Energy-Efficient Trajectory Planning for Long-Endurance Unmanned Systems: A Comparative Analysis

Abstract

Energy efficiency represents the primary constraint limiting endurance of battery-powered unmanned systems, necessitating trajectory planning algorithms that minimize energy consumption while satisfying mission objectives and operational constraints. This paper presents a comprehensive comparative analysis of energy-efficient trajectory planning methods across aerial, ground, and marine unmanned platforms. We systematically examine energy consumption models for different vehicle types: aerodynamic drag and induced drag for fixed-wing UAVs, rotor power requirements for multicopters, rolling resistance and drivetrain losses for ground vehicles, and hydrodynamic drag for underwater vehicles, analyzing model fidelity trade-offs between computational simplicity and prediction accuracy. The study categorizes trajectory planning approaches into graph-based methods (A*, Dijkstra with energy-weighted costs), sampling-based planners (RRT*, PRM* with energy metrics), optimization-based techniques (nonlinear programming, optimal control, dynamic programming), and learning-based approaches (reinforcement learning, imitation learning from optimal demonstrations). For each method, we analyze the incorporation of energy models into planning objectives, examining single-objective formulations minimizing energy consumption and multi-objective frameworks balancing energy efficiency with mission time, path smoothness, and safety margins. Particular attention is devoted to handling vehicle dynamics constraints including maximum velocity, acceleration limits, and minimum turning radius, as well as environmental factors such as wind fields for aerial vehicles, terrain slopes for ground vehicles, and ocean currents for marine platforms. The comparative analysis develops a unified simulation framework implementing twelve prominent trajectory planning algorithms across standardized scenarios: point-to-point navigation, area coverage, and multi-waypoint missions in environments with obstacles and dynamic disturbances. Performance evaluation encompasses energy consumption, computation time, path quality metrics (smoothness, clearance), and robustness to model uncertainties and environmental variations. Simulation results reveal that optimization-based methods achieve 15-30% energy savings compared to geometric planners but require significantly longer computation time, while learning-based approaches offer real-time performance after offline training with near-optimal energy efficiency. We investigate the impact of planning horizon length, demonstrating that longer horizons enable exploitation of environmental features (thermal updrafts for soaring, favorable currents) but increase computational burden and sensitivity to prediction errors. The paper examines velocity profile optimization showing that speed variation along fixed geometric paths yields substantial energy savings, particularly for vehicles with nonlinear energy-speed relationships. Advanced techniques are analyzed including chance-constrained optimization for energy-efficient planning under uncertainty, model predictive control for online replanning with receding horizon, and hierarchical planning decomposing long-duration missions into energy-optimal segments. Application-specific considerations are discussed for different mission types: surveillance requiring loitering with minimal energy expenditure, delivery missions prioritizing total mission time, and exploration balancing energy efficiency with information gathering.