Spatio-Temporal Reach and Escape Logic (STREL)

• STREL is a formal logic for specifying and monitoring spatio-temporal properties in mobile, networked cyber-physical and multi-agent systems.
• It extends Signal Temporal Logic by incorporating spatial modalities over dynamic graphs, allowing expressive reasoning about inter-agent relations coupled with temporal requirements.
• STREL employs efficient offline and online monitoring algorithms, using techniques like BFS and weighted automata to verify both Boolean and robust quantitative semantics.

Spatio-Temporal Reach and Escape Logic (STREL) is a formal logical framework for specifying and monitoring spatio-temporal properties of mobile, networked, and spatially distributed cyber-physical systems (CPS) and multi-agent systems (MAS). STREL uniformly extends Signal Temporal Logic (STL) by adding spatial modalities interpreted over dynamic, weighted graphs, allowing concise and expressive reasoning about inter-agent spatial relations coupled with temporal requirements. STREL supports both Boolean and robust (quantitative) semantics, making it suitable for both correctness verification and robustness analysis under uncertainty or distributional shift (Zhao et al., 3 Apr 2025, Nenzi et al., 2021, Bartocci et al., 2019).

1. Formal Syntax and Derived Modalities

STREL formulas are built from atomic predicates on agent or node attributes and extended by Boolean, temporal, and spatial operators. The canonical grammar is:

$$\psi ::= \text{True} \mid \pi \mid \neg\psi \mid \psi_1 \land \psi_2 \mid \psi_1 U_I \psi_2 \mid \psi_1 R_{[d_1,d_2]} \psi_2 \mid \mathcal{E}_{[d_1,d_2]}\psi$$

• Atomic predicates ($\pi$): pointwise checks on node or agent attributes or physical quantities.
• Boolean connectives ($\neg$, $\land$, $\lor$): standard propositional logic.
• Temporal operators (e.g., $U_I$ for “until” over interval $I$): as in STL, act pointwise over time.
• Spatial reach ($R_{[d_1,d_2]}$): “From the current agent, there is a route of graph-path length in $[d_1,d_2]$ along which $\psi_1$ holds up to the first node where $\pi$0 holds.”
• Spatial escape ($\pi$1): “There is a path from the agent to some node at graph-path length in $\pi$2 along which $\pi$3 holds at every node.”

Derived modalities:

Operator	Definition	Informal Semantics
$\pi$4	$\pi$5	“$\pi$6 holds somewhere within distance $\pi$7”
$\pi$8	$\pi$9	“$\neg$0 holds everywhere within $\neg$1”
$\neg$2	$\neg$3	“Eventually $\neg$4 holds in interval $\neg$5”
$\neg$6	$\neg$7	“Always $\neg$8 holds in interval $\neg$9”

Parameterization by distance metrics allows expressive encoding of notions such as “within $\land$0 hops,” “within latency $\land$1,” or “within Euclidean range $\land$2” (Nenzi et al., 2021, Bartocci et al., 2019).

2. Boolean and Robust Semantics

STREL provides both qualitative (Boolean) and quantitative (robustness) semantics.

Boolean semantics assign $\land$3 (true) or $\land$4 (false) to formula satisfaction at node $\land$5 and time $\land$6. Spatial connectives quantify over routes in the dynamic spatial graph:

• $\land$7: there exists a route with accumulated distance in $\land$8 from $\land$9 to $\lor$0, with $\lor$1 on all intermediates and $\lor$2 at $\lor$3.
• $\lor$4: there exists a route and node $\lor$5 at minimal distance in $\lor$6 where $\lor$7 holds throughout the path (Zhao et al., 3 Apr 2025, Nenzi et al., 2021, Bartocci et al., 2019).

Quantitative (robust) semantics assign real-valued robustness scores $\lor$8, defined recursively:

• Boolean connectives become min/max,
• Existential quantifiers become sup,
• Universal quantifiers become inf,
• Negation induces sign change.

Soundness: $\lor$9 implies satisfaction, $U_I$0 implies violation. Robust semantics are stable under small perturbations of input signals (Zhao et al., 3 Apr 2025, Visconti et al., 2021).

Semirings generalize the underlying aggregation, supporting Boolean, max/min, and tropical settings (Bartocci et al., 2019).

3. Spatial Operators: Intuition, Expressiveness, and Derived Modalities

STREL's spatial primitives—reach and escape—model non-local, path-dependent agent interactions via the system's underlying graph structure.

• Spatial reach ($U_I$1): Encodes causality or propagation: “agent $U_I$2 can reach a region where $U_I$3 holds by traversing agents satisfying $U_I$4 within a certain path-cost.” Used to describe connectivity, information flow, or risk propagation.
• Spatial escape ($U_I$5): Models the existence of a safe or resource-rich corridor, i.e., “agent $U_I$6 can escape through a region where $U_I$7 holds along the entire path.”

Derived modalities enable high-level constructs, e.g.,

• Somewhere: Existence of a property within a spatial neighborhood,
• Everywhere: Universal coverage within a region,
• Surround: “While staying in $U_I$8, one cannot escape without hitting $U_I$9.”

STREL can express properties such as coverage, fault tolerance, containment regions, or dynamic boundaries (Bartocci et al., 2019, Nenzi et al., 2021).

4. Monitoring Algorithms and Practical Tooling

Offline and online monitoring algorithms for STREL evaluate the satisfaction or robustness of specifications against observed (or predicted) spatio-temporal traces.

• Offline monitoring is performed by syntax-directed bottom-up traversals. For spatial operators, bounded flooding (e.g., BFS with distance cutoff) efficiently enumerates feasible paths; Dijkstra or Floyd–Warshall methods are used for escape evaluation. Complexity is $I$0 for reach (per time slice; $I$1 number of nodes), $I$2 for escape in dense graphs (Bartocci et al., 2019, Nenzi et al., 2021, Bartocci et al., 2021).
• Online/incremental monitoring is supported for imprecise signals: interval arithmetic yields partial guarantees (certainly safe, certainly violated, or uncertain), and algorithms use sliding-window data structures for temporal operators (Visconti et al., 2021).
• Automaton-based monitoring: Weighted alternating finite automata (AFA) representations of STREL formulas enable both offline and online monitoring via symbolic state updates, supporting dynamic topologies and system mobility (Balakrishnan et al., 27 Mar 2025).
• Principal tool: MoonLight is an open-source Java/Matlab monitoring tool supporting both Boolean and robust semantics, scalable to large graphs ($I$3 nodes for reach) (Bartocci et al., 2021).

5. Application Domains and Case Studies

STREL has been applied to a range of spatially distributed CPS, with extensive experimental validation.

• Ad hoc wireless and sensor networks: Fault-tolerant routing, battery-aware connectivity, coverage invariants (Nenzi et al., 2021, Bartocci et al., 2021).
• Epidemiology and social behavior: Spatio-temporal risk zoning, spread containment via spatial reach predicates (e.g., “if a susceptible is near an infected within 2 days, infection spreads within 7 days”) (Nenzi et al., 2021, Mohammadinejad et al., 2021).
• Environmental monitoring: Monitoring pollution/adverse condition escape within geographic radii, with interval semantics to handle missing/noisy sensor data (Visconti et al., 2021).
• Multi-agent and robotic systems: Connectivity, coordination, spatial progress properties under motion and communication constraints; robust runtime verification under distributional drift for drone swarms (Zhao et al., 3 Apr 2025, Alsalehi et al., 2021).
• Learning interpretable properties: Unsupervised algorithms discover and refine STREL formulas as cluster descriptors over spatio-temporal datasets, yielding succinct logical explanations for emergent behavior (Mohammadinejad et al., 2021).

6. Relationship to Other Spatio-Temporal and Geometric Logics

STREL generalizes STL by enabling reasoning about relational spatial structure in MAS and networked CPS.

• Comparison with SpaTiaL: SpaTiaL focuses on metric and geometric predicates (e.g., proximity, orientation) and quantitative STL combinations, targeting fine-grained geometric manipulation; STREL captures topological reachability, escape, and covering on arbitrary graphs, emphasizing multi-hop and connectivity properties (Luo et al., 15 Dec 2025).
• STREL's spatial modalities are richer than simple neighborhood quantifiers, as they can express path-dependent, non-local dependencies and are invariant under isometries for Euclidean-distance parameterizations (Bartocci et al., 2019).

7. Robustness Under Uncertainty and Distribution Shift

Robust quantitative semantics in STREL are essential in practice, especially for runtime verification leveraging predictive models subject to distributional shift. Boolean satisfaction is brittle to small errors, while robustness margins quantify system safety under model uncertainty.

Distributionally robust runtime verification (RPRV) for STREL leverages conformal prediction to compute lower bounds on future robustness, accounting for both prediction error and deployment-time deviation from training-time statistics. Probabilistic soundness guarantees are maintained even beyond the i.i.d. setting (Zhao et al., 3 Apr 2025).

• (Zhao et al., 3 Apr 2025) Distributionally Robust Predictive Runtime Verification under Spatio-Temporal Logic Specifications, 2025
• (Nenzi et al., 2021) A Logic for Monitoring Dynamic Networks of Spatially-distributed Cyber-Physical Systems, 2021
• (Bartocci et al., 2019) Monitoring Mobile and Spatially Distributed Cyber-Physical Systems, 2019
• (Visconti et al., 2021) Online Monitoring of Spatio-Temporal Properties for Imprecise Signals, 2021
• (Balakrishnan et al., 27 Mar 2025) Monitoring Spatially Distributed Cyber-Physical Systems with Alternating Finite Automata, 2025
• (Mohammadinejad et al., 2021) Mining Interpretable Spatio-temporal Logic Properties for Spatially Distributed Systems, 2021
• (Alsalehi et al., 2021) Neural Network-based Control for Multi-Agent Systems from Spatio-Temporal Specifications, 2021
• (Luo et al., 15 Dec 2025) NL2SpaTiaL: Generating Geometric Spatio-Temporal Logic Specifications from Natural Language for Manipulation Tasks, 2025
• (Bartocci et al., 2021) MoonLight: A Lightweight Tool for Monitoring Spatio-Temporal Properties, 2021