Tracking the Race Between Deep Reinforcement Learning and Imitation Learning -- Extended Version
This work provides insights into the trade-offs between DRL and IL for sequential decision-making problems, which is incremental as it benchmarks existing methods on a known domain.
The paper compared deep reinforcement learning (DRL) and imitation learning (IL) on the Racetrack benchmark, finding that DRL produces more foresighted agents that avoid risky states, while IL leads to riskier paths, with DRL performing best overall.
Learning-based approaches for solving large sequential decision making problems have become popular in recent years. The resulting agents perform differently and their characteristics depend on those of the underlying learning approach. Here, we consider a benchmark planning problem from the reinforcement learning domain, the Racetrack, to investigate the properties of agents derived from different deep (reinforcement) learning approaches. We compare the performance of deep supervised learning, in particular imitation learning, to reinforcement learning for the Racetrack model. We find that imitation learning yields agents that follow more risky paths. In contrast, the decisions of deep reinforcement learning are more foresighted, i.e., avoid states in which fatal decisions are more likely. Our evaluations show that for this sequential decision making problem, deep reinforcement learning performs best in many aspects even though for imitation learning optimal decisions are considered.