Technology Category
Search and Optimization: Learning to SearchPlanning, Routing, and Scheduling: Planning with Language ModelsIntelligent Robots: Learning & Optimization for ROB
Transformer-Guided Deep Reinforcement Learning for Optimal Takeoff Trajectory Design of an eVTOL Drone
🤖 AI Summary
This paper addresses the minimum-energy takeoff trajectory optimization problem for electric vertical takeoff and landing (eVTOL) unmanned aerial vehicles. We propose a Transformer-guided deep reinforcement learning (DRL) framework that leverages the Transformer’s dynamic attention mechanism to adaptively prioritize critical state dimensions during policy search, thereby enhancing state-space exploration efficiency and improving policy convergence quality. The method jointly optimizes power allocation and wing-tilt angle control while rigorously satisfying key takeoff constraints—including minimum vertical displacement and prescribed horizontal velocity. Experimental results demonstrate that, compared to conventional DRL approaches, our method reduces training steps by 75%, achieving convergence in only 4.57×10⁶ steps. Moreover, the optimized trajectory attains 97.2% of the energy efficiency of a high-fidelity simulation benchmark. These results validate the proposed method’s superior convergence speed, control accuracy, and physical feasibility.
Technology Category
Application Category
📝 Abstract
The rapid advancement of electric vertical take-off and landing (eVTOL) aircraft offers a promising opportunity to alleviate urban traffic congestion. Thus, developing optimal takeoff trajectories for minimum energy consumption becomes essential for broader eVTOL aircraft applications. Conventional optimal control methods (such as dynamic programming and linear quadratic regulator) provide highly efficient and well-established solutions but are limited by problem dimensionality and complexity. Deep reinforcement learning (DRL) emerges as a special type of artificial intelligence tackling complex, nonlinear systems; however, the training difficulty is a key bottleneck that limits DRL applications. To address these challenges, we propose the transformer-guided DRL to alleviate the training difficulty by exploring a realistic state space at each time step using a transformer. The proposed transformer-guided DRL was demonstrated on an optimal takeoff trajectory design of an eVTOL drone for minimal energy consumption while meeting takeoff conditions (i.e., minimum vertical displacement and minimum horizontal velocity) by varying control variables (i.e., power and wing angle to the vertical). Results presented that the transformer-guided DRL agent learned to take off with $4.57 imes10^6$ time steps, representing 25% of the $19.79 imes10^6$ time steps needed by a vanilla DRL agent. In addition, the transformer-guided DRL achieved 97.2% accuracy on the optimal energy consumption compared against the simulation-based optimal reference while the vanilla DRL achieved 96.3% accuracy. Therefore, the proposed transformer-guided DRL outperformed vanilla DRL in terms of both training efficiency as well as optimal design verification.
Problem
Research questions and friction points this paper is trying to address.
Optimizing eVTOL takeoff trajectories for minimal energy consumption
Overcoming training difficulties in deep reinforcement learning methods
Improving training efficiency and accuracy in optimal control design
Innovation
Methods, ideas, or system contributions that make the work stand out.
Transformer-guided DRL for eVTOL trajectory optimization
Transformer explores realistic state space each step
Reduces training time by 75% compared to vanilla DRL
🔎 Similar Papers
