Motion and Behavior Planning Library

A modular motion planning library for autonomous vehicles, built with a scalable and extensible system architecture to support rapid integration, experimentation, and evaluation of diverse planning algorithms. The framework includes core autonomy subsystems such as environment simulation, agent modeling, motion primitive generation, trajectory prediction, vehicle dynamics propagation, and collision-aware trajectory validation. It aims to build a scalable codebase to enable eplxoration of Motion and Behavior Planning algorithms.

Example

System Architecture

The main modules include-

Environment

• Maintains world state
• Updates all agents
• Handle interaction between dynamic entities

Agent

Abstract representation of any world entity

• Static Obstacles
• Dynamic Obstacles

Agent Model

Defines state transition dynamics for an agent

• Bicycly model
• Learned models

Primitive Generator

Generate dynamically feasible candidate trajectories for the ego vehicle at given tinmestep

Collision Checker

Validate generated trajectories against the changing environment

Main Planner Loop

The planner operated in a receding-horizon loop:

1. Update environment
2. Generate candidate motion primitives
3. Collision-check trajectories
4. Select valid trajectory
5. Execute first segment
6. Replan

Build Instructions

mkdir build
cd build
cmake ..
cmake --build .

Run the program

plan.exe