.
纹状体D2的状态可以描述为:The state of striatum D2 can be described as:
其中，ui SD2是纹状体D2，i通道的神经元输入，ai SD2是该通道神经元的状态，yi SD2是该通道神经元的输出，wCSD2是大脑皮层到纹状体D2的权重,1-λ描述的是多巴胺对纹状体D2的抑制作用，εSD2为纹状体D2的输出阈值。Among them, u i SD2 is the striatum D2, the neuron input of the i channel, a i SD2 is the state of the channel neuron, y i SD2 is the output of the channel neuron, w CSD2 is the cerebral cortex to the striatum D2 The weights, 1-λ describe the inhibitory effect of dopamine on striatal D2, and ε SD2 is the output threshold of striatal D2.
苍白球外核可以描述为：The outer nucleus of the globus pallidum can be described as:
其中，ui GPe是苍白球外核，i通道的神经元输入，ai GPe是该通道神经元的状态，yi GPe是该通道神经元的输出，wSD2GPe是纹状体D2到苍白球外核的权重,εGPe是苍白球外核的输出阈值。Among them, u i GPe is the outer nucleus of the globus pallidus, the neuron input of the i channel, a i GPe is the state of the neuron of the channel, y i GPe is the output of the neuron of the channel, w SD2GPe is the striatum D2 to the globus pallidum The weight of the outer nucleus, εGPe is the output threshold of the outer nucleus of the globus pallidus.
底丘脑核的模型可描述为:The model of the subthalamic nucleus can be described as:
其中，ui STN是底丘脑核，i通道的神经元输入，ai STN是该通道神经元的状态，yi STN是该通道神经元的输出，wGPeSTN是苍白球外核到底丘脑核的权重,εSTN是底丘脑核的输出阈值。Among them, u i STN is the subthalamic nucleus, the neuron input of the i channel, a i STN is the state of the channel neuron, y i STN is the output of the channel neuron, w GPeSTN is the subthalamic nucleus of the external nucleus of the globus pallidum Weight, ε STN is the output threshold of the subthalamic nucleus.
根据解剖学上研究，底丘脑核对于苍白球内核的投射十分的分散，因此在描述苍白球内核的数学模型时，令苍白球内核的输入包括其他左右通道的底丘脑核出。According to anatomical studies, the projection of the subthalamic nucleus to the globus pallidus nucleus is very scattered. Therefore, when describing the mathematical model of the globus pallidus nucleus, the input of the globus pallidus nucleus includes the output of the subthalamic nucleus of other left and right channels.
描述为：described as:
其中，ui GPi是苍白球内核，i通道的神经元输入，ai GPi是该通道神经元的状态，yi GPi是该通道神经元的输出，wSD1GPi是纹状体D1到苍白球内核的权重,wSTNGPi为底丘脑核到苍白球内核的输出权重，εGPi是苍白球内核的输出阈值，值需要注意的是，wCSD1、wCSD2、wSTNGPi为正值表征激励性连接、wSD1GPi、wSD2GPe、wGPeSTN为负值，表征抑制性连接。Among them, u i GPi is the globus pallidus kernel, the neuron input of the i channel, a i GPi is the state of the neuron of the channel, y i GPi is the output of the neuron of the channel, w SD1GPi is the striatum D1 to the globus pallidus kernel weight, w STNGPi is the output weight from the subthalamic nucleus to the globus pallidus core, εGPi is the output threshold of the globus pallidus core, the values need to be noted that w CSD1 , w CSD2 , and w STNGPi are positive values representing incentive connections, w SD1GPi , w SD2GPe , w GPeSTN are negative values, representing inhibitory connections.
步骤四：校正基底神经节的通道模型参数：当机器人的整体运行情况在以用户偏好为标准的条件下有所偏差时，进行对步骤三中模型各项参数的校正。Step 4: Correct the channel model parameters of the basal ganglia: When the overall operation of the robot deviates from the user preference as the standard, correct the parameters of the model in step 3.
最终按照苍白球内核输出对机器人行为进行选定。Finally, the behavior of the robot is selected according to the output of the globus pallidum kernel.
本发明也可在不同功能，不同数量规模，不同检测目标，不同运动方式与目标检测方式的机器人集群中应用。以上对本发明的具体实施例进行了描述。需要理解的是，本发明并不局限于上述特定实施方式，本领域技术人员可以在权利要求的范围内作出各种变形或修改，这并不影响本发明的实质内容。The present invention can also be applied in robot clusters with different functions, different numbers and scales, different detection targets, different motion modes and target detection modes. Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810600769.5A CN108897315A (en) | 2018-06-12 | 2018-06-12 | A kind of Multi-agent Team Formation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810600769.5A CN108897315A (en) | 2018-06-12 | 2018-06-12 | A kind of Multi-agent Team Formation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108897315A true CN108897315A (en) | 2018-11-27 |
Family
ID=64344681
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810600769.5A Withdrawn CN108897315A (en) | 2018-06-12 | 2018-06-12 | A kind of Multi-agent Team Formation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108897315A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119151A (en) * | 2019-05-24 | 2019-08-13 | 陕西理工大学 | A kind of target point distribution method of robot formation forming |
| CN111077887A (en) * | 2019-12-12 | 2020-04-28 | 南京理工大学 | A multi-robot comprehensive obstacle avoidance method using pilot following method |
| CN111082709A (en) * | 2019-12-18 | 2020-04-28 | 南京理工大学 | Electric anti-backlash control method based on basal ganglia |
| CN111207754A (en) * | 2020-02-28 | 2020-05-29 | 上海交通大学 | Multi-robot formation positioning method and robot equipment based on particle filter |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103941728A (en) * | 2014-04-24 | 2014-07-23 | 北京航空航天大学 | Formation changing method for intensive autonomous formations of unmanned aerial vehicle |
| CN105527960A (en) * | 2015-12-18 | 2016-04-27 | 燕山大学 | Mobile robot formation control method based on leader-follow |
| CN108646550A (en) * | 2018-04-03 | 2018-10-12 | 江苏江荣智能科技有限公司 | A kind of multiple agent formation method of Behavior-based control selection |
-
2018
- 2018-06-12 CN CN201810600769.5A patent/CN108897315A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103941728A (en) * | 2014-04-24 | 2014-07-23 | 北京航空航天大学 | Formation changing method for intensive autonomous formations of unmanned aerial vehicle |
| CN105527960A (en) * | 2015-12-18 | 2016-04-27 | 燕山大学 | Mobile robot formation control method based on leader-follow |
| CN108646550A (en) * | 2018-04-03 | 2018-10-12 | 江苏江荣智能科技有限公司 | A kind of multiple agent formation method of Behavior-based control selection |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119151A (en) * | 2019-05-24 | 2019-08-13 | 陕西理工大学 | A kind of target point distribution method of robot formation forming |
| CN110119151B (en) * | 2019-05-24 | 2022-02-11 | 陕西理工大学 | A Target Point Allocation Method for Robot Team Formation |
| CN111077887A (en) * | 2019-12-12 | 2020-04-28 | 南京理工大学 | A multi-robot comprehensive obstacle avoidance method using pilot following method |
| CN111077887B (en) * | 2019-12-12 | 2022-06-28 | 南京理工大学 | A multi-robot comprehensive obstacle avoidance method using pilot following method |
| CN111082709A (en) * | 2019-12-18 | 2020-04-28 | 南京理工大学 | Electric anti-backlash control method based on basal ganglia |
| CN111207754A (en) * | 2020-02-28 | 2020-05-29 | 上海交通大学 | Multi-robot formation positioning method and robot equipment based on particle filter |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110608743B (en) | Multi-unmanned aerial vehicle collaborative route planning method based on multi-population chaotic grayling algorithm | |
| CN108897315A (en) | A kind of Multi-agent Team Formation | |
| CN110806759B (en) | An aircraft route tracking method based on deep reinforcement learning | |
| CN106502250B (en) | The path planning algorithm of multi-robot formation in three-dimensional space | |
| CN100568144C (en) | A multi-behavior fusion automatic navigation method for mobile robots in unknown environments | |
| CN108362284A (en) | A kind of air navigation aid based on bionical hippocampus cognitive map | |
| CN114037050A (en) | Robot degradation environment obstacle avoidance method based on internal plasticity of pulse neural network | |
| CN108646550B (en) | Multi-agent formation method based on behavior selection | |
| CN113759935A (en) | Intelligent group formation movement control method based on fuzzy logic | |
| CN111707274B (en) | Energy-optimal spacecraft continuous dynamic obstacle avoidance trajectory planning method | |
| Rastogi et al. | Sample-efficient reinforcement learning via difference models | |
| Liu | A novel path planning method for aerial UAV based on improved genetic algorithm | |
| CN109227550A (en) | A kind of Mechanical arm control method based on RBF neural | |
| CN116203950A (en) | A CTU Path Planning Method Combining Particle Swarm Optimization and Dynamic Window Method | |
| CN115352466A (en) | Method and system for longitudinal control of automatic driving based on GRU network regulation mechanism | |
| Storms et al. | Equating user performance among communication latency distributions and simulation fidelities for a teleoperated mobile robot | |
| Jiang et al. | A Neural Network Controller for Trajectory Control of Industrial Robot Manipulators. | |
| CN110658816A (en) | Mobile robot navigation and control method based on intelligent assembly | |
| RU2406103C1 (en) | Method of contolling movement of dynamic object on path | |
| Abdulla et al. | Roll control system design using auto tuning LQR technique | |
| CN118153621A (en) | Multi-agent trapping method based on double-layer diagram attention reinforcement learning | |
| Zhang et al. | A Bio-Inspired Integration Model of Basal Ganglia and Cerebellum for Motion Learning of a Musculoskeletal Robot | |
| Du et al. | A Learning-Based Nonlinear Model Predictive Control Approach for Autonomous Driving | |
| Khaleghi et al. | Analysis of uav/ugv control strategies in a dddams-based surveillance system | |
| RU2445670C1 (en) | Method of controlling movement of dynamic object on space trajectory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181127 |
|
| WW01 | Invention patent application withdrawn after publication |