RobotSchedule

Generating Schedules for Robotic Workloads

Installation

Install the package in editable mode:

pip install -e .

Examples

The repository includes several example scripts in src/scripts/:

Basic Scheduling Examples

testing.py - Parameter sweep for scheduling workloads with transfer times

• Performs a parameter sweep over different numbers of jobs and machines
• Generates synthetic sequential workloads with transfer times
• Saves runtime results to CSV and schedule plots to plots/ directory
• Run: python src/scripts/testing.py

testing_simple.py - Simple scheduling without transfer times

• Demonstrates scheduling with zero transfer times (simplified problem)
• Creates a workload with 4 sequential jobs on 3 machines (CPU, GPU, FPGA)
• Saves plot to plots/simple_test.png
• Run: python src/scripts/testing_simple.py

Advanced Scheduling Examples

testing_iree.py - Schedule IREE dispatch graphs on dual-core device

• Schedules real neural network dispatch graphs (Fast, Dronet, and 5 MLP instances)
• Models a dual-core device with CPU_P (performant, 1.5x faster) and CPU_E (efficient)
• Creates dependency chains: Fast → Dronet, and MLP0 → MLP1 → MLP2 → MLP3 → MLP4
• All MLP instances share the same color in the visualization
• Saves plot to plots/iree_combined_schedule.png
• Run: python src/scripts/testing_iree.py

packing_demo.py - Demonstrates greedy vs. convex packing algorithms

• Compares greedy and convex packing approaches for workload scheduling
• Shows trade-offs between solution quality and computation time
• Saves plots to plots/greedy_schedule.pdf and plots/convex_schedule.pdf
• Run: python src/scripts/packing_demo.py

additional_obj_demo.py - Scheduling with additional objectives

• Demonstrates scheduling with nominal start times and periodic constraints
• Useful for real-time scheduling scenarios
• Saves plot to plots/additional_objectives_schedule.png
• Run: python src/scripts/additional_obj_demo.py

Output

Most examples generate:

• Schedule visualizations saved to plots/ directory
• Runtime statistics printed to console
• Some examples also save CSV files with performance metrics