基于在线调整的群体聚合形态动态生成方法

Swarm Pattern Generation Based on Online Adjustment

目前大多数用于群体聚合形态生成的基因调控网络模型(GRN)不仅需要先验知识进行设计,而且在设计前假设环境是已知的。这些局限使在动态的、未知的环境下对GRN的设计往往是一个反复试错的过程。为此,本研究提出一种基于在线优化群体聚合形态动态生成方法。该方法将探测到的目标及其周围障碍物信息作为GRN的输入,通过在线调整策略动态生成适应当前环境的群体形态。首先,该方法设计一个基于在线优化的分层GRN,上层结构负责群体形态的自适应生成,下层结构负责引导群体机器人移到群体形态上。其次,该模型根据所检测到的环境信息,实时、自动地调用两种不同的围捕策略。具体而言,当检测到目标与障碍物距离大于等于预设的安全距离时,通过在线优化GRN的参数,生成自适应的群体形态对目标进行围捕。反之,则采用群体与障碍物合作的策略对目标进行围捕。最后,通过设计两个典型的仿真场景实验进一步验证了所提方法在受限环境下对动态目标围捕的有效性。

Currently,most gene regularity networks(GRNs)for generating swarm patterns usually require some prior knowledge and the design of GRNs is usually based on the assumption that the environment is known in advance.Therefore,the design of GRNs is generally a trial-and-error process,especially in the dynamic and unknown environments.To address these issues,this paper proposes a swarm pattern generation method by using a GRN with an online optimization mechanism.The positions of targets and the obstacles around the targets are as the input of the GRN,and an online adjustment approach is employed to generate swarm patterns which are suitable for current environments.Firstly,the proposed method designs a hierarchical gene regulation network model based on online optimization.The upper structure is responsible for the adaptive swarm patterns,and the lower structure is responsible for guiding the swarm robots to move to the swarm patterns.Secondly,the proposed model can generate suitable swarm patterns in real time when the environment changes.Specifically,when the distance between a target and obstacles is detected to be greater than or equal to a preset safety distance,the target is rounded up by generating an adaptive swarm patterns through online optimization of GRN parameters.Conversely,a cooperative strategy between the swarm patterns and the obstacles is used to round up the target.Finally,two typical simulation scenarios are set to verify the effectiveness of the proposed method.