Multimodal sensing for socially compliant safe robot navigation in human-aware indoor environments using group-aware pose estimation and modified RRT

Multimodal sensing and perception are crucial for socially assistive robots (SARs) to navigate effectively in human-aware indoor environments. This article aims to equip SARs with the multimodal perception capability to assist visually impaired individuals in safely navigating and perceiving their surroundings, thereby enhancing their social engagement. The focus is on enabling visually impaired users to identify their destination in social scenarios and move independently toward it. A novel group-aware pose estimation (GAPE) algorithm is implemented to identify a suitable position for a new group member. In addition, a modified RRT* algorithm for path planning, an obstacle avoidance framework, and a follower-aware controller are used to guide the user toward the goal. Multimodal feedback is provided through audio and a custom-designed hand-held direction indicator, the NAVI-Stick. The proposed GAPE algorithm operates in real-time, achieving a pose metric accuracy of over 90% for groups with fewer than six members. Trials conducted with a blindfolded participant demonstrated effective guidance toward the desired human group. The maximum Euclidean distance between the user’s final position and the goal position across 15 trials was 0.98 m. The follower-aware controller exhibited an average Euclidean-distance-based positional error of 0.0943 m.