Intervals are often used to describe temporal events in knowledge representation. In traditional physics, an event $E$ is usually instantaneous, so it can be characterized by the moment of time $t(E)$ when this event occured. In common sense reasoning, an "event" is usually something that takes some time, e.g., "I am crossing the street". Such a non-instantaneous event $E$ is better characterized by a time interval $[t^-(E),t^+(E)]$, where $t^-(E)$ is the starting time of the event and $t^+(E)$ is its ending time. Possible positions of two intervals characterize possible temporal relations between two events (e.g., "$E$ is before $E'$", or "$E$ starts with $E'$", etc.).
Interval formalisms that are currently used in knowledge representation use intervals from the real number line (standard time). However, one of the major potential applications of such formalisms is planning different actions. Depending on which actions we choose, one and the same event may take different amount of time. It is therefore reasonable, instead of considering time as a real number line (standard time), to consider a branching tree-like time (physical time) in which at any moment when a decision can change the future, the time line branches into two (or more) lines that describe possible futures. In this new formalism, events are intervals in physical time. In this case, in addition to traditional relations between intervals, we get new relations of the type "$E$ and $E'$ are events in different possible worlds, so they cannot influence each other".
These new relations are similar to the ones introduced in special relativity, where in addition to three possibilities of Newtonian physics ("$E$ precedes $E'$"; "$E'$ precedes $E$", and "$E$ and $E'$ are simultaneous"), we get the fourth one: "$E$ and $E'$ are spatially separated and thus, cannot influence each other". Because of this similarity, the authors call their new interval formalism relativistic.