Individual agents, such as robots or drones, can cooperate and finish a task when communication channels are open. What happens, though, if their technology is insufficient or the signals are jammed, making communication impossible? Researchers from the University of Illinois at Urbana-Champaign began with this more challenging task. They came up with a way to teach many agents to work together by using multi-agent reinforcement machine learning, a type of artificial intelligence.
Huy Tran, an aeronautical engineer in Illinois, noted that it is simpler when agents can communicate with one another. But we wanted to achieve this in a decentralized manner, so that they don’t communicate with one another. “We also concentrated on circumstances in which it is unclear what the various duties or responsibilities of the agents should be.”
Because it’s unclear what one agent should do in contrast to another agent, Tran claimed that this scenario is far more complicated and has a harder difficulty.
The intriguing topic, according to Tran, is how humans can gradually learn to work together to achieve a goal.
By developing a utility function that alerts the agent when it is acting in a way that is beneficial to the team or useful to the individual, Tran and his colleagues employed machine learning to find a solution to this issue.
With team goals, “it’s difficult to determine who helped us win,” he remarked. “We created a machine learning method that enables us to recognize when a single agent contributes to the overall team goal. If you compare it to sports, one soccer player may score, but we also want to know about the teammate’s contributions, such as assists. “Understanding these delayed impacts is challenging.”
The algorithms that the researchers used can also spot when an agent or robot is acting in a way that isn’t helpful to the end result. The robot simply chose to do something that wasn’t helpful to the end result, not necessarily anything that was bad.
They tested their algorithms by simulating games like Capture the Flag and StarCraft, a popular computer game.
Watch Huy Tran show how deep reinforcement learning can help robots figure out what their next move should be in the game Capture the Flag.
“StarCraft can be a little bit more unpredictable — we were excited to see our method work well in this environment too.”
Tran says that this kind of algorithm is useful in a wide range of real-world situations, such as military surveillance, robots working together in a warehouse, managing traffic signals, coordinating deliveries by self-driving cars, and controlling the grid.
According to Tran, when Seung Hyun Kim was a mechanical engineering undergraduate student, he developed the majority of the theory underlying the concept; Neale Van Stralen, an aerospace undergraduate, assisted with the implementation. Both students received guidance from Tran and Girish Chowdhary. The work was recently presented to the AI community at the peer-reviewed conference on autonomous agents and multi-agent systems.
