Appendix D — Hybrid Dynamical Systems Background

Purpose: Provide mathematical background on hybrid dynamical systems and their application to RelationWorld's state space.

D.1 What Is a Hybrid Dynamical System?#

A hybrid dynamical system (HDS) combines:

1. Continuous dynamics: smooth evolution governed by differential equations.
2. Discrete dynamics: mode switches or jumps (resets) at specified conditions.

Formal definition (simplified for this context):

A hybrid dynamical system is a 6-tuple $(Q, X, f, G, R, \mathrm{Init})$ where:

Execution:

• Within mode $q$, the state $x(t) \in X(q)$ evolves via $\dot{x} = f_q(x)$.
• When $x(t)$ hits guard $G_{q,q'}$, a discrete transition fires: $q$ changes to $q'$, and $x$ is reset via $R_{q,q'}$.
• In the new mode $q'$, continuous evolution resumes.

Key property: The state space is stratified — not a single manifold, but a union of different-dimensional spaces.

D.2 Application to RelationWorld#

Mapping RelationWorld to the HDS Framework#

The Six Events as Discrete Transitions#

D.3 Continuous Dynamics: Yang--Mills Gradient Flow#

The Flow Equation#

Within a fixed topology $(n, E)$, the continuous dynamics are driven by Yang--Mills gradient flow:

where:

• $h = (h_1, \ldots, h_n) \in G^n$ is the gauge transformation vector.
• $\mathcal{E}_{F^\circ}$ is the Yang--Mills (flattening) energy on the fruit interior (Chapter 7).
• $\nabla_{h_i}$ is the gradient w.r.t. the $i$-th factor.

Convergence (Theorem H, Chapter 7): For fixed topology, this gradient flow converges to a critical point (local minimum) of $\mathcal{E}_{F^\circ}$.

Geometric interpretation: The flow minimizes the "curvature" of the connection, driving it toward flatness. In the flat limit (all holonomies $\Omega = e$), the system reaches a stable equilibrium.

Phase Portrait#

The phase portrait of the continuous dynamics within a topology $(n, E)$ has:

• Stable fixed points (local minima of $\mathcal{E}_{F^\circ}$).
• Saddle points (saddle critical points).
• Unstable manifolds (rare; the energy is Morse-like).

Trajectories may approach different fixed points depending on initial conditions (basin of attraction).

D.4 Discrete Dynamics: Guard Crossings and Resets#

Topology-Change Events#

When the system trajectory $(n(t), E(t), w(t), g(t))$ hits a guard condition, a discrete transition fires.

Guard: CONTACT(i,j)#

Condition: External command or intrinsic rule triggers edge activation.

Effect:

• Edge support expands: $E \to E \cup \{(i,j), (j,i)\}$.
• Dimension of continuous state increases: $|P(n, E)| \to |P(n, E \cup \{(i,j),(j,i)\})|$.
• The reset map initializes the new edge: $w_{ij} := \varepsilon > 0$ (small perturbation), $g_{ij} := e$ (identity).

Consequence for fruits: Adding an edge may create new fruits (if the new edge has low conductance) or merge existing fruits (if the new edge has high conductance).

Guard: BIRTH(v)#

Condition: External command introduces a new node.

Effect:

• Node set expands: $V \to V \cup \{v\}$; $n \to n+1$.
• Gauge group grows: $G^n \to G^{n+1}$.
• Initial edge specification: typically a pre-computed set $E_v$ of edges from $v$ to existing nodes.

Consequence for fruits: The new node may form its own fruit or join existing fruits, depending on weights.

Guard: DEATH(v)#

Condition: External command removes a node.

Effect:

• Node set shrinks: $V \to V \setminus \{v\}$; $n \to n-1$.
• All edges incident to $v$ are deleted: $E \to E \setminus E_v$.
• The fruit containing $v$ is recomputed (may split or vanish entirely).

Threshold-Based Guards: SPLIT/MERGE#

SPLIT guard condition:

When the Cheeger conductance $\phi(F)$ decreases below the threshold $\theta$, a new fruit "splits off." Formally, the fruit detection algorithm in Chapter 4 recomputes $\mathfrak{F}_t$ and finds a new component.

MERGE guard condition:

When the conductance of a union drops below $\theta$, two fruits merge.

Reset for SPLIT/MERGE: The discrete topology $(n, E)$ does not change. Only the derived structures $(\mathfrak{F}_t, \Sigma_t)$ are recomputed. This is a "mode switch" within the same continuous state space $P(n,E)$.

D.5 Switching and Transversality#

Transversal Crossings#

A guard crossing is transversal if the trajectory crosses the guard set transversely (not tangentially). Formally:

Definition D.1 (Transversal crossing): The trajectory $x(t)$ is transversal to guard $G_{q,q'}$ at time $t^*$ if:

(the velocity is not parallel to the guard surface).

Consequence: Near a transversal crossing, the solution is locally unique, and the reset map applies cleanly.

Zeno Behavior and Chattering#

Zeno behavior occurs when a system undergoes infinitely many discrete transitions in finite time.

Example: If SPLIT/MERGE guards are very close, the system might oscillate rapidly between modes.

Mitigation: Hysteresis (guard conditions with threshold width) prevents Zeno. In RelationWorld, we can impose:

• Splitting threshold: $\phi < \theta - \delta$ (strictly below).
• Merging threshold: $\phi > \theta + \delta$ (strictly above).

This creates a dead band and prevents oscillation.

D.6 Lyapunov Stability and Convergence#

Energy Function#

Define the total Lyapunov function for the hybrid system:

where $\mathcal{E}(x)$ is the Yang--Mills energy of the configuration $x$ in mode $q$, and the topological cost penalizes topology changes (or encourages them, depending on design goals).

Stability Theorem#

Theorem D.1 (Hybrid stability): The hybrid evolution satisfies:

provided:

• Yang--Mills gradient flow is applied continuously (Theorem H, Chapter 7).
• Discrete events are sufficiently sparse (no Zeno behavior).

Consequence: The system converges to an attractor (equilibrium or limit cycle in the state space).

D.7 Examples from RelationWorld#

Example 1: Stable Equilibrium#

Scenario: A 5-node complete graph with two fruits $F_1, F_2$ well-separated (high inter-fruit conductance threshold).

1. Continuous phase: Yang--Mills flow flattens the connection within each fruit.
2. Equilibrium: Both fruits stabilize with $\Omega(\triangle) \approx e$ (flat). The system converges to an energy minimum.
3. No guards triggered: Topology remains fixed; fruits remain fixed.
4. Limit behavior: The system sits at a fixed point $x^* \in P(n,E)$ with $\dot{x} = 0$.

Example 2: SPLIT Event#

Scenario: A single fruit $F$ with weights slowly decreasing due to exogenous DEFORM.

1. Continuous phase: $\phi_t(F)$ increases gradually as weights decrease.
2. Guard trigger: At time $t^* = 100$, $\phi_t(F)$ crosses $\theta$ from below.
3. Discrete transition: Fruit detection recomputes $\mathfrak{F}_{t^*}$. Now, $F \notin \mathfrak{F}_{t^*}$, and two new fruits $F_1, F_2$ appear.
4. Reset: The discrete topology $(n, E)$ is unchanged; the recomputation of $\mathfrak{F}$ is the reset.
5. Continued evolution: Yang--Mills flow continues in the same topology, now with two fruits instead of one.

Example 3: BIRTH Event (Node Addition)#

Scenario: At time $t^* = 50$, an external event triggers node $v = 6$ to appear in a 5-node graph.

1. Pre-BIRTH: Mode $q = (5, E)$, continuous state $x \in \mathbb{R}_{\ge 0}^{|E|} \times G^{|E|}$.
2. Guard trigger: Command BIRTH(6).
3. Reset:
   • New mode: $q' = (6, E \cup E_6)$ where $E_6 = \{(6,j), (j,6) : j \in J\}$ (pre-specified neighborhoods).
   • New continuous state: $x' \in \mathbb{R}_{\ge 0}^{|E \cup E_6|} \times G^{|E \cup E_6|}$.
   • Old edges retain their parameters; new edges from node 6 are initialized (small weights, identity transits).
4. Continued evolution: Yang--Mills flow in the new 6-node topology.

D.8 Formal Relationships to Chapter 0#

Chapter 0 (Formal State Space) defines:

• State space $\mathcal{S} = \bigsqcup_{(n,E)} C(n,E)$ (configuration spaces per topology).
• Gauge quotient: $C(n,E) = P(n,E) / G^n$.
• Event maps: transitions between strata.

This appendix provides the dynamical systems framework:

• The hybrid automaton formalization of the evolution.
• Guard conditions specifying when transitions occur.
• Reset maps governing how continuous state changes during transitions.
• Convergence analysis via Lyapunov functions.

Connection:

• The configuration space $C(n,E)$ (Definition 0.4) is the semantic state space (gauge-equivalence classes).
• The raw parameter space $P(n,E)$ (Definition 0.2) is the computational representation.
• The hybrid system evolves in $P(n,E)$ but should be understood modulo gauge equivalence (i.e., in $C(n,E)$).

D.9 Literature and Further Reading#

Hybrid dynamical systems are a well-developed area. Key references:

1. Liberzon (2003) — Switching in Systems and Control. Covers switched linear systems and stability conditions.
2. Goebel et al. (2012) — Hybrid Dynamical Systems: Modeling, Stability, and Robustness. General theory with applications.
3. van der Schaft & Schumacher (2000) — An Introduction to Hybrid Dynamical Systems. Detailed mathematical treatment.

Gauge theory on graphs:

• Baez & Lauda (2010) — discrete analogues of continuous gauge theory.
• Landi & Lanteri — applications to combinatorial geometry.

Yang--Mills flow:

• Rade (1994), Donaldson (2002) — continuous setting; the discrete analogue in Chapter 13 extends these ideas.

D.10 Open Problem: Hybrid Attractors for RelationWorld#

Problem D.1: Characterize the attractors of the hybrid system $(Q, X, f, G, R)$ governing world evolution. Specifically:

1. For what initial configurations do stable multi-fruit equilibria exist?
2. Can the system exhibit limit cycles or strange attractors (chaotic behavior)?
3. What is the basin of attraction for a given attractor?

Understanding these questions would provide a complete picture of "typical" world behavior over long timescales.