TriDeliver: Cooperative Air-Ground Instant Delivery with UAVs, Couriers, and Crowdsourced Ground Vehicles

arXiv:2604.09049 · cs.RO, cs.HC · Submitted 2026-04-10 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: Today's paper: "TriDeliver: Cooperative Air-Ground Instant Delivery with UAVs, Couriers, and Crowdsourced Ground Vehicles".

Dev: Instant delivery, shipping items before critical deadlines, is essential in daily life.

Rosa: First, who's behind it and why it matters.

Title and authors: Rosa: Let's talk about who wrote this piece and what they're calling their system "TriDeliver: Cooperative Air-Ground Instant Delivery with UAVs, Couriers, and Crowdsourced Ground Vehicles." It sounds like a pretty comprehensive name for such an integrated setup.

Dev: The team includes Junhui Gao, Yan Pan, Qianru Wang, Wenzhe Hou, Yiqin Deng, Liangliang Jiang, and Yuguang Fang as the authors of this work. I'm interested in seeing how their specific expertise in robotics and control translates into a practical system like this.

Taro: From an autonomy standpoint, the title suggests they are aiming for a holistic solution by integrating air, ground human power, and ground vehicle resources into one cooperative framework.

Rosa: It really emphasizes the cooperative nature of the work; it’s not just about optimizing one agent in isolation but making sure these three distinct agents complement each other's strengths for instant delivery.

Dev: The implication here is that they are trying to solve a complex scheduling problem where you need to coordinate different operational constraints and capabilities across multiple platforms simultaneously.

Taro: That complexity is definitely there, especially when considering the city-scale parcel assignment which the paper points out can be an NP-hard problem when you have three cooperating agents involved.

The paper's summary: Rosa: The summary of "TriDeliver: Cooperative Air-Ground Instant Delivery with UAVs, Couriers, and Crowdsourced Ground Vehicles" shows that the main goal is to achieve efficient instant delivery by integrating these three agents through a hierarchical cooperative framework.

Dev: Essentially, they designed a Transfer Learning based algorithm to extract knowledge from courier behavior—their historical delivery patterns—and then transfer that learned knowledge over to both the UAVs and the crowdsourced ground vehicles with some fine-tuning.

Taro: So, the process starts by modeling what couriers do, like their preferences and decision functions, and then using those learned models to guide how the UAVs and GVs operate for parcel assignment.

Rosa: That knowledge transfer is key because it lets the autonomous agents benefit from the rich experience of human delivery agents without having to learn everything from scratch in a completely new environment.

Dev: The paper also details how they handle the remaining parcels that don't fit neatly into those models by solving an optimization problem, specifically formulating it as a Generalized Assignment Problem with Assignment Restriction, or GAPAR.

Taro: That GAPAR part suggests they have a system in place to manage the residual tasks intelligently once the preferred assignments are made by the learned models.

The paper's improvements: Rosa: The paper outlines several ways they improve upon prior approaches, focusing on how this cooperative structure handles things like ground traffic congestion, which is a big concern for both couriers and GVs.

Dev: They specifically mention that UAVs are used to bypass existing ground traffic jams to ensure urgent parcels get delivered rapidly, which minimizes operational costs while boosting efficiency.

Taro: I noticed the authors also focus on how this system improves the impact on original tasks for crowdsourced GVs, showing a significant reduction in negative externalities when integrated into this cooperative model.

Rosa: That's interesting because it suggests that by coordinating everything hierarchically, they can reduce the adverse effects on things like taxi passenger experience by a substantial amount, which is important for real-world deployment.

Dev: They quantify these improvements quite well; for instance, they report reducing delivery cost by sixty-five point eight percent compared to state-of-the-art cooperative delivery methods involving UAVs and couriers.

Conclusion: Rosa: Wrapping up the discussion on "TriDeliver: Cooperative Air-Ground Instant Delivery with UAVs, Couriers, and Crowdsourced Ground Vehicles," the main implication is that this hierarchical cooperative framework offers a way to achieve substantial efficiency gains by learning from human behavior.

Dev: The paper demonstrates that by transferring knowledge from couriers to UAVs and GVs, they can significantly lower delivery costs and improve time reliability compared to previous methods.

Taro: What really stands out is how the system handles the uncertainty inherent in city-scale parcel assignment, moving it from a purely hard problem to one where learned models provide strong initial scheduling knowledge.

Rosa: So, when we look at the future work, they seem focused on scaling this up and testing it outside of controlled lab environments to see how robust this cooperative model is in a messy real world.

Dev: I'm keen to know how the loop rate performs when these different agents interact dynamically in a live setting, because that's where we need to ensure the system doesn't introduce unacceptable latency or failure modes.

Taro: And I think testing its robustness against unexpected disruptions, like sudden environmental changes or unpredictable demand spikes, will be crucial for proving its real-world applicability beyond the tested scenarios.

City University of Hong Kong

cs.RO, cs.HC

Submitted: 2026-04-10

Updated: 2026-10-07

Code: https://github.com/cbdog94/STL

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 74/100

The gist: Instant delivery, shipping items before critical deadlines, is essential in daily life.

Key concepts

Transfer Learning: This is an AI technique where knowledge learned from one task (courier behavior) is used to improve performance on a different, related task (UAV and GV routing). The system learns how couriers decide what to do based on their history and then adapts that decision-making logic for the other delivery agents.
Courier Preference Extraction: This step involves collecting detailed historical data about how human couriers make choices, such as ordering times, detour distances, and velocity. This data is used to train a shared network that learns the courier's decision function, fC(x), which predicts the best delivery action based on the current situation.
Generalized Assignment Problem with Assignment Restriction (GAPAR): This mathematical optimization problem is used to assign any remaining parcels that none of the three delivery models (UAV, Courier, GV) prefer. It ensures that every unassigned parcel is allocated to one of the agents while respecting specific constraints related to their capabilities and limitations.
Delivery Models: These are the specific rules defining how each agent operates. For instance, UAVs have energy consumption models and path planning rules, couriers have payload limits, and GVs have different modes like OD-pair delivery or halfway delivery, all governed by defined time and distance constraints.

Terminology

Fine-Tuning — 12× in this paper · explained in 5 episodes
Transfer Learning — 12× in this paper · explained in 10 episodes
Cooperation — 9× in this paper · explained in 2 episodes
Ablation — 3× in this paper · explained in 1 episode
Ablation study — 2× in this paper · explained in 1 episode
Activation Function — 2× in this paper · explained in 1 episode
Adam — 2× in this paper · explained in 2 episodes
Core — 2× in this paper · explained in 2 episodes
Generalized Assignment Problem with Assignment Restriction — 2× in this paper
International Conference on Data Engineering — 2× in this paper
Robustness — 2× in this paper · explained in 5 episodes
Consistency — 1× in this paper · explained in 3 episodes
Energy Efficiency — 1× in this paper · explained in 3 episodes
Failure Modes — 1× in this paper · explained in 3 episodes
Hybrid Approach — 1× in this paper · explained in 4 episodes
Loss Function — 1× in this paper · explained in 2 episodes
Mobile Edge Computing — 1× in this paper · explained in 2 episodes
Q-learning — 1× in this paper · explained in 3 episodes
Reinforcement Learning — 1× in this paper · explained in 4 episodes

Summary

Instant delivery, shipping items before critical deadlines, is essential in daily life. TriDeliver proposes the first hierarchical cooperative framework integrating human couriers, Unmanned Aerial Vehicles (UAVs), and crowdsourced ground vehicles (GVs) to achieve efficient instant delivery by leveraging a Transfer Learning-based algorithm to extract knowledge from courier behavior and transfer it to UAVs and GVs.

The gist

TriDeliver reduces the delivery cost by 65.8% versus state-of-the-art cooperative delivery by UAVs and couriers, while further improving performance in terms of delivery time (−17.7%), delivery cost (−9.8%), and impacts on original tasks of crowdsourced GVs (−43.6%).

Framework Overview

The proposed TriDeliver system integrates three distinct agents: dedicated UAVs, human couriers, and crowdsourced GVs to achieve efficient instant delivery by strategically leveraging their complementary strengths. The workflow involves the following steps:

Courier Preference Extraction and Model Training (Sec. V-A): Courier delivery preferences are extracted from historical data to train a model representing courier decision functions, denoted as fC(x).
Knowledge Transfer and Fine-tuning for GVs and UAVs (Sec. V-B): This learned model is transferred to the delivery models of UAVs and GVs with fine-tunings, respectively.
Remaining Parcel Assignment (Sec. V-C): Parcels not preferred by any of the three models are assigned by solving an optimization problem, formulated as a Generalized Assignment Problem with Assignment Restriction (GAPAR).

Delivery Models

The paper defines specific delivery models for each agent:

(a) UAV Delivery Model:

UAVs' capabilities involve departing from charging stations to restaurants, delivering parcels while searching for others, and returning to charging stations. Path planning utilizes a sampling-based algorithm to avoid collisions and no-fly zones. Energy consumption is modeled by the equation:

∆E(u) = X Σ I−1 i=0 σ(w(i)) × DistU (l(i), l(i + 1)) v(u), where σ(w) denotes the energy consumption rate.

The delivery time limit for a parcel is defined by Eq.(3): t(p) ≤ ∆t, ∀p ∈ P(u).

(b) Courier Delivery Model:

Couriers follow a path involving picking up and dropping off parcels, limited by payload capacity (Eq. 5: n(i) ≤ nmax) and delivery time limits (Eq. 6: t(p) ≤ ∆t). The delivery cost is determined by the distance traveled:

s(c, p) = 3.15 × DistC (lo(p), ls(p)).

(c) GV Delivery Model:

GVs participate in three cases: Origin-Destination (OD)-pair Delivery, Halfway Delivery, and Unoccupied Delivery. OD-pair delivery requires a spatial limit d(b, p) ≤ dmax and a temporal limit tpu(b, p) ≤ ∆tpu. The cost for OD-pair delivery is defined as:

s(b, p) = 2 × 2.7 × (d(b, p) + d′(b, p)).

Transfer Learning Strategy

The core innovation lies in the Transfer Learning (TL)-based algorithm designed to extract knowledge from couriers’ behavioral history and transfer it to UAVs and GVs with fine-tunings.

Courier Preference Extraction: Features such as ordering time, restaurant location, detour distance, velocity, delivery cost, payload, and remaining delivery time are collected into the feature space XC.
Decision Function Training: A network (Shared Network) is trained using Binary Cross Entropy Loss to represent the courier's decision function fC(x).
Knowledge Transfer and Fine-tuning: The model is transferred to GVs and UAVs, where feature spaces XB and XU are defined by replacing courier payload indicators with GV status or UAV carrying weight, respectively. Parameters are updated using optimization techniques (Eq. 16 and 17) based on the loss function L to minimize deviation from true labels YB and YU.

Remaining Parcel Assignment

Parcels not assigned based on the preference models (PC, PB, PU) form the set P'.

Improvements for AI systems

As a fastidious and diligent researcher, I have analyzed the TriDeliver framework, which integrates human couriers, UAVs, and crowdsourced Ground Vehicles (GVs) using a Transfer Learning (TL)-based approach for cooperative parcel assignment.

Here are the specific improvements that can be made to AI systems based on this paper:

The core improvement lies in developing a unified, intelligent dispatch system that moves beyond simple heuristic optimization by leveraging learned human behavior.

A sophisticated, multi-modal Decision Engine capable of real-time task allocation across heterogeneous agents (UAVs, Couriers, GVs).
A robust knowledge transfer mechanism (Transfer Learning) that allows the wisdom of human couriers to be effectively distilled and applied to the decision-making models of autonomous systems (UAVs and GVs).

The improved AI system will be a tri-modal cooperative delivery manager with the following capabilities:

A single, optimized dispatch service that intelligently routes parcels by selecting the best agent based on learned preferences, maximizing fleet utilization while respecting dynamic constraints.
Enhanced UAV path planning and energy management that is informed by predicted demand patterns derived from historical courier behavioral data (e.g., knowing which areas/times couriers prefer).
Adaptive Ground Vehicle (GV) task assignment that dynamically adjusts its participation in Unoccupied Delivery or Halfway Delivery based on real-time traffic conditions and predicted demand for its specific service type, ensuring stable supply during peak hours without over-provisioning.
A dynamic pricing/reward system that optimizes the trade-off between maximizing delivery speed and minimizing the negative impact on ground agents (couriers/drivers), leading to lower overall operational costs for the delivery company while maintaining high service quality for customers.

In summary, this AI system will transform parcel logistics from a fragmented approach into a synergistic ecosystem where every agent understands the needs of its counterparts, resulting in:

A reduction in delivery cost (up to 65.8%).
Significant improvement in delivery time reliability (up to 17.7% faster).
Minimization of negative externalities on ground agents (e.g., reduced adverse impacts on taxi passenger experience by up to 43.6%).

Sources

Adam: A Method for Stochastic Optimization