基于人机协同的无人系统自主性评估方法

Autonomy Evaluation of Unmanned Systems for Human-robot Teamwork

针对智能无人系统的自主能力评价问题,开展了基于人机协同的无人系统自主性评估方法技术研究。首先,分析了现有自主性评估方法未充分考虑无人系统的内部模型表示和算法实现方式的不足,提出了将无人系统的感知与认知、决策与规划、行动与控制等三个基础能力维度的协同性和学习性作为其自主性评估的主要因素。其次,设计了包含人操控、人委派、人监督、混合主动、全自主等五个自主等级的自主性评估表,揭示了每个自主等级的协同性和学习性内涵。最后,提出了一种基于无人系统基础能力加权求和的无人系统综合自主等级量化计算方法,并且通过案例验证了自主等级量化参考表的有效性。所提出的方法有望广泛应用于各类无人系统的自主性评估。

In order to evaluate the levels of autonomy(LOA)for intelligent unmanned systems,a novel autonomy evaluation method is proposed in the context of human-robot teamwork.Firstly,the shortcomings of the current methods are discussed that they have not sufficiently considered how the internal models are represented and how the algorithms are used for the unmanned systems.Furthermore,the collaboration and learning ability of the three basic ability dimensions of perception and cognition,decision-making and planning,and action and control of unmanned systems are proposed as the main factors for their autonomy evaluation.Then,a autonomy evaluation table is designed that includes five levels of autonomy:human control,human delegation,human supervision,hybrid initiative,and full autonomy,and the connotations of teamwork and learning are explained in details for each level of autonomy.Finally,a quantitative calculation method of synthetical LOA is proposed using weighted sum of basic capabilities of unmanned system,and the effectiveness of the quantitative reference table of autonomous level is verified by a case study.The proposed method can be widely applied to the autonomy evaluation of various unmanned systems.