Joint Decision-Making in Robot Teleoperation: When are Two Heads Better Than One?

TL;DR

The paper tackles whether two human operators can exceed a single operator in dynamic teleoperation by applying Maximum Confidence Slating (MCS) to joint robot-controller decisions. Through a large online experiment with 100 participants controlling two robots under latency variation, the study shows that MCS-based dyads outperform the best individual, with gains modulated by individual skill levels and confidence calibration. Key insights reveal that similar performance and well-calibrated confidence within the dyad maximize gains, while large skill gaps or poor calibration can reduce or negate benefits. The findings support using confidence-based joint decision mechanisms to enhance human-robot collaboration in time-sensitive, spatiotemporal control tasks and motivate future work on human-AI dyads and decision-support integration.

Abstract

Operators working with robots in safety-critical domains have to make decisions under uncertainty, which remains a challenging problem for a single human operator. An open question is whether two human operators can make better decisions jointly, as compared to a single operator alone. While prior work has shown that two heads are better than one, such studies have been mostly limited to static and passive tasks. We investigate joint decision-making in a dynamic task involving humans teleoperating robots. We conduct a human-subject experiment with $N=100$ participants where each participant performed a navigation task with two mobiles robots in simulation. We find that joint decision-making through confidence sharing improves dyad performance beyond the better-performing individual (p<0.0001). Further, we find that the extent of this benefit is regulated both by the skill level of each individual, as well as how well-calibrated their confidence estimates are. Finally, we present findings on characterising the human-human dyad's confidence calibration based on the individuals constituting the dyad. Our findings demonstrate for the first time that two heads are better than one, even on a spatiotemporal task which includes active operator control of robots.

Figures (12)

• Figure 1: Web based robot navigation simulator. The red and green curves represent trajectories driven by a participant from our user study who was experiencing low and high delay in control of the mobile robot.
• Figure 2: Screenshots showing driving trials scenario and question queries (\ref{['fig:procedure-2']}: environment setup; \ref{['fig:procedure-4']}: robot choices, \ref{['fig:procedure-5']}: user confidence levels) for each participant. There are total of 100 trials per participant.
• Figure 3: Accuracy in robot selection for high-confidence, higher-performing, low-confidence, lower-performing, and random participants from the paired dyads.
• Figure 4: Variation in accuracy gain as participants of different success rates are paired together. Pairing similar performing participants leads to higher benefit.
• Figure 5: Variation in accuracy gain as participants of different success rates are paired together. Blue: both participants being below the mean skill level, and red: both participants being above the mean skill level.