Safety in mobile robot navigation is an essential aspect, but it is often accompanied by a trade-off of efficiency in different navigation steps. For global planning, safety is frequently handled by inflated obstacles. Larger margins to the obstacles increase safety while lessening the path efficiency, such as longer path length. For local motion planning tasks, safety becomes even more critical because of dynamic objects in the robot's proximity but also accompanies substantial trade-offs, such as highly low-speed operations. In such aspects of safety and efficiency, we propose safe and efficient global path planning and local motion planning methods.

For global path planning, we proposed the allowable speed of navigation for safety, which limits the maximum speed based on the clearances in the environment. The corresponding cost formulates the traveling time, leading to planning a time-efficient global path. A new map representation is proposed, named Hierarchical Topology Map (HTM) and Hierarchical Topology Map with Explicit Corridor (HTM-EC), and incorporated with the proposed safety-aware navigation speed. HTM is a double-layered data structure of a topology graph and the corresponding metric skeleton points of a map, which returns a feasible, on-skeleton path in an extremely short time. HTM-EC is an extended map expression with the Explicit Corridor that incorporates the nearest obstacle points along with skeleton points into HTM, which returns an optimal, off-skeleton (i.e., metric) path with corridor optimization of the given on-skeleton path. When they are used together, safety-aware time-efficient paths can be planned with light computation. The efficacy of the safety-aware allowable speed and lighter computations have been verified with simulations and experiments.

For local motion planning, we incorporate a pedestrian navigation model to address efficiency. The pedestrian model seeks the desired direction of motion that minimizes the remaining distance to the destination. The cognitive collision locations are computed, which enables far-sighted path planning by incorporating future configurations of the surroundings. Safety guarantee is analyzed for the output of the pedestrian-model-based motion planner. Finally, to address the behavioral uncertainties, the degree of cooperation (DoC) is proposed that helps prediction of the other objects' velocities with online estimation. The simulation results with the different number of agents show that the pedestrian model can generate a smooth path between the agents. As the local planner uses fixed-length visual measurement such as LiDAR, the proposed planner is scalable regardless of the number of moving obstacles as well as the types of obstacles (i.e., static or dynamic).