Unlocking User-Friendly Robotic Control: Adaptive Control Methods for Collaborative Robotic Arms

By Dr. Sampath LonkaReviewed by Susha Cheriyedath, M.Sc.Jul 11 2023

Robotic arms are increasingly used in collaborative settings, requiring users to control multiple degrees of freedom (DoFs) for object manipulation. Traditional input devices necessitate mode switches to access individual DoFs, leading to cognitive load. Adaptive DoF Mapping Controls (ADMCs) have reduced mode switches but failed to decrease the perceived workload.

In a recent submission to the arXiv* server, researchers proposed multimodal feedback in real-time to address this issue, comparing current and suggested mappings. They compared two approaches in a VR study: continuous recommendations and discrete thresholds. The results indicate a decrease in task completion time, a reduction in the number of mode switches, and a lower perceived workload. Continuous and Threshold methods perform similarly, highlighting the need for user-centered customization options to improve usability and user acceptance of robotic technologies.

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.

Background

Robotic devices, driven by advancements in semi-autonomous technologies, are increasingly finding their way into various aspects of our lives. Collaborative robotic arms, in particular, offer versatility in industrial assembly lines and mobility assistance. However, the introduction of (semi-) autonomous actions has posed challenges, leading to stress among end-users.

Previous studies have shown that manual control reduces stress and increases user satisfaction. To address these challenges, ADMCs have been proposed, combining the benefits of autonomous actions with manual controls. The current study evaluates two novel ADMC methods, Continuous and Threshold, compared to a classic control method to assess their impact on task completion time, mode switches, workload, and user experience. The findings highlight significant reductions in task completion time, mode switches, and perceived workload, underscoring the need for customizable configurations based on user preferences.

Related work

Research in collaborative robotic solutions has been prominent, with a particular focus on two main areas of investigation. Firstly, studies have explored robot motion intent designs to facilitate effective collaboration between humans and robots sharing the same space. Augmented reality (AR) has been used to visually convey robot motion intent, making interactions more intuitive and natural.

Secondly, ADMCs have been proposed to streamline robot control methods and reduce the need for frequent mode switches. Previous studies have demonstrated the effectiveness of ADMCs in decreasing mode switches but have also highlighted challenges such as high cognitive demand, difficulties in understanding mappings, and a lack of predictability.

Proposed methods

Building upon previous work, the current study developed a virtual reality (VR) simulation to compare different ADMC methods with a non-adaptive baseline condition, referred to as Classic. Two improvements were implemented: visualizing both current and forthcoming DoF mapping suggestions to enhance predictability and simplifying the input by reducing it to a single DoF to decrease cognitive demand. The control method for ADMCs involved a task-specific script that assessed the gripper's rotation, position, and finger position in relation to a target. The ADMC system provided five movement options (modes) in order of perceived usefulness.

The study evaluated two ADMC methods:

1) Continuous: This control method provides continuous feedback on robot motion intent, allowing users to evaluate and switch to updated movement suggestions. Two directional indicators, a light blue arrow, and a dark blue arrow, represent the selected mode and the optimal suggestion, respectively. While enhancing transparency, this approach may increase distractions, mode switches, and perceived workload.

2) Threshold: Threshold, as opposed to Continuous, employed discrete and multifaceted feedback to signal optimized movement suggestions. Users were notified of updated suggestions through a vibration pulse and sound signal. This method reduced perceived workload as users were not constantly evaluating visual feedback, but it may have increased task completion time and a perceived loss of control.

Experiments

The effectiveness of the ADMC methods was evaluated through a controlled experiment using a VR simulation with 24 participants. The participants performed training trials and measured trials across three conditions: Classic, Continuous, and Threshold. Several dependent variables were measured, including task completion time, mode switches, perceived workload, subjective assessment, and participants' rankings of the control methods. The results demonstrated significant advantages of the ADMC methods compared to the Classic method.

Results

The findings indicated that both the Continuous and Threshold methods significantly reduced task completion time and mode switches compared to the Classic method. Surprisingly, the ADMC methods also significantly reduced perceived workload compared to the Classic method, contrary to previous findings. There were no significant differences between Continuous and Threshold in the analyzed metrics, indicating their equal efficiency. Participant preferences varied between the two methods, highlighting the need for individualization options and the possibility of combining ADMC with static suggestions for enhanced usability.

Conclusions

In conclusion, the ADMC methods evaluated in the present study show promise for improving human-robot interaction in collaborative robotics. The results underscore the importance of usability, safety, and end-user acceptance in designing control methods for collaborative robotic arms.

The study's limitations included the use of a VR simulation environment, emphasizing the need for real-world validation. Future work should focus on identifying factors influencing user preferences and optimizing control methods for real-world applications.

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.