Efficient Large-Scale Multi-Drone Delivery using Transit Networks
Authors: Shushman Choudhury, Kiril Solovey, Mykel J. Kochenderfer, Marco Pavone
JAIR 2021 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the efficiency of our approach on simulations with up to 200 drones, 5000 packages, and transit networks with up to 8000 stops in San Francisco and the Washington DC Metropolitan Area. Our framework computes solutions for most settings within a few seconds on commodity hardware and enables drones to extend their effective range by a factor of nearly four using transit. |
| Researcher Affiliation | Academia | Shushman Choudhury EMAIL Department of Computer Science Stanford University, CA, 94035 USA Kiril Solovey EMAIL Mykel J. Kochenderfer EMAIL Marco Pavone EMAIL Department of Aeronautics and Astronautics Stanford University, CA, 94035 USA |
| Pseudocode | Yes | Algorithm 1: Merge Split Tours Algorithm 2: The multi-agent level of Enhanced CBS for MAPF-TN. |
| Open Source Code | Yes | 1. The code for our work is available at https://github.com/sisl/Multi Agent Allocation Transit.jl. |
| Open Datasets | Yes | We ran simulations with two large-scale public transit networks in San Francisco (SFMTA) and the Washington Metropolitan Area (WMATA). We used the open-source General Transit Feed Specification2 data for each network. 2. https://developers.google.com/transit/gtfs |
| Dataset Splits | No | The paper describes generating 'randomly generated locations' and running '100 trials' or '30 trials' for evaluation. However, it does not specify exact train/test/validation dataset splits or provide a random seed for reproducibility of these random generations, which is crucial for recreating the exact experimental scenarios in a machine learning context. |
| Hardware Specification | Yes | We implemented our approach in the Julia programming language for fast numerical simulations and tested it on a machine with a 6-core 3.7 GHz 16 Gi B RAM CPU. |
| Software Dependencies | No | The paper mentions 'Julia programming language' but does not specify a version number. It also refers to 'General Transit Feed Specification2 data' but this is a data format, not a software dependency with a version. |
| Experiment Setup | Yes | We set the drone flight range constraint conservatively to 7 km and the average speed to 25 kph, based on DJI Mavic 2 specifications.3 For the much larger WMATA area, we used a flight range of 10 km to enable more feasible solutions. ... We randomly sampled the integer carrying capacity of any transit edge C(e) from {3, 4, 5}, representing single and double buses, and set the suboptimality factor for ECBS to 1.1. |