ZJU-FAST-Lab/ego-planner-swarm
Ego Planner Swarm
Ego-planner-swarm is an autonomous drone swarm trajectory planner and decentralized multi-agent navigation system. It provides algorithmic software tools for calculating collision-free movement paths for single or multiple multicopters using decentralized onboard computing and local peer communication without relying on central ground stations or servers.
The software encompasses algorithmic components for drone motion planning, including asynchronous background replanning loops that adapt flight paths immediately to newly perceived obstacles or moving agents. Movement paths are represented mathematically as continuous piecewise polynomial curves that allow analytical derivatives for velocity and acceleration constraints, while initial safe graph routes are generated through cluttered environments using safe flight corridors extracted from volumetric distance fields. These paths are refined iteratively by combining smoothness penalties with repulsive potential fields derived directly from local occupancy map gradients, while accounting for the velocity, acceleration, and dynamic limits of multicopters.
The repository includes a simulation environment that models quadrotor physical dynamics, motor responses, and motion constraints in software, alongside hardware-accelerated sensor simulation that generates synthetic depth images and point clouds to mimic onboard perception.
Features
- Multi-Agent Trajectory Planners - Computing safe and collision-free movement paths for multiple autonomous robots operating simultaneously in crowded spaces.
- Spline Trajectory Generators - Represents smooth flight paths mathematically as continuous piecewise polynomial curves that allow instant analytical derivatives for velocity and acceleration constraints.
- Path Planning Algorithms - The robotics framework calculates collision-free movement vectors and safe flight paths concurrently for multiple autonomous aerial robots using decentralized onboard planning algorithms.
- Autonomous Drone Swarm Trajectory Planners - An algorithmic software tool that calculates collision-free movement paths for single or multiple multicopters using decentralized onboard computing.
- Decentralized Multi-Agent Navigation Systems - A navigation system that coordinates movement vectors for autonomous aerial vehicles simultaneously without central server coordination.
- Decentralized Swarm Optimization - Computes collision-free multi-agent paths independently onboard each vehicle using local communication and peer state sharing without a central coordinator.
- Gradient-Based Safe Trajectory Optimizers - Refines movement paths iteratively by combining smoothness penalties with repulsive potential fields derived directly from local occupancy map gradients.
- Aerial Navigation Simulations - Guiding groups of aerial vehicles through complex environments using decentralized algorithms without relying on central ground stations.
- Quadrotor Dynamics Simulators - The robotics framework calculates physical movement and response to motion commands either through simplified kinematic conversion or full differential equation dynamics modeling.
- Quadrotor Physics Simulators - Modeling the real-world flight dynamics, motor responses, and motion constraints of multicopter drones in software.
- Robotics Simulators - Simulating fleets of autonomous vehicles in shared virtual environments to test coordination algorithms before deploying them to real hardware.
- ROS Simulation Bridges - A simulation environment that models quadrotor physical dynamics and generates synthetic depth sensor data for testing robotic algorithms.
- Synthetic Sensor Simulation - The robotics framework generates synthetic depth images and point cloud measurements using graphics hardware to mimic real onboard sensor perception in testing environments.
- Kinodynamic State Machine Simulators - Models drone movement constraints directly within the path search phase by accounting for velocity, acceleration, and dynamic limits of multicopters.
- Topological Path Searchers - Generates initial safe graph routes through cluttered environments using safe flight corridors extracted from volumetric distance fields before numerical refinement.
- Asynchronous Recalibration Loops - Runs continuous replanning cycles in dedicated background threads to adapt flight paths immediately to newly perceived obstacles or moving agents.
