End-to-End Deep Imitation Learning: Robot Soccer Case Study

In imitation learning, behavior learning is generally done using the features extracted from the demonstration data. Recent deep learning algorithms enable the development of machine learning methods that can get high dimensional data as an input. In this work, we use imitation learning to teach the robot to dribble the ball to the goal. We use B-Human robot software to collect demonstration data and a deep convolutional network to represent the policies. We use top and bottom camera images of the robot as input and speed commands as outputs. The CNN policy learns the mapping between the series of images and speed commands. In 3D realistic robotics simulator experiments, we show that the robot is able to learn to search the ball and dribble the ball, but it struggles to align to the goal. The best-proposed policy model learns to score 4 goals out of 20 test episodes.

Effective Multi-Robot Spatial Task Allocation using Model Approximations

Real-world multi-agent planning problems cannot be solved using decision-theoretic planning methods due to the exponential complexity. We approximate firefighting in rescue simulation as a spatially distributed task and model with multi-agent Markov decision process. We use recent approximation methods for spatial task problems to reduce the model complexity. Our approximations are single-agent, static task, shortest path pruning, dynamic planning horizon, and task clustering. We create scenarios from RoboCup Rescue Simulation maps and evaluate our methods on these graph worlds. The results show that our approach is faster and better than comparable methods and has negligible performance loss compared to the optimal policy. We also show that our method has a similar performance as DCOP methods on example RCRS scenarios.