Safe Bipedal Locomotion and Navigation in Uncertain Environments

This dissertation tackles the challenge of enabling bipedal robots to navigate safely in dynamic, uncertain environments. Unlike industrial robots in structured settings, bipedal robots face significant complexities in real-world scenarios due to challenges in locomotion, obstacle avoidance, and environmental unpredictability. This research introduces a hierarchical approach involving task planning, motion planning, and body control to enhance navigation safety, structured across three frameworks. The first framework addresses task and motion planning in a partially observable environment with dynamic obstacles, using Linear Temporal Logic (LTL) to ensure safe navigation even when obstacles are out of view. The second framework focuses on social navigation, employing a Social Zonotope Network (SZN) to predict pedestrian paths and plan socially acceptable paths for the robot, enabling it to avoid collisions and navigate within a social environment. The third framework addresses terrain uncertainty in a heterogeneous team search and rescue context, where both bipedal and aerial robots cooperate to explore the environment. Here, a terrain-aware model predictive controller helps the bipedal robot navigate unknown, uneven terrain by optimizing paths for safety and traversability.