C. Gehring and D. Precup. Smart exploration in reinforcement learning using the absolute temporal difference errors. AAMAS, 2013. link