Route Replanning¶
Replanning procedure — instantiates Oriented Goal Wayfinding
Discards a route that has become invalid and synthesizes a fresh whole route from the current position to the goal.
Route Replanning is the global recovery move: when the route currently being followed has become invalid — a segment is blocked, newly-revealed terrain turns out to be impassable, or orientation confidence has collapsed — it discards the whole plan and synthesizes a fresh route from the current position to the goal. Its defining idea is that it operates at the level of the entire route, not the individual step: it does not shuffle back one move and try a neighbor; it throws away the invalidated plan and computes a new one wholesale. This is what keeps a goal reachable in a space that changes after planning — the mechanism that says "the old way is dead; here is a completely different way," reactively and from scratch, as often as the world forces it.
Example¶
A freight dispatcher runs a fleet of delivery trucks across a metro area, each on a route planned that morning. Mid-afternoon, a highway bridge that several routes depend on is closed for emergency repair — a segment that was passable at planning time is now impassable. Replanning triggers: for every affected truck, the system takes the truck's current position and its remaining stops and computes an entirely new route around the closure, rather than nudging the old plan. One truck already past the bridge is untouched; three approaching it are handed fresh routings through surface streets; one is redirected to reorder its stops entirely because the new geography makes the old sequence wasteful. The old routes are simply abandoned. Minutes later the fleet is moving on plans that did not exist an hour ago, each valid for the world as it now is.
How it works¶
The procedure sits behind a trigger and a synthesizer. The trigger watches for invalidation: an obstacle report, a revealed unknown region that turns out non-traversable, or orientation confidence dropping below threshold. When it fires, the synthesizer takes the current position, the goal, and the updated map — now including whatever new obstacle prompted the replan — and generates a fresh, complete route, discarding the sunk plan without sentiment. The new route is committed and followed until the next invalidation. The whole design tension is in the trigger: replan too eagerly and the agent thrashes, forever abandoning barely-troubled plans; replan too reluctantly and it fixates on a route the world has already broken. Hysteresis on the trigger — a margin between "wobbling" and "genuinely invalid" — is what keeps the mechanism from either extreme.
Tuning parameters¶
- Trigger sensitivity — how invalid a route must be before replanning fires. Sensitive triggers adapt fast but thrash on noise; sluggish triggers are stable but fixate on broken plans.
- Replan scope — recompute the whole route, or only the affected leg. Full replans find globally better routes; partial replans are faster and less disruptive.
- Frequency cap / hysteresis — a floor on time or distance between replans. Prevents oscillation, at the cost of slower response to a genuinely changed world.
- Cost model — what the new route optimizes (time, distance, safety, reversibility). Different models yield very different reroutes for the same obstacle.
- Commit stance — commit hard to the new route, or hold it tentatively pending confirmation the obstacle is real.
When it helps, and when it misleads¶
Its strength is keeping the goal reachable in dynamic spaces where obstacles appear after the plan was made. When a route breaks, replanning restores a valid path instead of leaving the agent stranded on a dead plan or grinding through local dodges that cannot escape a large obstacle. Formal dynamic-replanning algorithms exist precisely for this — recomputing an optimal path efficiently as the map changes.[1]
It fails at both extremes of its trigger. Too trigger-happy and it thrashes: replanning so often on transient noise that the agent lurches between routes and never makes progress. Too reluctant — or paired with sunk-cost instinct — and it exhibits route fixation, clinging to a plan the world has already invalidated. Replanning on noise rather than signal is the classic misuse, burning effort re-deriving routes that did not need changing. The guarding discipline is hysteresis and a frequency cap on the trigger, plus distinguishing a real, persistent obstacle from a momentary wobble before tearing up the plan.
How it implements the components¶
backtracking_and_reroute_rule— the reroute half of that rule: abandon the invalidated route and synthesize a fresh whole route to the goal.incomplete_map_boundary— a newly-revealed, non-traversable unknown region is a primary replan trigger; the procedure incorporates it into the new plan.orientation_confidence_threshold— a confidence drop below threshold is itself a trigger to replan rather than push on.
It does not implement candidate_move_set — step-level neighbor search with single-step undo is local_search_with_backtracking, which works locally where this works globally. Nor does it own the static up-front waypoint_sequence: the planned chain of subgoals is waypoint_decomposition; replanning reacts and regenerates rather than decomposing in advance.
Related¶
- Instantiates: Oriented Goal Wayfinding — it fills the loop's global recovery role when a route becomes invalid.
- Consumes: waypoint_decomposition — it regenerates the route (and its waypoint chain) that decomposition originally laid out.
- Sibling mechanisms: breadcrumb_trail · dead_reckoning_loop · guided_traversal_protocol · landmark_based_wayfinding · local_search_with_backtracking · map_matching_and_recalibration · orientation_overlay · turn_by_turn_route_guidance · waypoint_decomposition
Editorial Notes¶
Form Classification¶
Form family: Control, Automation & Runtime
Rationale: Route Replanning operates by senses route invalidation and replaces the active route with a synthesized feasible alternative during operation. That concrete deployed or enacted form is Control, Automation & Runtime under the frozen taxonomy.
Nearest alternative: Decision, Gate & Allocation — Although Decision, Gate & Allocation can support this mechanism, the frozen evidence makes its operative form the act that senses route invalidation and replaces the active route with a synthesized feasible alternative during operation; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Robotics & Automation
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Recomputing a collision-free route from the robot's current state after the prior path becomes invalid is canonical autonomous-motion planning. NASA mobile-robot navigation work directly joins localization, obstacle sensing, path planning, and replanning in robotic operation; computer science supplies the algorithms.
Related originating lineages:
- Computer Science & Software Engineering — Route Replanning's terminology and operating form—discards a route that has become invalid and synthesizes a fresh whole route from the current position to the goal—are rooted most directly in computer science and software-engineering practice.
- Engineering & Design — Engineering design, reliability, and systems-safety practice supplies a parallel or contributing lineage for the mechanism's defining operation: discards a route that has become invalid and synthesizes a fresh whole route from the current position to the goal.
- Operations Research — Vehicle routing independently reoptimizes routes under disruption.
Review resolution: The blind reviewers disagreed on primary lineage (robotics_automation versus computer_science); authoritative or primary research supports robotics_automation as the best historical origin. Recomputing a collision-free route from the robot's current state after the prior path becomes invalid is canonical autonomous-motion planning. NASA mobile-robot navigation work directly joins localization, obstacle sensing, path planning, and replanning in robotic operation; computer science supplies the algorithms. The cited NASA, Mobile Robot Navigation directly supports the defining operation used in that choice. All independently supported contributing domains are retained without an arbitrary cap, while domain_reach=multi_domain records later applicability separately from provenance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
References¶
[1] D* ("Dynamic A", Anthony Stentz, 1994) and its successors compute a shortest path and then efficiently *repair it as the map changes — edges become blocked or costs update — rather than replanning from scratch each time. It is the canonical algorithmic answer to "the route just became invalid; give me a fresh one, fast." registry ↩