Imitating Cost Constrained Behaviors in Reinforcement Learning
Speaker (s): 
SHAO Qian
PhD Candidate
School of Computing and Information Systems
Singapore Management University
| Date:
Time:
Venue:
| | 9 May 2024, Thursday
11:00am - 11:30am
Meeting room 4.4, Level 4
School of Computing and Information Systems 1,
Singapore Management University,
80 Stamford Road, Singapore 178902
Please register by 8 May 2024.
|
|
About the Talk
Imitation learning traditionally focuses on mimicking expert behavior without considering constraints like fuel or time limits, which are crucial in real-world scenarios such as self-driving delivery vehicles. This paper introduces methods for adapting imitation learning to obey such constraints, employing techniques like Lagrangian-based optimization, meta-gradients, and alternating gradient methods for cost violations. Our empirical results demonstrate that our meta-gradient approach significantly outperforms conventional imitation learning methods in scenarios with strict cost constraints.
This is a Pre-Conference talk for 34th International Conference on Automated Planning and Scheduling (ICAPS 2024).
About the Speaker
SHAO Qian is a Ph.D Candidate in Computer Science at the SMU SCIS, supervised by Prof. CHENG Shih-Fen. Her research interests are last-mile logistics, imitation learning and approximate dynamic programming.
