Robust planning for heterogeneous UAVs in uncertain environments

Abstract

Future Unmanned Aerial Vehicle (UAV) missions will require the vehicles to exhibit a greater level of autonomy than is currently implemented. While UAVs have mainly been used in reconnaissance missions, future UAVs will have more sophisticated objectives, such as Suppression of Enemy Air Defense (SEAD) and coordinated strike missions. As the complexity of these objectives increases and higher levels of autonomy are desired, the command and control algorithms will need to incorporate notions of robustness to successfully accomplish the mission in the presence of uncertainty in the information of the environment. This uncertainty could result from inherent sensing errors, incorrect prior information, loss of communication with teammates, or adversarial deception. This thesis investigates the role of uncertainty in task assignment algorithms and develops robust techniques that mitigate this effect on the command and control decisions. More specifically, this thesis emphasizes the development of robust task assignment techniques that hedge against worst-case realizations of target information. A new version of a robust optimization is presented that is shown to be both computationally tractable and yields similar levels of robustness as more sophisticated algorithms. This thesis also extends the task assignment formulation to explicitly include reconnaissance tasks that can be used to reduce the uncertainty in the environment. A Mixed-Integer Linear Program (MILP) is presented that can be solved for the optimal strike and reconnaissance mission. This approach explicitly considers the coupling in the problem by capturing the reduction in uncertainty associated with the reconnaissance task when performing the robust assignment

(cont.) of the strike mission. The design and development of a new addition to a heterogeneous vehicle testbed is also presented.