Cloud Hosting

(Credit: MIT News, with figures from iStock)

A new machine-learning based approach can control drones and autonomous vehicles more effectively and efficiently in difficult conditions, according to its developers at the Massachusetts Institute of Technology (MIT) and Stanford University.

The technique, designed for dynamic environments where conditions can change rapidly, could help an autonomous vehicle learn to compensate for slippery road conditions to avoid going into a skid. Other potential applications include allowing a robotic free-flyer to tow different objects in space, or enabling a drone to closely follow a downhill skier despite being buffeted by strong winds.

The researchers approach incorporates structures from control theory into the process for learning a model. It does this in such a way that leads to an effective method of controlling complex dynamics, an MIT announcement said, such as those caused by wind on the trajectory of a flying vehicle. The structures are like hints that can help guide how to control a system, the announcement added.

The focus of our work is to learn intrinsic structure in the dynamics of the system that can be leveraged to design more effective, stabilising controllers, said assistant professorNavid Azizan from MIT. By jointly learning the systems dynamics and these unique control-oriented structures from data, were able to naturally create controllers that function much more effectively in the real world.

The technique immediately extracts an effective controller from the model, the announcement said, as opposed to other machine-learning methods that require a controller to be derived or learned separately with additional steps. With this structure, the researchers approach is also able to learn an effective controller using less data than other approaches. This could help their learning-based control system achieve better performance, faster, in rapidly changing environments.

This work tries to strike a balance between identifying structure in your system and just learning a model from data, said lead authorSpencer M Richards, a graduate student at Stanford University. Our approach is inspired by how roboticists use physics to derive simpler models for robots. Physical analysis of these models often yields a useful structure for the purposes of control one that you might miss if you just tried to naively fit a model to data. Instead, we try to identify similarly useful structure from data that indicates how to implement your control logic.

The researchers found that their method was data-efficient, achieving high performance even with little data. It could reportedly model a highly dynamic rotor-driven vehicle using only 100 data points, for example. Methods that used multiple learned components saw their performance drop much faster with smaller datasets.

This efficiency could make the technique especially useful in situations where a drone or robot needs to learn quickly in rapidly changing conditions.

The general approach could also be applied to many types of dynamical systems, from robotic arms to free-flying spacecraft operating in low-gravity environments.

The work was supported, in part, by the NASA University Leadership Initiative and the Natural Sciences and Engineering Research Council of Canada. The research will be presented at the International Conference on Machine Learning (ICML), running this week at the Hawaii Convention Centre.

Want the best engineering stories delivered straight to your inbox? TheProfessional Engineeringnewslettergives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, clickhere.

Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.

Read the original post:
Drones stay on course in difficult conditions thanks to machine ... - Professional Engineering