Human-Centric Co-Adaptation for Optimal Human-Machine Coordination

Abstract

In this paper, we investigate the optimal co-adaptation strategy between humans and machines in order to maximize the performance outcome in completing a coordinated task. Although prior works identified the optimal machine adaptation for human’s specific condition, it is still not clear how to design an optimal co-adaptation strategy which enables the machine to adapt to the human concurrent with human adaptation. To achieve the optimal co-adaptation between human and machine, the adaptation strategy should maximize the immense potential of the human adaptation while minimizing the resources required of the machine. To accomplish this, we propose a novel human-centric co-adaptation strategy of the machine. In our strategy, the machine initially waits for the human to adapt to machine. Once human adaptation arrives to the plateau, the machine starts to adapt. Rather than addressing the error fully on its initial adaptation, the machine still provides headroom for the human to adapt further, since the changed condition might augment the capability of human adaptation. We call this strategy human-centric co-adaptation, as the machine adapts based on the human’s capability. We implemented the test setup by measuring the step distance of a human using a treadmill alongside an optical tracking system. We tested our strategy alongside two other adaptation strategies: single sided adaptation, where only the human adapts to the machine, and co-adaptation without a strategy, where both human and machine concurrently adapt to each other without priority. Our results indicate that, with an accuracy task, machine co-adaptation did not reduce the error in easy target. However, if the target became challenging for the human, machine co-adaptation successfully lowered the error. When the human-centric strategy was applied to the co-adaptation, the error was not further reduced but the dependency to the machine was reduced.