A Query is evaluated to find out whether some statement is true or false. If the statement has variables, then the query is true if there is a substitution of the variables that makes the corresponding statement true. Queries can be constructed using the Formula class, and are translated into query pipelines internally. The result of an evaluation is an answer, which in the case of KnowRob, can refer to a positive answer (i.e. "yes"), a negative answer (i.e. "no"), or a neutral answer (i.e. "unknown"). Additionally, answers can be marked as "uncertain" to allow a conflicting but certain answer to overrule the uncertain one.

Query Pipelines

KnowRob represents query pipelines as a form of communication graph where Tokens are passed between nodes. The construction and evaluation of such pipelines is encapsulated in the QueryPipeline class. Each node in the graph is a query processor that reacts on input tokens by producing output tokens. The graph is constructed by recursively translating a query into a set of query stages. The query stages are implemented as classes that inherit from the TokenStream class. For example, the TokenBroadcaster class takes each input token and sends it to all of its output streams.

Each token is either an Answer or a ControlToken. An Answer indicates the result of a query, and can be positive, negative, or neutral. A ControlToken is used to control the flow of the query pipeline. Currently, only the EndOfEvaluation control token is used to indicate that the evaluation of a query has finished. This is needed as each stage can produce multiple output tokens for each input token, so it must indicate when it has finished producing output tokens.

There is a special syntax available to create query pipelines using the C++ stream operator >>. The expression s1 >> s2 creates a link between two query stages where s1 must be a broadcast stream (i.e. one with dedicated outputs), and s2 is a stage that follows s1 (i.e. it receives outputs of s1as input).

Query Answers

The Answer class is used to represent the result of a query evaluation. It can be positive (AnswerYes), negative (AnswerNo), or neutral (AnswerDontKnow). Answers can be well-founded or not. A well-founded answer is one that is based on certain knowledge only, while a non-well founded answer is (at least partially) based on assumptions. For example, a negative answer can be well-founded if a sub-system can prove that the answer cannot be positive.

Query Parsing

The QueryParser class provides a simple interface to parse queries and terms. Below table lists some examples: