DoorDash Deploys Multi-Agent Reinforcement Learning to Optimize Food Delivery Dispatch
Researchers at DoorDash have published a paper describing a deployed reinforcement learning system that dynamically adjusts dispatch optimization weights in their food-delivery marketplace. The system uses delayed operational signals—such as delivery speed, courier utilization, and merchant congestion—to train store-level policies that shift tradeoffs between delivery quality and batching efficiency. The work demonstrates a practical approach to safely applying offline-trained RL policies in a live, large-scale economic and logistics environment.
A team from DoorDash has detailed a production reinforcement learning system accepted at the ICML 2026 Workshop on Reinforcement Learning from World Feedback. Rather than replacing the existing combinatorial assignment optimizer, the system introduces a store-level policy that selects discrete multipliers to adjust the optimizer's balance between delivery quality and batching efficiency. The policy is trained offline using logged marketplace data and employs Double Q-learning with a conservative regularizer to mitigate out-of-distribution value overestimation—a common challenge in offline RL. Execution is decentralized at the store level while training uses centralized data, fitting a multi-agent framework. In a production switchback experiment, the offline-trained policy successfully increased batching rates and reduced courier-side time costs without degrading customer-facing delivery metrics. The paper highlights how delayed, noisy, and coupled real-world feedback from a live marketplace can be used to safely adapt operational decision policies.
What's missing
The paper does not appear to disclose the scale of the switchback experiment (e.g., number of stores, geographic markets, or duration), nor does it quantify the magnitude of improvements in batching rates or courier time costs with specific figures in the abstract. Long-term stability of the deployed policy under distribution shift also remains an open question.
What different sources said
- arXiv cs.AICenter
Multi-Agent Reinforcement Learning from Delayed Marketplace Feedback for Objective-Weight Adaptation in Three-Sided Dispatch
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.