Creating and verifying stable AI-controlled systems in a rigorous and versatile way

Neural networks have made a seismic impact on how engineers design controllers for robots, catalyzing more adaptive and efficient machines. Still, these brain-like machine-learning systems are a double-edged sword: Their complexity makes them powerful, but it surely also makes it difficult to ensure that a robot powered by a neural network will safely accomplish its task.

The normal option to confirm safety and stability is thru techniques called Lyapunov functions. For those who can discover a Lyapunov function whose value consistently decreases, then you definately can know that unsafe or unstable situations related to higher values won’t ever occur. For robots controlled by neural networks, though, prior approaches for verifying Lyapunov conditions didn’t scale well to complex machines.

Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and elsewhere have now developed recent techniques that rigorously certify Lyapunov calculations in additional elaborate systems. Their algorithm efficiently searches for and verifies a Lyapunov function, providing a stability guarantee for the system. This approach could potentially enable safer deployment of robots and autonomous vehicles, including aircraft and spacecraft.

To outperform previous algorithms, the researchers found a frugal shortcut to the training and verification process. They generated cheaper counterexamples — for instance, adversarial data from sensors that would’ve thrown off the controller — after which optimized the robotic system to account for them. Understanding these edge cases helped machines learn the best way to handle difficult circumstances, which enabled them to operate safely in a wider range of conditions than previously possible. Then, they developed a novel verification formulation that permits using a scalable neural network verifier, α,β-CROWN, to supply rigorous worst-case scenario guarantees beyond the counterexamples.

“We’ve seen some impressive empirical performances in AI-controlled machines like humanoids and robotic dogs, but these AI controllers lack the formal guarantees which might be crucial for safety-critical systems,” says Lujie Yang, MIT electrical engineering and computer science (EECS) PhD student and CSAIL affiliate who’s a co-lead creator of a brand new paper on the project alongside Toyota Research Institute researcher Hongkai Dai SM ’12, PhD ’16. “Our work bridges the gap between that level of performance from neural network controllers and the security guarantees needed to deploy more complex neural network controllers in the true world,” notes Yang.

For a digital demonstration, the team simulated how a quadrotor drone with lidar sensors would stabilize in a two-dimensional environment. Their algorithm successfully guided the drone to a stable hover position, using only the limited environmental information provided by the lidar sensors. In two other experiments, their approach enabled the stable operation of two simulated robotic systems over a wider range of conditions: an inverted pendulum and a path-tracking vehicle. These experiments, though modest, are relatively more complex than what the neural network verification community could have done before, especially because they included sensor models.

“Unlike common machine learning problems, the rigorous use of neural networks as Lyapunov functions requires solving hard global optimization problems, and thus scalability is the important thing bottleneck,” says Sicun Gao, associate professor of computer science and engineering on the University of California at San Diego, who wasn’t involved on this work. “The present work makes a very important contribution by developing algorithmic approaches which might be a lot better tailored to the actual use of neural networks as Lyapunov functions on top of things problems. It achieves impressive improvement in scalability and the standard of solutions over existing approaches. The work opens up exciting directions for further development of optimization algorithms for neural Lyapunov methods and the rigorous use of deep learning on top of things and robotics basically.”

Yang and her colleagues’ stability approach has potential wide-ranging applications where guaranteeing safety is crucial. It could help ensure a smoother ride for autonomous vehicles, like aircraft and spacecraft. Likewise, a drone delivering items or mapping out different terrains may benefit from such safety guarantees.

The techniques developed listed here are very general and aren’t just specific to robotics; the identical techniques could potentially assist with other applications, comparable to biomedicine and industrial processing, in the longer term.

While the technique is an upgrade from prior works when it comes to scalability, the researchers are exploring how it could actually perform higher in systems with higher dimensions. They’d also wish to account for data beyond lidar readings, like images and point clouds.

As a future research direction, the team would love to supply the identical stability guarantees for systems which might be in uncertain environments and subject to disturbances. As an illustration, if a drone faces a powerful gust of wind, Yang and her colleagues need to ensure it’ll still fly steadily and complete the specified task. 

Also, they intend to use their method to optimization problems, where the goal could be to attenuate the time and distance a robot needs to finish a task while remaining regular. They plan to increase their technique to humanoids and other real-world machines, where a robot needs to remain stable while making contact with its surroundings.

Russ Tedrake, the Toyota Professor of EECS, Aeronautics and Astronautics, and Mechanical Engineering at MIT, vice chairman of robotics research at TRI, and CSAIL member, is a senior creator of this research. The paper also credits University of California at Los Angeles PhD student Zhouxing Shi and associate professor Cho-Jui Hsieh, in addition to University of Illinois Urbana-Champaign assistant professor Huan Zhang. Their work was supported, partially, by Amazon, the National Science Foundation, the Office of Naval Research, and the AI2050 program at Schmidt Sciences. The researchers’ paper will likely be presented on the 2024 International Conference on Machine Learning.