An Approach for Multi-Robot Opportunistic Coexistence in Shared Space

This thesis considers a situation in which multiple robots operate in the same environment towards the achievement of different tasks. In this situation, please consider that not only the tasks, but also the robots themselves are likely be heterogeneous, i.e., different from each other in their morphology, dynamics, sensors, capabilities, etc. As an example, think about a "smart hotel": small wheeled robots are likely to be devoted to cleaning floors, whereas a humanoid robot may be devoted to social interaction, e.g., welcoming guests and providing relevant information to them upon request. Under these conditions, robots are required not only to co-exist, but also to coordinate their activity if we want them to exhibit a coherent and effective behavior: this may range from mutual avoidance to avoid collisions, to a more explicit coordinated behavior, e.g., task assignment or cooperative localization. The issues above have been deeply investigated in the Literature. Among the topics that may play a crucial role to design a successful system, this thesis focuses on the following ones: (i) An integrated approach for path following and obstacle avoidance is applied to unicycle type robots, by extending an existing algorithm [1] initially developed for the single robot case to the multi-robot domain. The approach is based on the definition of the path to be followed as a curve f (x;y) in space, while obstacles are modeled as Gaussian functions that modify the original function, generating a resulting safe path. The attractiveness of this methodology which makes it look very simple, is that it neither requires the computation of a projection of the robot position on the path, nor does it need to consider a moving virtual target to be tracked. The performance of the proposed approach is analyzed by means of a series of experiments performed in dynamic environments with unicycle-type robots by integrating and determining the position of robot using odometry and in Motion capturing environment. (ii) We investigate the problem of multi-robot cooperative localization in dynamic environments. Specifically, we propose an approach where wheeled robots are localized using the monocular camera embedded in the head of a Pepper humanoid robot, to the end of minimizing deviations from their paths and avoiding each other during navigation tasks. Indeed, position estimation requires obtaining a linear relationship between points in the image and points in the world frame: to this end, an Inverse Perspective mapping (IPM) approach has been adopted to transform the acquired image into a bird eye view of the environment. The scenario is made more complex by the fact that Pepper’s head is moving dynamically while tracking the wheeled robots, which requires to consider a different IPM transformation matrix whenever the attitude (Pitch and Yaw) of the camera changes. Finally, the IPM position estimate returned by Pepper is merged with the estimate returned by the odometry of the wheeled robots through an Extened Kalman Filter. Experiments are shown with multiple robots moving along different paths in a shared space, by avoiding each other without onboard sensors, i.e., by relying only on mutual positioning information. Software for implementing the theoretical models described above have been developed in ROS, and validated by performing real experiments with two types of robots, namely: (i) a unicycle wheeled Roomba robot(commercially available all over the world), (ii) Pepper Humanoid robot (commercially available in Japan and B2B model in Europe).