Drones are enhancing delivery efficiency and energy conservation through advanced online learning techniques. The Department of Computer Science at TU Darmstadt has pioneered a strategy utilizing a multi-agent simulator to achieve superior delivery success rates and reduced delivery times compared to traditional models.
This innovative approach enables drones to predict when they need recharging, thereby improving resource planning and benefiting mixed fleets with diverse energy profiles. This leads to more autonomous and energy-efficient delivery systems. The researchers, led by Professor Roderich Groß, will present their study, titled 'Ready, Bid, Go! On-Demand Delivery Using Fleets of Drones with Unknown, Heterogeneous Energy Storage Constraints,' at the 24th International Conference on Autonomous Agents and Multiagent Systems in Detroit on May 21, 2025. The study has also been selected as a finalist for the Best Paper Award.
The system employs an auction-based method where drones bid on tasks based on their battery level and confidence in completing the delivery. Counterintuitively, the drone with the least confidence is selected, refining its understanding of its capabilities and promoting smarter resource allocation. This paves the way for more reliable and energy-optimized autonomous delivery systems.