People like to use the free utility called Levelator, but this is mostly to balance out levels that vary widely over a long program and are tedious to ride manually. Still, you might want to try it, it is free and works very fast with adrag-and-drop action, much like an audio version of Mpeg Streamclip, only not as multi-functioned.
If there was a real, easy answer to your problem, somebody would be rich from marketing it and we’d all know about it already and use it.
You might try downward expansion on the rumbles, applying parametric EQ, some compression and gating and limiting on the noise, but when you start with bad raw material, you really can’t make it good, only perceptually less bad. And the amount of time this is going to take you will be significant.
The COW pro audio forum would have better specific suggestions for you, but my guess is their number one suggestion will be to not stint on the expense of having someone competent with the right tools doing the audio job in the first place. I’m truly sorry that’s not a helpful answer for your immediate problem. Every project failure I’ve ever had was pretty much based on this failure; not spending nearly enough time and money getting the audio right from the beginning. Audiences forgive bad lighting and camerawork, but they will NOT sit still for bad audio.