Untitled

Abstract

Abstract. Recent algorithmic and theoretical advances in reinforcement learning (RL) have attracted widespread interest. RL algorithms have appeared that approximate dynamic programming on an incremental basis. They can be trained on the basis of real or simulated experiences, focusing their computation on areas of state space that are actually visited during control, making them computationally tractable on very large problems. If each member of a team of agents employs one of these algorithms, a new collective learning algorithm emerges for the team as a whole. In this paper we demonstrate that such collective RL algorithms can be powerful heuristic methods for addressing large-scale control problems.Elevator group control serves as our testbed. It is a difficult domain posing a combination of challenges not seen in most multi-agent learning research to date. We use a team of RL agents, each of which is responsible for controlling one elevator car. The team receives a global reward signal which appears noisy to each agent due to the effects of the actions of the other agents, the random nature of the arrivals and the incomplete observation of the state. In spite of these complications, we show results that in simulation surpass the best of the heuristic elevator control algorithms of which we are aware. These results demonstrate the power of multi-agent RL on a very large scale stochastic dynamic optimization problem of practical utility.

Adaptive Discrete Communication Bottlenecks with Dynamic Vector Quantization

“…Bottlenecking inter-agent communication in multiagent RL. A wide range of multi-agent applications have benefitted from inter-agent message passing including distributed smart grid control (Pipattanasomporn et al, 2009), consensus in networks (You & Xie, 2011), multi-robot control (Ren & Sorensen, 2008), autonomous vehicle driving (Petrillo et al, 2018), elevators control (Crites & Barto, 1998) and for language learning in two-agent systems (Lazaridou et al, 2017). An important challenge in MARL is how to facilitate communication among interacting agents, especially in tasks requiring synchronization (Scardovi & Sepulchre, 2009;Wen et al, 2012).…”

Understanding the Importance of Efficient Visitor Flow Within Tokyo Skytree

“…For example, [6] presents a methodology to improve elevator performance by applying machine learning. Crites et al [7] present a similar approach, focusing on safety and control issues. Similarly, Ho and Fu [8] uses a learning algorithm to optimize elevator performance.…”

Application of Reinforcement Learning in Decision Systems: Lift Control Case Study

“…The real-world scenario introduces uncertainties that necessitate adaptive decision-making without complete knowledge of future events. Heuristic algorithms have traditionally tackled this issue [3], with recent exploration of solutions based on artificial intelligence algorithms [4][5][6][7].…”