VehiclePrimitives
Personal project | ongoing
Overview
A library of motion primitives for vehicle navigation and control - short, pre-computed maneuver segments (e.g. cubic spiral paths, speed-matched transitions) that a planner can stitch together into a full trajectory instead of solving the whole path from scratch every time. Built to generate feasible trajectories for cars, trucks, and autonomous robots.
MPTreeRRT
The primitives feed an MPTreeRRT planner: a sampling-based motion planner that explores the primitive space the way an RRT explores continuous state space, built on the sampling-based approach from the MPTree publication (MPTree: A Sampling-based Vehicle Motion Planner for Real-time Obstacle Avoidance, IFAC-PapersOnLine 2024). Rather than sampling raw control inputs, it samples from the library of dynamically-feasible primitives, which keeps every candidate trajectory kinematically valid by construction.
Status
Public C++ library with Python bindings, ongoing personal project.