A Novel Hybrid Aerial/Ground Mobile Manipulator

Abstract

This research is concerned with the mechanical design of a mobile manipulating un- manned ground vehicle (MM-UGV) coupled with an existing commercial unmanned aerial vehicle (UAV) to create a novel hybrid aerial/ground mobile manipulator. A hybrid robotic system capable of manipulating in both aerial and ground environments is a new research direction for field robotics. The hybrid system has the potential of stimulating new research and engineering challenges as well as providing multipurpose robotic systems for industrial applications. A bilevel optimization-based strategy is presented for making important design choices, such as the selection of gear ratios, electric DC motors, manipulator link lengths, and UGV base length. The objective is to minimize the overall mass of the MM-UGV such that hybridization given certain constraints is possible. Constraints related to workspace, dynamic tip-over stability, actuator torque/force limits, and battery properties are incorporated into the formulation in order to ensure maneuverability of the system. Design specifications such as the expected range of end-effector forces, operating surface grade, and various position, velocity and acceleration variables are input to the optimization problem. The resulting problem is formulated as a robust mixed integer bilevel nonlinear program (MINBP), in which some of the constraints are derived from maximization/minimization over the operational variables to ensure constraint satisfaction in all possible ground/air operation scenarios. Optimizing over the operational variable space is a novel technique compared to current research that focuses on optimizing robotic mechanical components for a single trajectory. The Branch-and-Sandwich Bilevel optimization algorithm (BASBL) was used to find a solution to the optimization problem. A parallelized version of the algorithm was implemented and deployed on an IBM BladeCenter computer cluster. Speedup and parallel efficiency of the algorithm show significant improvement over the serial approach. Improving the run time of the optimization is critical for iterative engineering design with different input parameters. A prototype using an optimal set of design parameters is constructed and results of a preliminary flight experiment are reported. The proposed design optimization methodology is rather general and could be applied to other robotic systems.