This project explores decentralized path-planning and coordination algorithms tailored for multi-agent, small-scale swarm robotics. Instead of relying on a centralized controller that introduces a single point of failure and high processing overhead, this architecture empowers individual agents to execute localized grid-based pathfinding dynamically. The system is designed around discrete grid synchronization, enabling scalable multi-robot routing without large performance spikes.

Core Architectural Components

The system's execution pipeline is split into explicit roles to optimize how agents perceive obstacles, interact with payload nodes, and resolve intersections:

• Track Robot Node

  Real-time kinematic path tracing where agents navigate an intersection network using local map frames. Each agent handles its own localization vector relative to its neighborhood.

• Solid Cargo Box (Task Allocation)

  Decentralized item-to-agent pairing mechanics. Agents identify static coordinates representing tasks, negotiate entry metrics with adjacent units, and schedule physical retrieval patterns without an over-arching central dispatcher.

• Line Target Goal

  End-point convergence management. Multi-agent streams converge onto designated workspace target points while dynamically updating cross-path priorities to avoid deadlock configurations.

A* Algorithm Visualization

Following shows how the A* algorithm works and how its further implemented to multiple robots.

Conflict Resolution & Localized Grid Consensus

A major focus of this decentralized setup is solving intersection deadlocks. Utilizing an interactive Simulation Monitor interface, agents share their path itineraries with nearby neighbors rather than broadcasting to the whole network.

When two agent trajectories cross on the coordinate plane, the units utilize localized intersection worker modes to stagger their movements seamlessly. This method optimizes routing efficiency and drastically minimizes total wait-times during simultaneous crossings.

A simpler visualization for understanding

Runtime Monitoring & Optimization Metrics

To analyze the efficiency of the underlying pathfinding matrix, the system features a live tracking engine that monitors multi-agent performance using key metrics:

• Fitness Scores

  A performance metric calculated per agent based on path efficiency, energy usage, and successful goal navigation. High-efficiency paths retain a higher fitness ceiling.

• Kinematic Resource Constraints

  The software continuously tracks individual agent Moves, Waits, and Turns. Measuring wait states provides direct feedback on the pathfinder's efficiency, allowing for iterative tuning of local prioritization behaviors.