Redundancy resolution for mobile manipulation
MetadataShow full item record
Mobile manipulation provides avenues to deploy dexterous, precise manipulators into environments that are not suited for fixed-base manipulators. To do this, complex decision making and control methods need to be implemented to properly allocate system resources to the task requirements and environmental constraints. This report presents a foundation from which complex decision making and control can be implemented on mobile manipulation systems. Currently, mobile manipulator modeling techniques vary widely and often contain a loss of generality. A readily deployable modeling technique that can incorporate multiple mobile platform and manipulator types is needed. The proposed modeling method combines the rate kinematic formulations of mobile platforms with rate kinematic formulations of manipulators in order to create a single kinematic model for the combined system. This modeling technique allows for preexisting Redundancy Resolution Techniques (RRTs) and performance criteria developed for manipulators to be used on mobile manipulation systems. The combined kinematic model is also useful for new performance criteria generation and inverse kinematics. Existing inverse kinematics techniques such as Resolved Rate can be implemented on the unified kinematic model to find input solutions to a given output. Mobile manipulation systems are redundant with respect to the end effector leading to the need for a decision making structure in order to properly allocate its resources. Currently, outside of UTRRRG mobile manipulation redundancy resolution is limited to incorporating small numbers of performance criteria. Gradient Projection (GP) is typically used to resolve the redundancy of these systems limiting the criteria selection to be smooth and continuous. This research provides a modeling technique and two new mobile manipulation specific RRTs that can be directly implemented with the Direct Search (DS) RRT to better generate appropriate solutions for given requirements. When implemented with DS, Mobile Manipulation Generate Options (MMGO) provides improved solutions for mobile manipulators over the existing Simple perturbation strategy without a computational performance hit. Mobile Manipulation Configuration Control (MMCC) is a method of separately prescribing a mobile platform path from the manipulator end effector path. It is often beneficial to be able to incorporate mobile robot path planners with mobile manipulation systems to avoid constraints while maintaining the task requirements of the end effector. This work also develops three new performance criteria which are specific to mobile manipulation. The proposed performance criteria include Wheel Slip Avoidance (WSA), Tip Over Avoidance (TOA), and Static Stability Availability (SSA). These criteria are checks to ensure a solution generated by a DS technique does not cause the mobile manipulator to slip on the ground or tip over. Combining existing RRTs and performance criteria with these new mobile manipulation specific RRTs and performance criteria provides a strong basis for developing complex Decision Making Strategies for mobile manipulation systems.