Temporal Reasoning without Transitive Tables

Representing and reasoning about qualitative temporal information is an essential part of many artificial intelligence tasks. Lots of models have been proposed in the litterature for representing such temporal information. All derive from a point-based or an interval-based framework. One fundamental reasoning task that arises in applications of these frameworks is given by the following scheme: given possibly indefinite and incomplete knowledge of the binary relationships between some temporal objects, find the consistent scenarii between all these objects. All these models require transitive tables – or similarly inference rules– for solving such tasks. We have defined an alternative model, S-languages - to represent qualitative temporal information, based on the only two relations of \emph{precedence} and \emph{simultaneity}. In this paper, we show how this model enables to avoid transitive tables or inference rules to handle this kind of problem.

💡 Research Summary

The paper tackles a fundamental difficulty in qualitative temporal reasoning: the reliance on large transitive tables (or equivalent inference rules) to compose binary temporal relations. Traditional frameworks such as Allen’s interval algebra or Vilain‑Kautz’s point algebra represent temporal knowledge as sets of binary relations and resolve consistency by consulting pre‑computed n × n transitivity matrices. These tables quickly become unwieldy as the number of basic relations grows, and they are especially problematic when the knowledge base is incomplete or contains disjunctive information.

To avoid this bottleneck, the authors introduce S‑languages, a formal‑language‑based model that uses only two primitive temporal concepts: precedence (A occurs before B) and simultaneity (A occurs at the same instant as B). The key idea is to encode each temporal object as a letter of an ordinary alphabet X, and to treat simultaneity as a set of letters—i.e., a letter of the S‑alphabet, which is defined as any non‑empty subset of 2^X. An S‑word is a sequence of such S‑letters; reading the word from left to right yields a timeline, while each S‑letter’s internal set indicates which objects co‑occur at that instant.

Three well‑known operations on formal languages are then repurposed for temporal reasoning:

1. Intersection – enforces that two S‑languages must be satisfied simultaneously, thus representing the conjunction of constraints.
2. S‑shuffle – a generalisation of the classic shuffle operator. Given two S‑words f and g, the S‑shuffle produces all possible interleavings where at each position at most one letter from each word appears, and where simultaneous events are merged into a single S‑letter. This operation replaces the composition step of a transitivity table.
3. Projection – a monoid morphism that erases all occurrences of a chosen subset of letters, allowing one to focus on a sub‑problem or to eliminate already‑resolved events.

The authors connect these operations to Delannoy paths, a combinatorial construct counting lattice paths that use north, east, and north‑east steps. By mapping the north step to a singleton S‑letter a, the east step to b, and the diagonal step to the simultaneous letter {a,b}, the set of all (p,q)‑Delannoy paths corresponds exactly to the (p,q)‑Delannoy language L(p,q) = { w ∈ {a,b}^{*} | Parikh(w) = (p,q) }. The cardinality D(p,q) satisfies the classic recurrence D(p,q)=D(p‑1,q)+D(p,q‑1)+D(p‑1,q‑1), providing a closed‑form way to count all possible temporal interleavings without enumerating a transitivity table.

Two illustrative case studies demonstrate the practical power of the approach:

• Allen’s unsatisfiable closed network – The 13 Allen relations are translated into S‑words (e.g., “before” becomes a·b, “overlaps” becomes a·{b,c}·d, etc.). By repeatedly applying intersection and S‑shuffle, the system automatically derives a contradiction (the resulting language becomes empty), confirming inconsistency without consulting any pre‑computed table.
• Manna‑Pnuelli resource allocation problem – Requests and resources are modeled as S‑letters; precedence constraints encode ordering of requests, while simultaneity encodes mutually exclusive resource usage. Projection is used to isolate each resource’s sub‑problem, and the global consistency is verified by intersecting the projected languages. Again, no transitivity tables are needed, yet the method reproduces the known feasibility results.

The paper’s contributions can be summarised as follows:

1. Elimination of transitive tables – By recasting temporal composition as language operations, the approach sidesteps the exponential blow‑up associated with traditional tables.
2. Natural handling of incomplete and disjunctive knowledge – Intersection and projection work directly on partially specified S‑languages, making the method robust to uncertainty.
3. A solid combinatorial foundation – The link to Delannoy numbers offers both theoretical insight and practical counting tools for the number of admissible temporal scenarios.
4. Demonstrated applicability – The two case studies show that S‑languages can reproduce classic results in Allen’s algebra and in concurrent resource allocation, suggesting broader applicability to planning, scheduling, and temporal planning under uncertainty.

Limitations are acknowledged. Since the model only explicitly supports precedence and simultaneity, richer Allen relations such as “overlaps”, “meets”, or “starts” must be expressed indirectly via combinations of the primitive relations, potentially requiring additional translation layers. Future work is proposed to extend the S‑language framework with hierarchical definitions for these derived relations, or to integrate S‑languages with existing relational algebras in a hybrid system.

In conclusion, the paper presents a compelling alternative to traditional temporal reasoning techniques. By leveraging formal language theory—intersection, shuffle, and projection—together with combinatorial Delannoy paths, it provides a scalable, table‑free method for checking the consistency of qualitative temporal knowledge. The approach opens new avenues for research on expressive yet computationally tractable temporal formalisms.