Policy learning in SE(3) action spaces
This work addresses a bottleneck in robotic manipulation by enabling higher-dimensional spatial action spaces, though it appears incremental as it builds on existing spatial action representation methods.
The paper tackles the limitation of spatial action representations in robotics, which are often confined to low-dimensional action spaces and short-horizon tasks, by proposing ASRSE3 to reduce action space dimensionality and SDQfD for large action spaces, showing that both methods outperform standard baselines in block construction tasks and are applicable to real robotics systems.
In the spatial action representation, the action space spans the space of target poses for robot motion commands, i.e. SE(2) or SE(3). This approach has been used to solve challenging robotic manipulation problems and shows promise. However, the method is often limited to a three dimensional action space and short horizon tasks. This paper proposes ASRSE3, a new method for handling higher dimensional spatial action spaces that transforms an original MDP with high dimensional action space into a new MDP with reduced action space and augmented state space. We also propose SDQfD, a variation of DQfD designed for large action spaces. ASRSE3 and SDQfD are evaluated in the context of a set of challenging block construction tasks. We show that both methods outperform standard baselines and can be used in practice on real robotics systems.