A ROS-based motion planning for socially-aware navigation

Today, we are surrounded by intelligent device that can move autonomously. In the future this presence will became more and more common. Some years ago, nobody had an intelligent device that could clean the house, but today is quite common see some robot cleaner. In the future we will have a coexistence between human and robotic device. This evolution has advantages and disadvantages. Our lives will became easier and the boring things will be done from a robot or a smart device. The presence of robots around us could create some issues. The main issue is the co-presence of human beings and robots in the same environment and the rule to manage the interaction between this two kind of universe. Moreover the behavior of robot must be accepted from humans. The robot would have the same way-of-thinking of a human. In this way the human beings could accept the presence of robot, and see the robot like a one of them. Today, this limit of robot is under development. The goal of this work is to develop an algorithm to manage the trajectory of robot. In particular, this trajectory must be accepted by humans. To do that, the behavior of robot must be similar to a human being. The human beings usually decide to move using a collision-free trajectory. Also they keep in mind the other movement, to avoid possible further collision. Moreover, this algorithm must work in a crowd environment. This work want to “anthropomorphizing” the robot behavior. In the literature we have a lot of trials. The biggest part of approaches try to solve the issue thinking only on the robot. To do a better solution, we must consider the other players already present on the environment. In this way, the robot move itself taking apart the information coming from the other players. The more common motion planning approaches use a reactive behavior. The robot see the obstacle and decide to avoid it. The human being instead, use a predictive behavior. The human elaborate before how to avoid the obstacle, seeing also the other player position on the environment. We can take inspiration from the game theory to solve this issue. For example, when we play chess, we try to predict the further moves of other player. In the same way, we can apply this idea to our algorithm. We take a decision knowing the current condition of player and environment, and we think how it will move in the next moments. The our actions are calculated also analyzing the possible future action of the other player. In this thesis we develop a non-cooperative game. In this way every player move itself without any bound and move until they reach the goal. We use the Nash equilibrium concept. This concept is applied on dynamic moves to reach a feasible solution between all the player. Also we analyzed the presence of physical object to avoid. In this way, the robot play with the other player but avoiding at the same time the physical object present on the environment. To do that we insert a cost function and a cost map on the algorithm. The robot knows its position and how much cost the further step to reach the goal. In this way, the robot decide to perform the minimum cost path, taking apart the other player that could change the cost and feasibility of path itself. The algorithm will be checked using a real-world surveillance videos, common used on this field. In this way we can compare the quality of the algorithm. In fact, we saw if the trajectory calculated by algorithm are smart or wrong. In a