Reconciling Conflicting Intents: Bidirectional Trust-Based Variable Autonomy for Mobile Robots

In the realm of semi-autonomous mobile robots designed for remote operation with humans, current variable autonomy approaches struggle to reconcile conflicting intents while ensuring compliance, autonomy, and safety. To address this challenge, we propose a bidirectional trust-based variable autonomy (BTVA) control approach. By incorporating diverse trust factors and leveraging Kalman filtering techniques, we establish a core abstraction layer to construct the state-space model of bidirectional computational trust. This bidirectional trust is integrated into the variable autonomy control loop. Real-time modulation of the degree of automation is achieved through variable weight receding horizon optimization. Through a within-group experimental study with twenty participants in a semi-autonomous navigation task, we validate the effectiveness of our method in goal transfer and assisted teleoperation. Statistical analysis reveals that our method achieves a balance between rapid response and trajectory smoothness. Compared with binary control switching, this method reduces operator workload by 14.3% and enhances system usability by 9.9%.