Human – Centered Robot Navigation – Towards a Harmoniously Human – Robot Coexisting Environment

Chi-Pang Lam, Chen-Tun Chou, Kuo

IEEE Transaction on Robotics, Vol. 27, No.1, February 2011

Summary

                  Harmonious coexistence of robots with humans is an in interest for improving HRI and the navigation of robots in human environments is rarely studies. Two are the main issue to be addressed: the robot should be able to move autonomously and safely in a human – robot environment in order to complete a specific task and the robot should behave in both human and robot friendly manner during operations, this last issue is the one covered in the authors research. In the literature there are different proposed solutions as traffic rules (which would require to regulate humans’ behavior, which of course is not possible), artificial potential fields (which risk to stuck in local minima), nearness diagram (ND) navigation (which is a reactive navigation approach based on laser scanner), people tracking and so on. The authors propose a human-centered sensitive navigation (HCSN), imposing 6 harmonious rules for human and robot coexistence in the same environment: 1) Collision-free rule; 2) Interference-free rule; 3) Waiting rule; 4) Priority rule; 5) Intrusion rule; 6) Human rule (human have the highest priority). Robots may then have there own internal rules. Also 6 sensitive field have to be defined. The Human Sensitive Field is around the human and assumes an egg shape, bing longer ahead of the human (elliptic) and shorter behind (circle) since it must consider human motion, this field has the highest priority. The Stationary Robot Working Field is the robot working (or just waiting) in a fixed location and has the scape of a disc, it is second in the priority rule. The Movable Robot Working Field is the robot working while moving, it is third in the priority rule and also has a donut shape. The Robot Normal Field is the idle o not working robot, it presents a disc shape and has lowest priority. The Human-Robot Stationary Joint Field present both human and robot working together in stationary position, it assumes a circle shape with the center in the middle between the human and robot. The Human-Robot Moving Joint field also assumes an egg shape (whit longest axis facing the operator). The two combination fields assume the same priority as the Human Sensitive Field, for simplicity the 6 states are denominated as follow in the same order: H1, R1, R2, R3, HR1, HR2. The HCSN design consist of a VLH (Virtual Laser Histogram) which collects all the virtual distance data within 5 meters, this information is used for sensitive-field sensing, while situation identification, together with sensitive field sensed information inputs the motion planner, which then inputs the controller, providing in the end a final output.

Unluckily humans are not easy to localize and paradoxically information regarding them is much less than the ones regarding robots. The motion planner involved the use of ND navigation to fin out a suitable free walking area, while the potential field approach is later used to establish different potential field near the walking area we choose. Different cases are presented, it must be noted that in the priority rule, highest priority may not necessary be assigned to the closed to obstacle, but there is a risk factor applied in order to under preview future closest obstacles, in order to ensure smooth motion.

The priority rule is then used in the calculation of the potential field, which its negative gradient is taken to define the attractive force. Similar computation are then obtained for repulsive force and for repulsive force associated with the soft field (which is the flexible field regarding R2 and R3). The overall force is then the sum of the three and this will determine the sub-goal position, with a matter of a D₀ factor which is a look-ahead distance which is kept smaller into the sensitive field, while the gain is kept smaller for more effective look ahead distance.

Key Concepts

Navigation, Human-Robot Cooperation, Human-Robot Interaction

Key Results

The method, although being complex, can be performed smoothly in real-time environments.