OBLOT Model in Swarm Robotics

• OBLOT is a minimal model of autonomous robots that operate via repeated Look–Compute–Move cycles while being oblivious, silent, and anonymous.
• It applies geometric and graph-based methods under varying scheduler regimes (FSYNC, SSYNC, ASYNC) to analyze coordination and pattern formation.
• Research using OBLOT focuses on symmetry, impossibility results, and computational bounds in distributed systems and swarm robotics.

The OBLOT model is the classical minimal model of autonomous mobile robots used in distributed computing and swarm robotics. In this model, robots are oblivious and silent: they have no persistent memory across activations and no explicit communication, and coordination is achieved only through observation of positions and motion. The model is normally phrased through repeated Look–Compute–Move cycles executed by anonymous, identical, homogeneous, and autonomous robots, either in the Euclidean plane or on graph vertices, under an adversarial scheduler that determines activation and timing. In the formulation used for arbitrary pattern formation on an infinite rectangular grid, OBLOT means “OBlivious LOOK–Compute–MOVE robots” with full and unobstructed visibility, no common coordinate agreement, and movement restricted to one adjacent grid node per move (Sharma et al., 2023).

1. Core definition and constitutive assumptions

In its standard form, the OBLOT model assumes robots that are anonymous, homogeneous, autonomous, oblivious, and silent. Anonymous means that robots carry no identifiers and are physically indistinguishable; homogeneous means that all robots execute the same deterministic algorithm; autonomous means that there is no central control; oblivious means that robots have no persistent memory from one activation to the next; and silent means that there is no explicit message exchange or external signaling (Sharma et al., 2023).

The model is also typically disoriented. In graph and geometric formulations alike, robots do not share a global coordinate system, global compass, or common chirality. In the infinite-grid formulation, each robot imposes its own Cartesian frame with origin at its current node and axes parallel to the grid lines, but there is “no global coordinate agreement and no common sense of clockwise versus counter-clockwise” (Sharma et al., 2023). In Euclidean formulations, each robot similarly has its own local coordinate system, and the classical definitions often emphasize that directions, axes, and orientation need not agree between robots (Heriban et al., 2021).

The standard OBLOT assumptions are intentionally weak. The literature repeatedly presents OBLOT as the weakest of the traditional robot models: robots have no writable memory, no lights, and no direct communication, and therefore can only exploit current geometric or graph-theoretic structure. In introductory comparisons, OBLOT is contrasted with stronger models such as FSTA, FCOM, and LUMI, which add persistent state, communication, or both (Flocchini et al., 2023).

A common formalization depends on the environment. On an infinite rectangular grid, a configuration of $k$ robots is a pair $(\mathcal G,f)$ where $\mathcal G=\mathcal P\times\mathcal P$ and $f:V(\mathcal G)\to\{0,1\}$ with $\sum_v f(v)=k$; initially $f(v)\le 1$ for all $v$ (Sharma et al., 2023). On general graphs, a configuration is often written $C=(G,\lambda)$, where $\lambda$ records robot occupancy per vertex; some works assume multiplicities are visible, while others explicitly assume that robots cannot detect multiplicities and therefore only observe whether a vertex is occupied (Cicerone et al., 4 Sep 2025). In the plane, the global state is the multiset or tuple of robot positions in $\mathbb R^2$ (Flocchini et al., 2023).

2. Look–Compute–Move semantics and scheduler models

OBLOT robots operate through the standard Look–Compute–Move cycle. In the grid formulation, when a robot is activated it first obtains the entire set of occupied grid nodes expressed in its local coordinates; it then runs the common deterministic algorithm on the snapshot and outputs either “stay” or one of the four neighboring nodes; finally, if it chose a neighbor, it moves along that single grid edge, otherwise it remains on its current node (Sharma et al., 2023). In continuous-space formulations, the same three-step pattern is used, but the computed destination is a point in $(\mathcal G,f)$0 and the robot moves rigidly in a straight line to that point (Flocchini et al., 2023).

The model is parameterized by the scheduler. The standard scheduler classes are FSYNC, SSYNC, and ASYNC. Under FSYNC, time is divided into global rounds and every robot executes one LCM cycle in lockstep. Under SSYNC, rounds still exist, but in each round an arbitrary nonempty subset of robots is activated. Under ASYNC, there are no common rounds, and arbitrary finite delays may separate Look, Compute, Move, and idle periods (Sharma et al., 2023). Some comparative works further refine asynchronous behavior through $(\mathcal G,f)$1-atomic, $(\mathcal G,f)$2-atomic, and $(\mathcal G,f)$3-atomic variants (Flocchini et al., 2023).

These scheduler distinctions materially affect solvability. A complete landscape result states that, with unlimited visibility, $(\mathcal G,f)$4, settling the question of whether semi-synchronous activation is strictly stronger than asynchronous activation for oblivious robots (Flocchini et al., 2023). In discrete synchronous settings, a related comparison gives $(\mathcal G,f)$5 through tasks solvable in full synchrony but not in semi-synchrony (Sharma et al., 2024). This does not alter the definition of OBLOT itself, but it is central to how the model is analyzed.

Several papers also study alternative schedulers that interact with OBLOT’s weak internal capabilities. Round-robin activation, in which robots are cyclically activated one at a time, is used to study gathering on highly symmetric graphs (Cicerone et al., 11 Nov 2025). Energy-constrained variants induce a scheduler in which a robot cannot be activated in two consecutive rounds, yielding a model denoted $(\mathcal G,f)$6 in cross-model analyses (Buchin et al., 2022). These are not changes to the robots’ intrinsic capabilities; they are changes in the adversarial environment controlling the same LCM discipline.

3. Visibility, movement, and environmental variants

The “classical” OBLOT model usually assumes full visibility. On grids, this is phrased as fully unobstructed visibility, meaning that in the Look phase a robot sees the entire current configuration (Sharma et al., 2023). In the Euclidean plane, the standard transparent setting likewise assumes unbounded visibility of all other robots (Heriban et al., 2021). However, later work studies how weakening visibility changes the computational landscape.

A major distinction is between full visibility and limited visibility. Under limited visibility, each robot sees only robots within a fixed radius $(\mathcal G,f)$7, and the visibility relation induces a visibility graph on the current configuration (Das et al., 2023). The visibility perspective yields strict separations: for OBLOT, full visibility is strictly more powerful than limited visibility under both FSYNCH and SSYNCH (Das et al., 2023). Limited visibility has also motivated specialized pattern-formation and near-gathering algorithms that require additional geometric assumptions such as small initial diameter, convex boundary, or hole-freeness (Hahn et al., 2024).

Another environmental distinction concerns transparency versus opaqueness. In transparent settings, all robots are visible as points. In opaque settings, line-of-sight may be blocked by closer robots, and this substantially weakens the model in conjunction with collision intolerance and asynchrony (Feletti et al., 2024). The resulting impossibility and separation results are often not consequences of obliviousness alone, but of obliviousness combined with obstructed visibility.

Movement rules also vary with the domain. In graph-based OBLOT, robots move along edges and never rest on edge interiors. On the infinite rectangular grid, the move set is exactly $(\mathcal G,f)$8 or stay, and collisions and opposite-direction edge traversals are forbidden by protocol design (Sharma et al., 2023). On general graphs, a robot chooses a destination in $(\mathcal G,f)$9, the closed neighborhood of its current vertex (Cicerone et al., 7 Sep 2025). In Euclidean OBLOT, the move is a rigid straight-line motion to a computed point (Flocchini et al., 2023).

These variants preserve the defining features of OBLOT—obliviousness, silence, anonymity, and homogeneity—but alter the sensing or motion substrate. A plausible implication is that many “OBLOT results” are not portable across domains without care: impossibility or solvability may depend as much on visibility and movement semantics as on the absence of memory and communication.

4. Symmetry, asymmetry, and impossibility structure

Symmetry is a central organizing principle of the OBLOT model. Because robots have no IDs, no memory, and no explicit communication, they cannot in general distinguish symmetric roles. In the infinite-grid APF formulation, a symmetry of a configuration is any non-trivial graph automorphism preserving occupancy, and a configuration is asymmetric if it has no such non-trivial symmetry (Sharma et al., 2023). The same basic idea reappears in graph-based gathering, where automorphism orbits determine which vertices or robots are indistinguishable (Cicerone et al., 4 Sep 2025).

This directly yields impossibility conditions. In the grid APF paper, deterministic Arbitrary Pattern Formation in the OBLOT model is stated to be possible exactly from asymmetric initial configurations; without asymmetry, “one cannot even elect a unique head or tail” (Sharma et al., 2023). In the Euclidean pattern-formation literature, the corresponding necessary condition is that the symmetry group of the target pattern must be contained in that of the initial configuration (Hahn et al., 2024). More broadly, OBLOT’s lack of persistent symmetry-breaking devices means that impossibility results are typically driven by automorphisms, rotational symmetry, or scheduler-induced replay of symmetric views.

Recent work formalizes this phenomenon through equivariant dynamical systems. For fully synchronous OBLOT swarms, the evolution map $\mathcal G=\mathcal P\times\mathcal P$0 commutes with rotational and reflective transformations together with robot relabelings, yielding an equivariance structure isomorphic to $\mathcal G=\mathcal P\times\mathcal P$1 (2503.07576). From this, one obtains a no-loss-of-symmetry theorem: if $\mathcal G=\mathcal P\times\mathcal P$2 is the isotropy subgroup of configuration $\mathcal G=\mathcal P\times\mathcal P$3, then $\mathcal G=\mathcal P\times\mathcal P$4, so a protocol can never break existing symmetries (2503.07576). Under local invertibility, symmetry cannot even increase, giving equality of isotropy groups.

A related limited-visibility study establishes an analogous symmetry-preservation theorem for synchronous oblivious swarms: if the global map $\mathcal G=\mathcal P\times\mathcal P$5 is locally invertible at $\mathcal G=\mathcal P\times\mathcal P$6, then $\mathcal G=\mathcal P\times\mathcal P$7 (Gerlach et al., 2024). This line of work is notable because it recasts classical impossibility reasoning as a structural consequence of equivariance and invertibility rather than as a task-specific ad hoc argument.

Symmetry arguments also underlie impossibility on graphs. In vertex- and edge-transitive graphs under round-robin, adjacent two-vertex occupied configurations and fully occupied configurations are ungatherable without multiplicity detection (Cicerone et al., 11 Nov 2025). On complete graphs, complete bipartite graphs, and other highly symmetric topologies, the same obstruction appears in more specialized forms (Cicerone et al., 11 Nov 2025). These results illustrate a persistent theme: in OBLOT, the absence of memory and communication makes automorphism structure a first-order computational resource.

5. Relation to stronger robot models and to computational power

The standard hierarchy of robot models places OBLOT at the bottom. In cross-model analyses, the four traditional models are $\mathcal G=\mathcal P\times\mathcal P$8, $\mathcal G=\mathcal P\times\mathcal P$9, $f:V(\mathcal G)\to\{0,1\}$0, and $f:V(\mathcal G)\to\{0,1\}$1, corresponding respectively to oblivious-and-silent robots, robots with finite states but no communication, robots with finite communication but no persistent state, and robots with both (Sharma et al., 2024). The general hierarchy stated across schedulers is

$f:V(\mathcal G)\to\{0,1\}$2

in one comparison framework (Flocchini et al., 2023).

This hierarchy reflects concrete differences in solvability. For example, Rendezvous is unsolvable in $f:V(\mathcal G)\to\{0,1\}$3 but solvable in $f:V(\mathcal G)\to\{0,1\}$4, where a one-bit signaling mechanism can break symmetry (Flocchini et al., 2023). Likewise, the literature repeatedly contrasts OBLOT with LUMI, whose persistent lights provide both communication and constant-size memory; the APF paper on the infinite rectangular grid explicitly presents LUMI as computationally stronger while emphasizing that OBLOT is the “classical minimal model” (Sharma et al., 2023).

The role of synchrony complicates this picture. In some domains, stronger scheduling can compensate partly for weaker robot capabilities. For discrete synchronous environments, a comparison table gives $f:V(\mathcal G)\to\{0,1\}$5 (Sharma et al., 2024). For two robots, a complete characterization finds a striking collapse under full synchrony: $f:V(\mathcal G)\to\{0,1\}$6 showing that perfect synchrony can substitute for memory and communication when $f:V(\mathcal G)\to\{0,1\}$7 (Kitamura et al., 28 Dec 2025). This is an exceptional two-robot phenomenon rather than a general property of OBLOT.

Another important refinement concerns the Compute phase itself. Classical OBLOT papers generally do not measure or bound internal computational resources; cost measures instead concern rounds, robot moves, or movement distance. A 2025 graph-based study shows that, if this absence of bounds is taken literally, unlimited computational power has major consequences: synchronous OBLOT robots on finite graphs can solve every problem in a broad class $f:V(\mathcal G)\to\{0,1\}$8 of formation and reachability tasks with the minimum possible number of rounds and robot-moves by building and solving the entire configuration hypergraph (Cicerone et al., 4 Sep 2025). The authors explicitly argue that unbounded Compute “trivializes” a large class of problems and that realistic computational limits are needed if one wants nontrivial impossibility and hardness phenomena on graphs (Cicerone et al., 4 Sep 2025).

This suggests that “weakness” in OBLOT is multidimensional. With respect to memory and communication, the model is minimal. With respect to per-cycle internal computation, the classical formalism is permissive unless complexity bounds are added explicitly.

6. Representative applications and current research directions

The OBLOT model serves as a baseline framework for classical coordination tasks such as gathering, rendezvous, arbitrary pattern formation, general pattern formation, uniform circle formation, and leader election. In the discrete APF problem on the infinite rectangular grid, an algorithm is given for fully asynchronous OBLOT robots with asymmetric initial configuration, and its performance is analyzed in terms of space complexity and number of moves (Sharma et al., 2023). In graph settings, OBLOT is used to study gathering under round-robin on vertex-transitive, edge-transitive, and non-vertex-transitive graphs, including hostile variants without multiplicity detection (Cicerone et al., 11 Nov 2025).

The model is also a testbed for comparative complexity measures. One work on unreliable sensors studies the maximum traveled distance as a fuel-consumption metric for gathering in OBLOT under FSYNC, SSYNC, and ASYNC, and analyzes how sensing errors affect rendezvous and leader election (Heriban et al., 2021). Another line investigates energy-constrained extensions, where recharge constraints modify the effective scheduler and can make certain tasks solvable that are impossible in the usual semi-synchronous setting (Buchin et al., 2022).

Recent research directions around OBLOT include three themes. First, there is increasing attention to landscape theorems that compare schedulers, visibility assumptions, and robot models in a unified way (Flocchini et al., 2023). Second, there is a move toward structural mathematics of symmetry, using equivariant maps, isotropy lattices, and visibility-graph automorphisms to characterize what oblivious swarms can preserve or gain (2503.07576). Third, there is a growing recognition that the classical model leaves the Compute phase effectively unbounded, and that explicit algorithmic complexity assumptions may be necessary for meaningful graph-based lower bounds and impossibility claims (Cicerone et al., 4 Sep 2025).

Across these developments, the defining idea of OBLOT remains stable: robots are oblivious, silent, anonymous, homogeneous, and autonomous, and every decision is recomputed from the current snapshot alone. The enduring interest of the model lies precisely in this minimality. It isolates what coordination can be extracted from geometry, topology, and scheduler structure when neither memory nor explicit communication is available.