Towards Human-Centered Haptic Shared Autonomy for Assisted Driving Systems

Mbanisi, Kenechukwu

Etd

Towards Human-Centered Haptic Shared Autonomy for Assisted Driving Systems

Public

Recent advances in sensor technology and declining computing costs have contributed to the rising adoption of intelligent assisted driving systems. These systems complement human efforts in vehicle navigation across various aspects of personal mobility including passenger vehicles, wheelchairs, and mobile telepresence robots. Shared autonomy (SA) formalizes this approach of synergistic collaboration of both the human and autonomous agent to achieve a common navigation task. Haptic shared autonomy (HSA) is a form of SA where both agents “blend” their control inputs by applying forces on a force-enabled control interface (such as a motorized steering wheel or haptic joystick). Existing research shows that HSA systems lead to improved driving task performance, increased situational awareness and reduced workload for the human. However, these systems still struggle with issues around control authority arbitration and intent misalignment which lead to undesirable conflict between the human and the autonomous agent, resulting in user dissatisfaction and, oftentimes, abandonment of the system altogether. This dissertation presents a human-centered design approach with four proposed contributions to address these open challenges of haptic shared autonomy in assisted driving. The first contribution provides a better understanding of the implications of anticipation uncertainty on human motor control in force-based interactions. The second contribution is SocNavAssist, a multimodal (haptic + visual) shared autonomy framework that provides intuitive and intelligent navigation assistance for teleoperated mobile robots in human environments. The third is an extension to SocNavAssist to enable socially-aware assistance adaptation to different driving objectives/styles when navigating in dense, human-populated environments. Lastly, a novel approach for modeling and simulating human driver control behavior for improved driving assistance system design is proposed.

Creator