Pritesh P. Narayan

Computationally adaptive multi-objective trajectory optimization for UAS with variable planning deadlines

This paper presents a new approach which allows for the computation and optimization of feasible 3D flight trajectories within real time planning deadlines, for Unmanned Aerial Systems (UAS) operating in environments with obstacles present. Sets of candidate flight trajectories have been generated through the application of maneuver automaton theory, where smooth trajectories are formed via the concatenation of predefined trim and maneuver primitives; generated using aircraft dynamic models. During typical UAS operations, multiple objectives may exist, therefore the use of multi-objective optimization can potentially allow for convergence to a solution which better reflects overall mission requirements. Multiple objective optimization of trajectories has been implemented through weighted sum aggregation. However, real-time planning constraints may be imposed on the multi-objective optimization process due to the existence of obstacles in the immediate path. Thus, a novel Computationally Adaptive Trajectory Decision (CATD) optimization system has been developed and implemented in simulation to dynamically manage, calculate and schedule system execution parameters to ensure that the trajectory solution search can generate a feasible solution, if one exists, within a given length of time. The inclusion of the CATD potentially increases overall mission efficiency and may allow for the implementation of the system on different UAS platforms with varying onboard computational capabilities. This approach has been demonstrated in this paper through simulation using a fixed wing UAS operating in low altitude environments with obstacles present.

Multi-Objective UAS Flight Management in Time Constrained Low Altitude Local Environments

This paper presents a new framework for Multi-Objective Flight Management of Unmanned Aerial Systems (UAS), operating in partially known environments, where planning time constraints are present. During UAS operations, civilian UAS may have multiple objectives to meet including: platform safety; minimizing fuel, time, distance; and minimizing deviation from the current path. The planning layers within the framework use multi-objective optimization to converge to a solution which better reflects overall mission requirements. The solution must be generated within the available decision window, else the UAS must enter a safety state; this potentially limits mission efficiency. Local or short range planning at low altitudes requires the classification of terrain and infrastructure in proximity as potential obstacles. The potential increase in the number of obstacles present further reduces the decision window in comparison to high altitude flight. A novel Flight Management System (FMS) has been incorporated within the framework to moderate the time available to the environment abstraction, path and trajectory planning layers for more efficient use of the available decision window. Enabling the FMS during simulation increased the optimality of the output trajectory on systems with sufficient computational power to run the algorithm in real time. Conversely, the FMS found sub-optimal solutions for the system with insufficient computational capability once the objective utility threshold was decreased from 0.95 to 0.85. This allowed the UAS to continue operations without having to resort to entering a safe state.

Embedding Human Expert Cognition Into Autonomous UAS Trajectory Planning

This paper presents a new approach for the inclusion of human expert cognition into autonomous trajectory planning for unmanned aerial systems (UASs) operating in low-altitude environments. During typical UAS operations, multiple objectives may exist; therefore, the use of multicriteria decision aid techniques can potentially allow for convergence to trajectory solutions which better reflect overall mission requirements. In that context, additive multiattribute value theory has been applied to optimize trajectories with respect to multiple objectives. A graphical user interface was developed to allow for knowledge capture from a human decision maker (HDM) through simulated decision scenarios. The expert decision data gathered are converted into value functions and corresponding criteria weightings using utility additive theory. The inclusion of preferences elicited from HDM data within an automated decision system allows for the generation of trajectories which more closely represent the candidate HDM decision preferences. This approach has been demonstrated in this paper through simulation using a fixed-wing UAS operating in low-altitude environments.