Planning in Dynamic Environments using Evolving Time-Indexed Graphs

In several scenarios of human-robot interaction, the robot operates in a dynamic environment and has limited knowledge of its environment as well as the surrounding humans. In such cases modeling and reasoning over time can be beneficial, since both the environment and robot’s model of the environment changes over time. Here, we discuss planning with explicit consideration of time using “evolving time-indexed graphs,” and study it in the context of robot navigation among humans. Following, the safe-interval path planning (SIPP) algorithm, we pose the planning problem as graph search with time included as part of the state space. We propose forward safe-interval planner (FSIP), an incremental extension of SIPP, for applications where the robot’s model of the environment evolves during execution. Initial evaluations of the algorithms for path planning are carried out using simulation within Robot Operating System (ROS). We discuss the performance of the incremental approach, its limitations and our on-going work.