The unified scheduling language designed for the space mission scheduling platform

Abstract

The Unified Scheduling Language (USL) is an easy-scheduling feature delivered with the next generation of mission scheduling platform of Beijing Aerospace Control Center. As a domain-specific language, it allows the mission designer define scheduling aspects of mission with natural-like language rather than various machine oriented parameters. Unlike general purpose planning languages such as NDDL (New Domain Definition Language), the USL is not designed self-completed but lies on various external information. The technology of USL processor involves both complier and Natural Language Processing (NLP). The complier technology transit texts into computer-oriented dependency grammar trees under strict conventions, which is traditional in the computer fields. While NLP is still a progressive field with intelligence computing and big data processing introduced in. NLP applies both structured algorithms and statistics based information to understand general purpose texts. But, with inevitable probabilities, NLP may leads to deviations. With complier and NLP approaches combined, The USL uses a series of conventions to try to avoid ambiguities and still keep the agile characteristic of natural language. The USL customized symbol system and part-of-speech system are established more abundant than general NLP. Processing USL usually covers preprocessing, Chinese word segmentation, part-of-speech tagging, dependency grammar tree parsing and finally executing the grammar tree. For each approach, a process template is clarified to introduce the basic mechanism of USL processor. The dependency grammar tree parsing is the core of USL processor, which directly determine the success chance. Building the grammar tree from a scheduling requirement description involves searching the most probable node dependent relations based on known patterns. Combined with apparent rules and stochastic approaches, the tree parser is able to support USL applications. The USL is designed extendable. By defining new dictionaries, terminal symbols and non-terminal symbols, the USL accepts new basic types of tree nodes. And a training framework is given to enrich the learning bases in case of misunderstanding new patterns of texts. Currently, the symbol system of USL is defined with dozens of basic symbols. A series of node-corresponding functions are developed to interpret each node of the grammar tree. The architecture of USL processor is implemented and initially proves valid with simple natural-like texts parsed correctly. The USL requires minimum training and mission designers are liberated from filling in aggravating data forms. With the USL, the intelligence of scheduling platform is brought to a new level.