Redirecting to original paper in 30 seconds...

Click below to go immediately or wait for automatic redirect

arxiv_ai 90% Match Research Paper Robotics Engineers,Aerospace Engineers,AI Researchers,UAV Developers 1 week ago

Navigation in a Three-Dimensional Urban Flow using Deep Reinforcement Learning

reinforcement-learning › robotics-rl
📄 Abstract

Abstract: Unmanned Aerial Vehicles (UAVs) are increasingly populating urban areas for delivery and surveillance purposes. In this work, we develop an optimal navigation strategy based on Deep Reinforcement Learning. The environment is represented by a three-dimensional high-fidelity simulation of an urban flow, characterized by turbulence and recirculation zones. The algorithm presented here is a flow-aware Proximal Policy Optimization (PPO) combined with a Gated Transformer eXtra Large (GTrXL) architecture, giving the agent richer information about the turbulent flow field in which it navigates. The results are compared with a PPO+GTrXL without the secondary prediction tasks, a PPO combined with Long Short Term Memory (LSTM) cells and a traditional navigation algorithm. The obtained results show a significant increase in the success rate (SR) and a lower crash rate (CR) compared to a PPO+LSTM, PPO+GTrXL and the classical Zermelo's navigation algorithm, paving the way to a completely reimagined UAV landscape in complex urban environments.
Authors (2)
Federica Tonti
Ricardo Vinuesa
Submitted
October 29, 2025
arXiv Category
cs.AI
arXiv PDF

Key Contributions

Develops an optimal navigation strategy for UAVs in 3D urban flows using flow-aware PPO with a GTrXL architecture. This approach incorporates richer information about turbulent flow fields, leading to improved navigation performance.

Business Value

Enhances the safety and efficiency of autonomous drone operations in urban areas, enabling reliable delivery, surveillance, and inspection services.

Paper Metadata

Innovation Type

Algorithmic

Deployment Feasibility

Moderate, requires high-fidelity simulation environments and robust onboard processing capabilities.

Limitations Addressed

Challenges of UAV navigation in complex, turbulent urban environments with unpredictable flow dynamics.

Performance Gains

Significant increase in success rate and lower crash rate compared to PPO+LSTM, PPO+GTrXL without secondary tasks, and classical navigation algorithms.

Technical Tags

Deep Reinforcement LearningUAV NavigationUrban FlowProximal Policy OptimizationGated Transformer eXtra LargeFlow-awareTurbulenceRecirculation Zones3D SimulationLSTM

Research Topics

Reinforcement LearningRoboticsAutonomous SystemsAerospaceFluid Dynamics

Methods & Architectures

Flow-aware Proximal Policy Optimization (PPO)Gated Transformer eXtra Large (GTrXL) architectureSecondary prediction tasks Proximal Policy Optimization (PPO)Gated Transformer eXtra Large (GTrXL)Long Short Term Memory (LSTM)

Applications & Tasks

Unmanned Aerial Vehicles (UAVs) Urban Navigation Autonomous Flight Optimal navigation in turbulent urban environmentsHandling complex 3D flow dynamicsImproving UAV safety and success rates Autonomous navigationPath planningCollision avoidance

Datasets & Benchmarks

Benchmarks
Success Rate (SR) • Crash Rate (CR)
Success Rate (SR)Crash Rate (CR)

Related Fields

Aerospace EngineeringRoboticsComputational Fluid Dynamics

Keywords

UAVNavigationReinforcement LearningUrban FlowPPOGTrXLTurbulenceAutonomous Systems3D NavigationRobotics

Academic Context

#Reinforcement Learning#Robotics#Autonomous Systems#Aerospace#Fluid Dynamics

Commercial Potential

Potential Products
Autonomous drone navigation systemsUAV fleet management softwareUrban air mobility solutions
Target Industries
LogisticsDelivery ServicesSurveillanceInfrastructure InspectionUrban Air Mobility
Use Case Examples
Autonomous drone delivery in citiesAerial surveillance of urban areasInspection of bridges and buildings
Competitive Edge
Outperforms existing RL and classical methods for UAV navigation in complex urban flow environments.
Market Opportunity
Rapidly growing market for commercial drones and autonomous flight.
Revenue Models
Licensing of navigation softwaredevelopment of autonomous drone services.

Resource Requirements

Compute Needs
High, for training deep RL agents in complex simulations.
Data Requirements
Requires high-fidelity 3D urban flow simulation environments.
Deployment Constraints
Computational limitations on UAVs, real-world turbulence unpredictability.
Scalability
Scalability depends on the efficiency of the RL algorithm and the simulation environment.
Regulatory Considerations
Regulations regarding UAV operation in urban airspace.

Production Readiness

Maturity Level

Research/Development

Time to Market

3-5 years for widespread commercial deployment.

Patent Potential

Moderate, for the flow-aware RL approach and GTrXL integration.

View Full Paper Back to Papers