An ADSR defines a sound's character: how fast the Attack rises, where the Decay settles, how loud the Sustain holds, and how long the Release fades.
Exponential curves usually sound more natural than linear ones — human loudness perception is logarithmic. JUCE's ADSR class is exponential by default, but you still need to map parameter ranges yourself.