Dynamic Allocation of Autonomy for Limited-Bandwidth Human-Robot Teams Based on Measures of Trust in the Human
Abstract
Task TO: The development of enabling technologies for dynamic autonomy allocation and behavior learning: (1) Human Signal Interpretation. The development of tools that extract supplementary information from human control signals provided under bandwidth constraints. These tools at a minimum will extract information about the quality of the control channel over which the signals are provided, and whether they agree with the autonomy, (2) Trust Computation. The development of tools that reason about the extent to which each teammate should be trusted to produce high quality control signals. Multiple ways to measure team performance will be expl9red (3) Demonstration. The effectiveness of the tools will be verified on a mobile robot performing a collaborative navigation task with a human teammate. Task Tl: The development and assessment of mechanisms to promote situational awareness: (1) Alert paradigm. The development of software to realize the three paradigms to alert the human of changes in autonomy level, (2) Assessment (Robotic Arm). The paradigms developed will be assessed via user studies {30 subjects in Year 2} with a robotic arm, where the human-robot teams perform collaborative manipulation tasks. Task T2: The development and assessment of algorithms for dynamic autonomy allocation: (1) Algorithms for Dynamic Autonomy Allocation. The development of algorithms to dynamically shift between allocations of autonomy between human and robot teammates under the proposed framework and using tools developed under task TO. (2), Assessment (Mobile Robot}. The algorithms developed under will be assessed via user studies (30 subjects in Year 3) with a mobile robot, where the human-robot teams perform collaborative navigation tasks, (3) Porting to Robotic Arm. The implementation of milestone MS, including integration of the tools developed under Task TO, on the robotic arm. (4) Assessment (Robotic Arm). The algorithm will be assessed via user studies (30 subjects in Year 4) with a robotic arm, where the human-robot teams perform collaborative manipulation tasks. Task T3: The development and assessment of various approaches for learning from human teammate control signals at different autonomy levels: (1) Algorithms for Learning. The development of three flavors of machine learning algorithms, that learn from human control signals at different autonomy levels, leveraging the tools developed under task TO, (2) Assessment (Mobile Robot}. The learning algorithms will be assessed and compared via user studies with a mobile robot (20 subjects in Year 3, 40 subjects in Year 4), where the human-robot teams perform collaborative navigation tasks. Task T4: The development autonomy behaviors for the subject studies with human-robot teams. To provide a rich validation domain for the user studies proposed under Tasks T2 and T3 however, the suite of autonomy behaviors available on the two robot platforms should be extended. (1) Mobile Robot. The implementation of a sufficiently rich suite of behaviors at each autonomy level, on the mobile robot platform. (2) Robotic Arm. The implementation of a sufficiently rich suite of behaviors at each autonomy level, on the robotic arm platform.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 03, 2016
- Source ID
- N000141612247
Entities
People
- Brenna Argall
Organizations
- Office of Naval Research
- Shirley Ryan AbilityLab
- United States Navy