Reinforcement Learning for Multi-Robot Path Planning

This research project investigates multi-agent reinforcement learning (MARL) for coordinating a fleet of mobile robots navigating a shared warehouse-like environment while avoiding collisions and minimizing total travel time. A simulated grid-world environment models robots, static obstacles, and dynamic congestion.

A decentralized Deep Q-Network (DQN) policy is trained per agent with a shared reward structure that penalizes collisions and rewards reaching goal cells efficiently, and the study compares this against a centralized planner baseline (A* with reservation-table collision avoidance) on metrics such as makespan, path length overhead, and deadlock frequency as fleet size scales.

The paper further examines how communication range between nearby robots affects coordination quality, and discusses the scalability limits of the trained policies as the number of agents increases, concluding with recommendations for hybrid centralized-decentralized architectures in real warehouse robotics deployments.

Detailed Project Description

Reinforcement Learning for Multi-Robot Path Planning demonstrates closed-loop motion control: sensor input is turned into motor commands so the machine moves precisely and safely.

How It Works
1. Motors/servos are driven through a driver board from the controller's PWM outputs.
2. Commands come from a remote (Wi-Fi/BLE/RF/gesture) or from an autonomous algorithm.
3. Sensors (IMU, encoders, distance, GPS, LiDAR) provide feedback for stabilisation and obstacle avoidance.
4. A control loop adjusts motor outputs several times per second.

Technology Roles
- Python
- PyTorch
- Reinforcement Learning
- Multi-Agent Systems
- Robotics Simulation

Implementation Steps
1. Collect the components and study each module's datasheet (pin-out, voltage, interface).
2. Wire the circuit on a breadboard following the pin mapping; power sensors from the correct rail (3.3 V / 5 V).
3. Install Python, PyTorch, OpenAI Gym / PettingZoo and the required libraries; flash a minimal test sketch for each module separately.
4. Write the main firmware: read sensors → process → drive outputs / send data.
5. Configure network or cloud credentials and test communication end-to-end.
6. Calibrate thresholds, add error handling (reconnect, sensor time-out) and run a long-duration test.
7. Finalise on a PCB / enclosure and document wiring, code and results.

Key Features
- Remote or autonomous control
- Sensor feedback
- Obstacle handling
- Adjustable speed / precision

Expected Output
On completion the student will have a machine that follows the commanded path or task reliably, with telemetry visible to the operator.

Deliverables
- Complete source code with comments
- Circuit diagram and pin mapping
- Project report with screenshots and results
- Demonstration-ready build