This Small Business Innovation Research (SBIR) Phase I project proposes a methodology and novel algo- rithms for Personal Service Robotics. Recognizing that, in unstructured environments, autonomous robotics alone is not yet reliable enough for performing complete tasks, and complete manual teleoperation is too labor intensive, this approach combines autonomous operation with human assistance in a Tiered Human- in-the-Loop framework. At all levels, the autonomy will request help when unexpected or anomalous events occur. For example, at high levels the user may be asked to identify objects or adjust plans. At the low level the user may be asked to remote control the robot end-e^ector to unstick a wedged object. Automation will focus on the most frequent task components, including motion planning, object grasping, and simple rigid-environment manipulation. The major innovations arise from (a) the design of limit-aware autonomy in each of these areas, requesting human intervention when needed and (b) multi-tiered human-assisted fallback options that necessitate greater involvement for severe failures or complex circumstances. Phase I will demonstrate an operator assisting a robot completing personal service tasks in non-industrial environ- ments. Phase II will parallelize for an operator supporting multiple robots, with the multiplicative e^ect demonstrating overall labor savings.
The broader impact/commercial potential of this project includes accelerating the deployment of robots in unstructured environments, extending the market for personal robotic services beyond single-purpose autonomous devices, and increasing access to in-home assistance in daily living for an aging population. With demographics shifting to an older population, we face a major shortage of manual labor, particularly for unskilled work. This will impact the o;ce and home environments, and dramatically a^ect the older population for whom daily tasks become increasingly burdensome, but also for whom appropriate service workers may not be available. Allowing an operator to assist robots performing personal service tasks can increase the productivity of such labor and perform greater personal service than traditionally possible. Meanwhile, customers are clearly ready to invest into robotic technology. For reference, even today's highly- limited personal and home robots created a $1.28 Billion global market. A human-in-the-loop personal service robot will greatly expand that market, as well as create markets for operator services and infrastructures. Delivering personal robotic service will have the immediate bene ts of reduced labor costs and increased productivity, while also allowing older adults to age in familiar housing and retain greater independence.
This Small Business Innovation Research (SBIR) Phase I project proposes a methodology and novel algorithms for Personal Service Robotics. Recognizing that, in unstructured environments, autonomous robotics alone is not yet reliable enough for performing complete tasks, and complete manual teleoperation is too labor intensive, this approach combines autonomous operation with human assistance in a Tiered Human-in-the-Loop framework. For example, at high levels the user can identify objects or adjust plans. At the low level the user can remote control the robot end-effector to unstick a wedged object. Automation focuses on the most frequent task components, including motion planning, object grasping, and simple rigid-environment manipulation. At the end of Phase I, we demonstrated a mobile manipulator performing a set of pilot household tasks, in the context of two applications: dinner table cleanup and sorting items retrieved from an office cleanup. The robot showed the capability to perform the tasks fully autonomously in subsets of the cases presented. When necessary, the robot was assisted by a human operator through an interface developed in the course of Phase I, and without direct visual or audio contact with the robot. This approach lead to successful task completion in all attempted trials. Throughout the course of development for this project, we found that the proposed tiered approach to HitL operation is key to enabling task completion while also improving efficiency. In this context, the robot can use the most autonomous methods possible given the situation, and fall back on the operator-driven alternative if needed. In particular, we developed and tested the following autonomous modules and autonomy assisted teleoperation tools: * for collision-free motion planning, we developed autonomous and operator-assisted modules that planned and executed arm trajectories as needed for manipulation tasks; * for grasping and placing objects we developed strategies ranging from fully autonomous object recognition and manipulation, to operator-selected grasps executed autonomously, and finally to tasks executed under complete operator control. We also performed extensive testing with a large number of teleoperators, studying the performance of these strategies in highly cluttered, complex environments. * we developed complete manipulation primitives, such as door or drawer opening, which can be executed under full manual control, or by taking advantage of automated trajectory generation tools. In summary, the results obtained and technology developed during Phase I include: * demonstration of a HitL robotic system performing complex manipulation tasks related to daily living activities in unstructured environments; * demonstration of a Tiered HitL system, with teleoperator input combined with autonomous execution of (sub-)tasks at different levels of autonomy; * motion planning tools for autonomous and HitL manipulation, enabling collision-free arm movement and base navigation; * a novel system for HitL execution of grasping tasks, comprising a number of different strategies each requiring a different level of operator involvement; * a method for assisted execution of manipulation tasks such as door and drawer opening using appropriate coordinate frames for generating gripper trajectories.
