Walk in the Parkour Pathfinding

A pathfinding system that uses parkour-style movement to navigate complex environments in an engaging fashion, capable of obstacle avoidance.

A truly enjoyable pathfinding experience.

Detailed project page

Detailed

Problem Definition

Using simple randwalks or A* for platforming paths in a voxel environment led to unenjoyable or boring routes that didn't respect player engagement constraints.

Solution and Outcome

I simulated player jump constraints in sampled environments and used that state to generate procedural platforming paths in an efficient way. The resulting algorithm produced more interesting routes that respected jump feasibility and engagement considerations.

Overview of Initial Knowledge and Procedure

In essence, it is a server side plugin to generate procedural parkour. This is done by Utilizing the property of a minecraft jump that the player can travel 4 blocks horizontally from 1 block of space and 1.25 vertically to create parabola with an extra term to make it 3D. From this we obtain a volume that contains all possible jump points. We then pre calculate all possible jump points into a list for fast access so that we do not need to recalculate every time. When picking a new jump, we just sample from the list and add its local position to the last jump.

Generating and visualizing the jumps

Reach Peak Spread

Static sampler preview. Drag to orbit, scroll to zoom.

Project Overview

This project is a 1.21 spigot plugin which runs on a spigot server and is made using their API. It was developed using a custom IntelliJ run configuration so the server could run inside the IDE, and a hotswap agent dynamically rebuilt classes without restarting the server. This dramatically improved iteration speed and made implementing pathfinding behavior much faster.

Commands Singletons Types Util Resources Class Overviews What I Learned

Commands

Player-run commands in game.

NextJumpCommand
PlaceBlockCommand
RandomJumpsCommand
StartParkourCommand
TestBlocksCommand
TestChunkHighestHeightCommand

Singletons

This was one of the key concepts learned during development.

JumpDataManager
JumpLists

Types

Fully instantiated objects to extend in future iterations.

ParkourJump
Checkpoint

Util

Helper classes used across systems.

BlockRemover
BlockSummoner
ChunkAnalyzer
JumpListFilters
JumpListGenerator
LocationTools
VectorTools
Parkour

Resources

jumps.json

Class Overviews

NextJumpCommand.java: A debug command that picks a random jump and places it relative to the player using BlockSummoner.java.
PlaceBlockCommand.java: A debug command that places a block at the player with animation using BlockSummoner.java.
RandomJumpsCommand.java: The current main game implementation, selecting random jumps from list instances supplied by JumpDataManager.java.
StartParkourCommand.java: The eventual primary implementation intended to fully execute the pathfinding algorithm.
TestBlocksCommand.java: A debug command that places all blocks at a chosen Y level from the jump list, relative to the player.
TestChunkHighestHeightCommand.java: Teleports the player to the highest block in the current chunk via methods from ChunkAnalyzer.java.
JumpDataManager.java: Singleton access point for jump-list instances so systems can share data without expensive re-instantiation.
JumpLists.java: Provides pre-sorted jump lists generated from source data for pathfinding logic.
ParkourJump.java: Main jump data object; designed for future feature extension.
Checkpoint.java: Early checkpoint and death-detection support.

BlockRemover.java: Animates block removal by translating and transforming downward.
BlockSummoner.java: Animates block placement by transforming from an offset position into world position.
ChunkAnalyzer.java: Provides methods to evaluate chunk height characteristics and highest points.
JumpListFilters.java: Filters jump candidates using heuristics from the pathfinding algorithm.
JumpListGenerator.java: Recursively builds jump sequences across n pathfinding iterations using prior outputs as seeds.
LocationTools.java: Utility methods for Spigot Location handling.
VectorTools.java: Vector math helpers used by the pathfinding calculations.
Parkour.java: Main plugin class; registers commands and hosts continuous-loop behavior.
jumps.json: Static jump data generated by a Python script and loaded by the plugin.

What I Learned

General Concepts

Organizing large projects into sensible packages and ownership boundaries.
Planning features and controlling scope creep.
Breaking complex systems into smaller, testable components.
Debugging across multiple classes and async workflows.
Singletons, static vs non-static objects, and why object cloning matters.
IntelliJ run configs and hotswap workflow for rapid iteration.
Core linear algebra concepts including atan2.
JSON import, arrays, hashes, iterators, enhanced loops, and package design.
Input streams and try/catch error handling patterns.

API Specific

Building a Spigot plugin and registering commands.
Using interfaces and command contracts.
Using sound and block enums effectively.
Using Spigot scheduler tasks for asynchronous behavior.
Working with World, Player, Location, Block, and Chunk.

Tags Robotics Forestry platform Virtual pivot steering

Robotics Forestry platform Virtual pivot steering Low-impact pathing Adjustable bogey suspension

Skills Mechanical engineering Iterative design Systems testing

Mechanical engineering Automotive steering geometry Prototype instrumentation Integration testing Engineering communication