Generate near-infinite permutations of complex, domain-specific, 3D environments
highlights
- Reinforcement learning to enable control for complex, non-linear multi-input multi-output control systems
- Squeeze out maximal performance by optimizing control settings with a reinforcement learning agent
- Optimizing control without the availability of an accurate model
revolutionizing control systems with reinforcement learning.
Traditionally, a lot of effort is put into decoupling inputs and linearizing these nonlinear MIMO systems to make a well-functioning control system. However, reinforcement learning offers a different approach where the algorithm learns the optimal behavior by itself. The challenge is to get high performance in terms of stability and or reaction speed in these non-linear MIMO control systems.
maximize your machine's performance.
We can use reinforcement learning (RL) techniques to optimize your machine’s performance in two ways. First, use RL to model the entire nonlinear MIMO control system when the system is too complex and non-linearizable. Secondly, use RL to tune settings in the traditional control system to squeeze out the last bit of performance.