Opening — Why this matters now

Circular economy rhetoric is everywhere. Circular economy decision-making is not.

Most end-of-life products still follow a depressingly simple rule: disassemble until it hurts, or stop when the operator gets tired. The idea that we might formally decide when to stop disassembling — based on value, cost, safety, and information — remains oddly underdeveloped. This gap is no longer academic. EV batteries, e‑waste, and regulated industrial equipment are forcing operators to choose between speed, safety, and sustainability under real constraints.

This paper steps into that gap with an unfashionable but powerful idea: triage requires memory.

Background — Context and prior art

Disassembly Sequencing Planning (DSP) has had a long, respectable life optimizing how to take things apart. Graphs, precedence matrices, heuristics, metaheuristics — the toolbox is full. What it mostly lacks is a way to answer a more uncomfortable question:

Should we keep going at all?

Existing DSP models typically assume:

  • Fixed goals (e.g. full disassembly)
  • Deterministic conditions
  • No explicit comparison between partial recovery and terminal routing

Circular Economy (CE) routing — reuse, repurpose, recycle, dispose — is often bolted on as an afterthought. Worse, it is usually heuristic, opaque, or policy-driven rather than value-driven. The result is decision-making that looks systematic on paper and improvisational on the shop floor.

Analysis — What the paper actually does

The authors introduce a state-augmented disassembly graph. This sounds technical because it is — but the intuition is simple.

The core insight

In a normal disassembly graph, different paths can lead to the same physical component. The graph forgets how you got there.

That is fatal for decision-making.

The paper fixes this by encoding disassembly history into the state itself. Each node becomes:

[ \tilde{v} = (v, \tau) ]

Where:

  • $v$ is the physical component
  • $\tau$ is the history of executed steps

This enforces the Markov property: every decision depends only on the current state, not hidden past actions. In practical terms, it means:

  • Safety checks know what has already been done
  • CE options know what information is available
  • Costs are accumulated correctly

And — crucially — the planner can now compare stopping vs continuing at every step.

Stop vs Continue as a formal choice

At each augmented state, the agent chooses between:

  1. Continue disassembly (pay cost now, maybe gain access later)
  2. Commit to a CE route (reuse, repurpose, recycle, dispose)

Utility is defined cleanly:

[ U_{v,k}(H_v, \tau) = \text{Revenue}{v,k}(H_v) - \text{Cost}{v,k}(\tau) ]

This makes the problem recursively solvable using deterministic value iteration — no black-box optimisation, no mystical heuristics.

Findings — What the EV battery example reveals

The EV battery case is where the framework stops being polite and starts being honest.

Three health scenarios, three outcomes

Scenario Optimal Decision Why
High health Reuse pack Further disassembly destroys value
Mixed health Extract + reuse one module Partial disassembly dominates whole-pack routing
Degraded Recycle entire pack Module-level recycling is economically punished

The last case is the most revealing.

When “optimal” is environmentally wrong

In the degraded scenario, recycling the entire battery pack yields higher economic utility than disassembling and recycling modules individually. This is not a modelling flaw — it is a policy diagnosis.

The framework exposes a misalignment:

  • Operator incentives favor shallow recycling
  • Environmental goals require deeper separation

The model does something rare in sustainability research: it quantifies where subsidies or regulation would actually matter.

Implications — Why this matters beyond batteries

Three implications stand out.

1. Circular economy needs sequential AI, not static optimisation

This is not a job for one-shot MILPs or static heuristics. Triage decisions unfold as information is revealed. The framework is structurally compatible with dynamic programming and reinforcement learning — without turning into an inscrutable black box.

2. Interpretability is a feature, not a compromise

Every decision decomposes into:

  • Health
  • Cost
  • Access
  • Constraints

This matters for regulated environments where “the model decided” is not an acceptable explanation.

3. Policy design can finally be data-driven

By making value gaps explicit, the framework provides targets for intervention. Instead of vague incentives for “more recycling,” policymakers can ask: Which step is economically irrational, and by how much?

Conclusion — The quiet power of remembering the past

This paper does not promise magical automation. It does something better.

It restores memory to disassembly.

By augmenting states with history, the authors turn CE triage from a heuristic art into a sequential decision problem with clear trade-offs and interpretable outcomes. The result is a framework that is adaptable, explainable, and — perhaps most importantly — honest about where economics and sustainability diverge.

Circular economy will not be saved by slogans. It will be saved by models that know when to stop.

Cognaptus: Automate the Present, Incubate the Future.