Humanoid robot motion planning & balance in dynamical environment

Dear, I would like to make humanoid robot simulation of complex tasks (ROS, Gazebo and others). Since there is many novel approaches how to do path planning, balance and object manipulation, Iam looking for a guidance & advices about: 1, What are possible approaches and its combinations that could possibly fit to my desired simulation (object manipulation, movement in dynamical environment)? 2, What would you think about my personal high-level idea how to combine RRT++, balance DRL with pre-defined path planning function in order to deliver complex motion planning in dynamical environment such as our daily interiors while saving computation time allowing real time tasks & actions?