Human Aware Robot Navigation

As robots grow their influence in our lives, it is important to develop methods which allow robots to work and move along-side humans more naturally without compromising our safety. In this paper, we work towards implementing motion planning techniques and their comparative analysis for human aware robot navigation. Our approach uses predicted human trajectories and a cost function to plan collision free paths while taking human/social comfort into account, our approach generates paths that is more like how a human would navigate, such as waiting for the other person to move, navigating along side another person, and making way when blocking the person’s path. We will be conducting experiments for the robot navigation in hospital environment using Unity simulation and ROS.

Experimental Demonstrations

Experiment 1:

Our first experiment tests robot navigation in a intersection space where humans are crossing the robot’s path. For this scenario we intend to make the robot wait till the human passes by before re-planning to move towards its goal. In this experiment we intend to employ TEB-local planner and the results are a follows, we can see that the TEB-local planner needs to be optimized for our application, after successfully optimizing the TEB-local planner, we can see the local path being planned in real-time in the Rviz window, and when a human is detected, a path is generated to avoid collision.

Experiment 2:

The second experiment tests robot navigation in a corridor with humans walking parallel to and opposite to the robot’s navigation direction. For this scenario we intend to make the robot give enough space for the human to pass by and not make the human uncomfortable. To evaluate this experiment where the robot needs to navigate with respect to the human’s movement, we make the robot move in direct path of the human as shown. Here, we can see that the global path planned (orange) is in collision with the human but TEB-local planner path (blue) avoids collision with the human. This experiment fails if the robot is too slow to avoid the human, or if the human’s movement is faster than the local planner’s ability to re-plan.

Experiment 3:

The third experiment demonstrates our robot’s global path planning capabilities and compare TEB and Timed Path Followers ability to re-plan for dynamic-moving obstacles and new static obstacles in the environment.