Gestalt Effects in Planning: Rush-Hour as an example
Gestalt Effects in Planning: Rush-Hour as an example
复制标题
规划中的格式塔效应：以高峰时段为例
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Lars Konieczny
中科院分区：
文献类型：
--
作者：
S. Bennati;S. Brüssow;Marco Ragni;Lars Konieczny
规划中的完形效应：以高峰时段为例Stefano Bennati（bennati@cognition.uni-freiburg.de）Sven Brussow（sven@cognition.uni-freiburg.de）马可·拉格尼（ragni@cognition.uni-freiburg.de）Lars Konieczny（lars@cognition.uni-freiburg.de）认知科学中心，弗赖堡大学摘要规划问题已经被广泛研究的图论性质，如步骤nec的数量，达到特定的目标状态或问题空间的大小是必要的。然而，这些结构特性并不能完全表征问题。在所提出的眼动追踪研究中，我们还调查了感知因素对规划问题解决方案的影响。虽然不影响解决方案的正确性，但结果表明，某些完形属性是导致偏离最佳计划的原因。保留字：搬迁规划;高峰时间完形介绍规划问题的特点可以由结构特性，如达到特定目标状态所需的步骤数或问题空间的大小，以及感知特性，如元素之间的颜色和空间关系。虽然结构要素已被广泛研究，但后者尚未得到足够的关注。原因可能是结构属性比感知属性更容易操纵。我们选择的问题，高峰时间，可以很容易地操纵知觉属性。Rush-Hour在一个6 × 6的网格上示意了一个拥挤的停车场。任务是为被其他车辆挡住的玩家的汽车开路。玩家的车总是红色的，水平对齐，并放置在第三排，出口所在的同一排。有两种类型的车辆：汽车（长度2）和卡车（长度3）。每辆车都有一个方向垂直或水平和颜色。所有车辆只能沿其纵向轴线沿着前后移动。游戏规则禁止移动车辆或通过另一辆车或打破周围的停车场的墙壁。目标是通过依次移动阻挡道路的车辆来清除通往出口的道路，这些车辆反过来又被其他车辆阻挡。游戏是定义明确的，可分解的，非动态的，只有一个目标。它也是PSPACE完备的（Flake & Baum，2002）。通常有一个以上的可能的解决方案，但只有少数是最佳的。我们将最优解定义为涉及最少可能移动次数的解。规划问题通常以排列问题为特征。从认知的角度来看，计划可以被定义为对行动步骤的预期或“实现特定目标或期望结果的程序”（Morris & Ward，2005）。图1：高峰时段示例配置。任务是重新安排车辆，以便红色汽车可以移动。只有左边的棋盘包含一个簇。来自不同领域的见解，例如伦敦，表明困难来自静态属性，例如规划深度，即，将初始状态转换为目标状态所需的移动次数（Kaller，Unterrainer，拉姆，& Halsband，2004; Kaller，拉姆，Küostering，& Unterrainer，2011）。Rush-Hour的一个相关属性是网格上的车辆数量，因为它增加了搜索树。动态特性，如反直觉移动的数量和循环移动序列的数量也会影响难度（Ragni，Steffenhagen，& Fangmeier，2011）。如果一个移动导致了到目标状态的更高距离，例如当目标车必须远离出口或进入出口阻塞位置时，则该移动是违反直觉的。如果一个移动序列中的一辆车被同一移动序列中的另一辆车挡住，则该移动序列是循环的。人类在两块具有相同结构特性的不同板上的表现可能不同。我们认为，这种差异的结果，从知觉属性，影响问题解决和规划过程的方式，值得更深入的调查。完形理论认为，在一定条件下，人们将一组不同的物体视为一个整体。例如，当物体彼此对齐或彼此非常接近时，它们将被视为更大物体的一部分（Koffka，1935; Köohler，1959; Wertheimer，1938）。关于高峰时段域，集群表示将板上的相邻对象分组在一起的元对象：它被定义为彼此相邻的两个或更多个车辆的组，使得它们的主轴平行（参见图1）。见图2）。在这种情况下，完形法则的邻近性和
Gestalt Effects in Planning: Rush-Hour as an example Stefano Bennati (bennati@cognition.uni-freiburg.de) Sven Br ussow (sven@cognition.uni-freiburg.de) Marco Ragni (ragni@cognition.uni-freiburg.de) Lars Konieczny (lars@cognition.uni-freiburg.de) Center for Cognitive Science, University of Freiburg Abstract Planning problems have been extensively studied with regard to graph theoretical properties such as the number of steps nec- essary to reach a specific goal state or the size of the problem space. These structural properties, however, do not completely characterize a problem. In the presented eye-tracking study we also investigated the influence of perceptual factors on the solution to a planning problem. While not affecting the cor- rectness of a solution, the results suggest that certain Gestalt properties are responsible for the deviation from optimal plans. Keywords: Move planning; Rush-Hour; Gestalt Introduction Planning problems can be characterized by structural proper- ties, such as the number of steps necessary to reach a specific goal state or the size of the problem space, and perceptual properties, such as colors and spatial relations between ele- ments. While structural elements have been widely studied, the latter have not receive as much attention. The reason may be that structural properties are easier to manipulate than per- ceptual properties. The problem of our choice, Rush-Hour, can be easily manipulated with regard to perceptual proper- ties. Rush-Hour schematizes a crowded parking lot on a 6 × 6 grid (cf. Fig. 1) and the task is to clear the way for the player’s car which is blocked by some other vehicles. The player’s car is always red, horizontally aligned and placed in the third row, the same row where the exit is. There are two types of vehicles: cars (length two) and trucks (length three). Each vehicle has an orientation–vertical or horizontal–and a color. All vehicles can only be moved forward and backward along their longitudinal axes. The game rules forbid moving a vehi- cle over or through another vehicle or breaking the walls that surround the parking lot. The goal is to clear the way to the exit by sequentially moving the vehicles that block the way, which are in turn blocked by others. The game is well-defined, decomposable, non dynamic and has only one goal. It is also PSPACE-complete (Flake & Baum, 2002). There is normally more than one possible solu- tion, but only few of them are optimal. We define an optimal solution as the solution that involves the least possible num- ber of moves. Planning problems are often characterized by permutation problems. From a cognitive perspective planning can be de- fined as the anticipation of action steps or “a procedure for achieving a particular goal or desired outcome” (Morris & Ward, 2005). Figure 1: Rush-Hour sample configurations. The task is to rearrange the vehicles such that the red car can be moved out. Only the board on the left contains a cluster. Insights from different domains, such as Tower of London, indicate that difficulties arise from static properties such as planning depth, i.e., the number of moves necessary to trans- form the initial state into a goal state (Kaller, Unterrainer, Rahm, & Halsband, 2004; Kaller, Rahm, K¨ostering, & Un- terrainer, 2011). A relevant property of Rush-Hour is the number of vehicles on the grid, as it increases the search tree. Dynamic proper- ties, such as the number of counter-intuitive moves and the number of circular move sequences have an influence on dif- ficulty as well (Ragni, Steffenhagen, & Fangmeier, 2011). A move is counter-intuitive if it results in a higher distance to the goal state, for example when the goal car has to be moved away from the exit or into an exit-blocking position. A move sequence is circular if one of its vehicles is blocked by an- other vehicle that is part of the same move sequence. Human performance on two different boards with the same structural properties may differ. We propose that this differ- ence results from perceptual properties that affect problem solving and planning processes in a way that deserves deeper investigation. Gestalt Theory asserts that under certain conditions people perceive a group of distinct objects as a holistic unit. For example, when objects are aligned to each other or are in close proximity to each other, they will be perceived as parts of a bigger object (Koffka, 1935; K¨ohler, 1959; Wertheimer, 1938). With respect to the Rush-Hour domain, a cluster rep- resents a meta-object that groups together adjacent objects on the board: It is defined as a group of two or more vehicles that are next to each other such that their major axes are parallel (cf. Fig. 2). In this context, the Gestalt laws of proximity and