Complementing human and robot capabilities is essential for many tasks, e.g. rehabilitation and collaborative manufacturing. However, it is still not clear how control between humans and robots should be shared in order to ensure efficient task execution and intuitive interaction. Game theory seems as a promising mathematical framework that allows: i) posing this challenge as a dynamic negotiation (game) among human and robot (players) and ii) solving it to obtain optimal solution. In this work, we propose a differential game-theoretic shared control approach for human-robot haptic collaboration with Nash equilibrium optimal solution. We validate the proposed approach experimentally in a scenario where human is physically coupled with a haptic device and interacts with a virtual reality to perform a trajectory tracking task.