Improving Gradient Estimation by Incorporating Sensor Data
This work addresses the challenge of noisy reward signals in reinforcement learning for researchers and practitioners, though it appears incremental as it builds on existing policy search methods.
The paper tackled the problem of inefficient gradient estimation in policy search algorithms by incorporating sensor data to reduce variance, resulting in faster learning as demonstrated both theoretically and empirically.
An efficient policy search algorithm should estimate the local gradient of the objective function, with respect to the policy parameters, from as few trials as possible. Whereas most policy search methods estimate this gradient by observing the rewards obtained during policy trials, we show, both theoretically and empirically, that taking into account the sensor data as well gives better gradient estimates and hence faster learning. The reason is that rewards obtained during policy execution vary from trial to trial due to noise in the environment; sensor data, which correlates with the noise, can be used to partially correct for this variation, resulting in an estimatorwith lower variance.