This repository contains supplementary data and demonstration videos for the paper on SOPP: Sequence-Oriented Pickup-and-delivery Planning.
The algorithm addresses the challenge of coordinating multiple autonomous agents graph-based environments to efficiently complete pick-and-drop tasks while respecting specified delivery sequences. Each task includes:
- Source location: Where the item is picked up
- Destination location: Where the item is delivered
- Delivery sequence: Constraint on the order in which items must be delivered
- Time delays: Optional task release times for testing with asynchronous task arrival
- agent velocity is assumed 1 unit distance per 1 unit of time
A full-scale warehouse environment (based on warehouse-10-20-10-2-1 available on https://movingai.com/benchmarks/mapf/index.html) with 1000 autonomous agents coordinating to complete 5000 pick-and-drop tasks.
- JSON format: warehouse-10-20-10-2-1.json
- Map format: warehouse-10-20-10-2-1.map
Safe location is specified in node 1 (top-left corner).
- warehouse_10_20_10_2_1_tasks.m - Random task set for warehouse scenario
| Metric | Value |
|---|---|
| Makespan | 3,022.80 time units |
| Number of Agents | 1000 |
| Number of Tasks | 5000 |
| Total agents time | 1,679,776.80 units |
| Total Travel Distance | 946,098.00 units |
| Number of planning steps | 36,309 |
A complex maze environment (based on maze-32-32-2 available on https://movingai.com/benchmarks/mapf/index.html demonstrating agent coordination in constrained spaces.
- JSON format: maze-32-32-2.json
- Map format: maze-32-32-2.map
Safe location is specified in node 186 (dead-end on the right).
- maze_32_32_2_random_1.tasks.m - Random task set for maze scenario
| Metric | Value |
|---|---|
| Makespan | 695.95 time units |
| Number of Agents | 1000 |
| Number of Tasks | 333 |
| Total agents time | 44,919.45 units |
| Total Travel Distance | 38,180.00 units |
| Number of planning steps | 8,608 |
A customized compact warehouse environment with intentional choke point.
- JSON format: grid_map_4_tight.json
- Tasks.txt - 56 tasks with delivery sequences
The same compact warehouse environment as Scenario 3, but with staggered task arrival times (delays). This demonstrates the algorithm's ability to handle asynchronous task releases and dynamic task assignment.
Key Feature: Each task has a release time delay (ranging from 20 to 1120 time units), simulating real-world scenarios where tasks arrive at different times and no prior knowledge of task sequence is known.
- JSON format: grid_map_4_tight.json
- Tasks.txt - 56 tasks with varying arrival delays and delivery sequences