AI-Assisted Climbing Level Design
Research Prototype — In Progress
Formalizing climbing-level design rules so a solver can validate, compare, and eventually generate playable routes.
Project Overview
This project explores how climbing-level design knowledge can be formalized into computational rules that allow an AI system to reason about playable routes.
The prototype represents climbing poses as structured states, evaluates reachability, center-of-mass stability, directional hold constraints, stamina, and limb transitions, then searches for valid routes through a designer-authored wall.
The current focus is level validation and route evaluation. Automatic hold generation is a planned next stage.
Research Question
How can climbing-level design knowledge—reachability, body stability, hold direction, stamina, and route difficulty—be formalized so an AI system can evaluate and eventually generate playable climbing routes?
Playable Prototype
The prototype allows climbing poses to be tested directly in Unity. Limbs can move between reachable holds while the system tracks support configuration, body position, stamina, and hold-specific constraints.
A playable movement test, alongside the solver’s state and route analysis below.
System Overview
Input holds / wall setup → State generation → Transition solver → Route validation / evaluation.
- Input Holds / Wall Setup
- Define regular holds, directional sidepull holds, start holds, foot holds, and finish conditions.
- State Generation
- Generate candidate hand-foot support states from reachable hold combinations.
- Transition Solver
- Evaluate limb transitions using reach limits, center-of-mass sampling, and directional stability constraints.
- Route Validation / Evaluation
- Search from the start state to the finish and compare feasible solutions using route-level heuristics.
Player State & Movement Rules
Each pose is a structured player state, beyond a character animation: left-hand, right-hand, left-foot, and right-foot contacts, body position, center of mass, stamina, and sidepull constraints. Each transition moves one limb to another reachable hold.
- Four limb contacts
- Each hand and foot can occupy a reachable hold or remain unsupported.
- Center of mass
- Body position is evaluated relative to the support configuration and contributes to pose stability.
- Stamina
- Body movement and active movement consume stamina. Supported poses can recover stamina.
- Hold properties
- Regular holds primarily constrain reach; sidepull holds add directional stability constraints.
- Support requirement
- Valid supported climbing generally requires at least one hand and one foot to remain attached.
- Reachability
- A limb may only transition to holds within its reachable movement range.
- Goal state
- Reach the designated finish through a valid sequence of states and transitions.
Directional Stability Visualization
Sidepulls constrain where the climber’s body can remain mechanically stable. The prototype samples candidate center-of-mass regions and evaluates stable, weak, strained, or invalid regions.
Transition Solver / Hold Reachability Validator
The validator enumerates candidate support states, tests hand and foot transitions, rejects configurations that violate reach or stability constraints, and searches for a complete path from the starting configuration to the finish.
- Input
- Designer-authored holds and support rules.
- Search
- Candidate support states and limb transitions.
- Validation
- Reachability, geometry, center-of-mass stability, and sidepull constraints.
- Output
- Reachable states, valid transitions, solvable routes, and route-level difficulty estimates.
Route Evaluation
Once a solvable route is found, alternative solutions can be compared using movement difficulty, pose stability, reach demand, directional hold strain, and route length.
The current prototype can identify the shortest feasible route and an estimated easiest route. These estimates are heuristic, are still being refined, and have not yet been validated against human player data.
Current Status
The project currently formalizes climbing-level rules and evaluates whether a designer-authored wall is traversable.
- Represent poses as structured support states and generate candidate hand-foot configurations from placed holds.
- Evaluate limb reachability and sidepull constraints; sample feasible center-of-mass regions.
- Search valid limb transitions and determine whether the finish is reachable from the start configuration.
- Compare valid routes with heuristics, including the shortest feasible and estimated easiest routes.
Hold placement is still primarily designer-authored. This is a level validation and analysis tool; automatic generation is a future stage.
What's Next
- Automatic hold generation
- Generate candidate layouts and use the solver to reject unsolvable or undesirable configurations.
- Designer-controlled constraints
- Specify target difficulty, route length, hold types, required mechanics, or the desired number of alternative routes.
- Improved difficulty modeling
- Refine heuristics using reach, pose stability, sidepull strain, stamina demand, and route-level statistics.
- Player modeling
- Compare routes for different simulated player capabilities or preferences rather than one idealized climber.
- Explainable design feedback
- Show why a route is difficult, unstable, unreachable, or overly constrained, beyond a valid / invalid result.
The long-term goal is not to replace the level designer, but to build an AI-assisted design tool that can generate, validate, compare, and explain climbing routes while preserving designer intent.