It is critical for advanced manufacturing machines to autonomously execute atask by following an end-user's natural language (NL) instructions. However, NLinstructions are usually ambiguous and abstract so that the machines maymisunderstand and incorrectly execute the task. To address this NL-basedhuman-machine communication problem and enable the machines to appropriatelyexecute tasks by following the end-user's NL instructions, we developed aMachine-Executable-Plan-Generation (exePlan) method. The exePlan methodconducts task-centered semantic analysis to extract task-related informationfrom ambiguous NL instructions. In addition, the method specifies machineexecution parameters to generate a machine-executable plan by interpretingabstract NL instructions. To evaluate the exePlan method, an industrial robotBaxter was instructed by NL to perform three types of industrial tasks {'drilla hole', 'clean a spot', 'install a screw'}. The experiment results proved thatthe exePlan method was effective in generating machine-executable plans fromthe end-user's NL instructions. Such a method has the promise to endow amachine with the ability of NL-instructed task execution.
展开▼
