Machine vision algorithms for robust pallet engagement and stacking

The Army has a critical need for the automation of logistics assets that reduce the cost of operations, while increasing the efficiency and safety of operations. These vehicles sense the world around them and react to the environment, thereby minimizing the need to modify the working environment. An important feature that needs to be addressed is pallet engagement: the ability to find cargo pallets, sufficiently localize them so that autonomous tactical forklifts can safely engage and interact with them, and then manipulate those containers for stacking and unstacking operations. In this paper, we detail our work with using the Light Capability Right Terrain Forklift (LCRTF) as the autonomous platform for safely engaging Joint Modular Intermodal Containers using machine vision techniques. We will: . Detail the requirements for the system; . Discuss the issues involved with the pallet, the vehicle, and the environment; . Discuss the sensors used; . Detail the machine vision techniques that failed to support this task, and why; . Detail the current effective methods for machine vision based pallet engagement; and . Discuss future efforts.