Flexible Control Authority With a Robotic Arm: Facilitating Seamless Transitions Between User and Robot Control in Multi-Action Manipulation Tasks.

点击翻译按钮获取中文摘要

This proposal discusses the development and evaluation of a collaborative Assistive Robotic Manipulator (ARM) which allows the Veteran to preference control authority (user moving the arm versus robot software moving the arm) at any point in a multi-action task: flexible control authority (FlexCA). Veterans with tetraplegia, amyotrophic lateral sclerosis (ALS), and multiple sclerosis (MS) can readily access electric power wheelchairs (EPWs) for their mobility needs but are still limited in performing manipulation tasks. Current options for manipulation assistance include low tech specialized manipulation tools, modifications of the environment, caregiver support, and robotic arms. ARMs for EPW users are capable of a variety of functions but are unintuitive to control. Manually driving the control input for an ARM requires reasoning about many different modes for moving the arm in three directions, rotating the arm’s wrist in three directions, and opening and closing the gripper. The addition of more intelligent robot software assistance can ease control burdens but should also adapt to a user’s changing abilities, preferences, and contexts. Previous work pre-assigns control authority, which creates a rigid order of actions that causes unexpected behavior if the user decides to interrupt the system. This career development award (CDA-1) addresses an unforeseen challenge, the ability to allow the user to change their control authority preference by creating a seamless transition between manual control and robot autonomous operation both within and between different functional task actions. This is achieved through a development aim and an evaluation aim. The first aim is to develop the FlexCA assistive dialogue control system for ARMs that tracks the user state within a kitchen task allowing the Veteran to initiate control authority. The first goal of this aim is to infer the current state within a multi-action task regardless of how the robot is controlled. This is formulated as a sequential stochastic state model that leverages environment observations and user input to successfully monitor and estimate the state allowing the robotic software to seamlessly pick up wherever the user leaves off. Input and feedback from the system is initiated through the development of a voice-activated user interface. The system is evaluated in two phases. The first phase involves a representative user that actively participates in the iterative design process, and a group of benchmark users. The second phase evaluates the FlexCA system among Veterans with limited arm function who use EPWs. Veterans are recruited to provide objective performance measures and subjective assessments of usability and task workload. This work empowers Veterans with upper limb impairment to perform realistic functional tasks with an ARM. By achieving our aims, algorithmic developments can be generalized towards other assistive devices, and inform future models for more appropriately matching assistive systems that adapt to the user positively impacting their quality of life. This CDA-1 provides training experience in assistive technology, clinical methodology, and human robot interaction that bridges the candidate’s previous experience in robotics and software industry with the development of assistive technology to increase Veteran independence. Immediate goals include constructing user-centric driven technology specific to rehabilitation robotics and assistive devices and understanding the methodology of human subject study design. Long-term career goals include future work as a principal investigator on topics that merge probabilistic techniques, such as predictive modeling and stochastic decision making, for more context-aware and user-centric assistive technology platforms that include smart home environments, robotic wheelchairs, and robotic assistive arms.