![]() How does the robot know which objects should be combined to construct the replacement tool? One possible solution is for the user to manually encode the correct object combination in the goal definition, and the search procedure would find it. In the absence of a hammer that is required for hammering a nail to complete the task, the robot may choose to construct a replacement for the hammer using the objects available to it. For example, consider a task where the goal of the robot is to hang a painting on the wall. However, the application of heuristic search algorithms to perform tool construction in the context of task planning can be challenging. Heuristic search algorithms, such as A * and enforced hill-climbing (EHC), have been successfully applied to planning problems in conjunction with heuristics such as cost-optimal landmarks (Karpas and Domshlak, 2009) and fast-forward (Hoffmann and Nebel, 2001), respectively. FGS enables efficient application of existing heuristic search algorithms in the context of task planning in order to perform tool construction by accounting for physical attributes of objects (e.g., shape, material) during the search for a valid task plan. In contrast, this is the first work to address the problem through the construction of replacement tools, by introducing a novel approach called Feature Guided Search (FGS). Existing work that addresses the problem of planning in the case of missing tools focuses on directly substituting the missing tool with available objects (Agostini et al., 2015 Boteanu et al., 2015 Nyga et al., 2018). ![]() The robot must then derive a task plan that involves constructing an appropriate replacement tool from objects that are available to it, and use the constructed tool to accomplish the task. Specifically, we address the scenario in which a robot is provided with a task that requires certain tools that are missing or unavailable. In this paper, we focus on the problem of tool construction in the context of task planning. The ability to improvise and construct necessary tools can greatly increase robot adaptability to unforeseen circumstances, enabling robots to handle any uncertainties or equipment failures that may arise (Atkeson et al., 2018). While the capability to construct tools is often regarded as a hallmark of sophisticated intelligence, similar improvisation capabilities are currently beyond the scope of existing robotic systems. In addition to humans, other primates and certain species of birds have also been shown to creatively accomplish tasks by constructing tools from available objects, such as sticks and stones (Jones and Kamil, 1973 Toth et al., 1993 Stout, 2011). The makeshift carbon dioxide filter constructed on board the Apollo 13 (Cass, 2005), and the jury-rigged ventilators built to combat equipment shortages during COVID-19 (Turner et al., 2020), are examples of human ingenuity in the face of uncertainty. Humans often show remarkable improvisation capabilities, particularly in times of crises.
0 Comments
Leave a Reply. |