Imitation Learning via Simultaneous Optimization of Policies and Auxiliary Trajectories
This addresses the challenge of data-efficient policy learning in robotics, particularly for tasks like lifting and obstacle avoidance, though it is incremental as it builds on existing imitation learning techniques.
The paper tackles the problem of imitation learning without access to oracular experts by introducing CoDE, a method that uses only fixed trajectory demonstrations, resulting in better generalization and more accurate behavior reproduction with fewer trajectories compared to standard behavioral cloning.
Imitation learning (IL) is a frequently used approach for data-efficient policy learning. Many IL methods, such as Dataset Aggregation (DAgger), combat challenges like distributional shift by interacting with oracular experts. Unfortunately, assuming access to oracular experts is often unrealistic in practice; data used in IL frequently comes from offline processes such as lead-through or teleoperation. In this paper, we present a novel imitation learning technique called Collocation for Demonstration Encoding (CoDE) that operates on only a fixed set of trajectory demonstrations. We circumvent challenges with methods like back-propagation-through-time by introducing an auxiliary trajectory network, which takes inspiration from collocation techniques in optimal control. Our method generalizes well and more accurately reproduces the demonstrated behavior with fewer guiding trajectories when compared to standard behavioral cloning methods. We present simulation results on a 7-degree-of-freedom (DoF) robotic manipulator that learns to exhibit lifting, target-reaching, and obstacle avoidance behaviors.