Robotic navigation in environments shared with other robots or humans remains challenging as the intentions of the surrounding agents are not directly observable. Moreover, interaction is crucial to enable safe and efficient navigation in crowded scenarios. Local trajectory optimization methods, such as model predictive control (MPC), can deal with those changes but require global guidance and trajectory predictions for the other agents, which is not a trivial problem in crowded scenarios. This presentation will present my recent work on learning-based methods to enhance local trajectory optimization methods. Firstly, I will introduce multimodal trajectory prediction models enabling autonomous vehicles with anticipatory behaviors. Secondly, I will present my research on learning interaction-aware policies to provide long-term guidance to a local trajectory optimization planner via deep Reinforcement Learning (RL). Our findings indicate that combining learning and optimization methods improves navigation performance substantially than solely learning or optimization-based motion planners in cluttered environments.