https://arxiv.org/pdf/2109.02301.pdf
Introduction: Addressing the challenges in controlling robots, various interfaces have been developed, ranging from direct control to asynchronous control. Direct control involves operators controlling the robot’s movements in real-time, while asynchronous control allows operators to send complex programs for the robot to execute autonomously. The concept of task-level authoring is introduced as a solution, enabling operators to control robots by specifying sequences of high-level steps. This approach provides flexibility to the operator, allowing for both periods of autonomy and direct control as needed.
Limitations: The current system does not explore how to recover from failures during action execution and relies on robust actions.
What is it? The proposed solution allows operators to specify shorter sequences of high-level commands, known as task-level authoring, to create periods of variable robot autonomy.
Advancements compared to previous research:
- Task-level authoring is simpler and does not require specialized knowledge compared to direct control.
- Task-level authoring allows for flexible adjustment of task length compared to asynchronous control. It enables operators to assign shorter tasks for situations that require human intervention and longer tasks for situations where the robot can operate autonomously.
Key techniques and methods:
- End users can break down task specifications and adapt them to the current interaction and environment needs.
- The system combines the reactivity of direct control with asynchronous operation, providing action-level and high-level instructions.
- The system utilizes an AR interface.
Validation: The effectiveness of the system is evaluated through objective measures such as task score and autonomy, as well as subjective measures including NASA TLX, SUS, and user preference.
Discussion: There may be further discussions on the topic.
Recommended future papers to read are not mentioned.