Socially-Aware Multi-agent Velocity Obstacle Based Navigation for Nonholonomic Vehicles

We present an algorithm for collision free and socially-aware navigation of multiple robots in an environment shared with human beings, other robots and with the presence of static obstacles. We formulate the problem as a constrained optimization problem, where the cost function is chosen in order for the robotic agents to exhibit bio-inspired behaviors, such as cooperation inside the group and cohesive motion. Some of the constraints are required to avoid collision between the agents and with other obstacles and emanate from the application of Velocity Obstacle approach. The nonholonomic dynamics of the vehicles, is managed through the application of the feedback linearization technique to map the velocities in the control values. In this paper we propose both centralized solution and a completely decentralized solution. The overall strategies are extensively tested in simulations.