This Analog Controller Allows Robots to Move Faster and More Efficiently

Researchers have demonstrated a self-balancing robot that utilizes memristors to form an extremely effective analog control system.

The analog controller uses memristors to enable an inverted pendulum robot to stay upright more efficiently than a traditional system. (📷: Wei Wu Research Group/University of Southern California)

Robots functioning in the real world are restricted by the amount of computing power they have. Computers are running faster and more efficiently, but they're struggling to keep up with robotic systems, which have access to better sensors and more data. A memristor could help robotics overcome this barrier through lower complexity, higher speed, and lower cost. Researchers from the University of Southern California and the Air Force Research Laboratory have demonstrated a self-balancing robot that utilizes memristors to form an extremely effective analog control system. The team's idea was inspired by the functional structure of the human brain.

The researchers integrated a memristor into an analog circuit running an algorithm that inputs data from the robot's accelerometer and gyroscope. By doing so, they developed a wholly analog and physical Kalman filter, which takes out the sensor signal's noise. Another memristor was used to convert the sensor data into a proportional-derivative (PD) controller. The team then combined the components to create an analogy system capable of doing most of the work needed to keep an inverted pendulum robot upright more efficiently than a conventional system.

Traditionally-controlled robots tend to be shakier due to the linearity of the dynamic system, which alternates quicker than the on-board controller can handle. Due to the memristors, the cycle time is considerably shortened from 3,034 microseconds to six microseconds, allowing the robot to balance more smoothly.

The robot is a hybrid system since the motor drivers and a digital computer communicates back and forth.

Offloading a large amount of computation onto the memristors enables the robot's higher brain functions to have more breathing room. In addition to performance-improvement, there is a reduction in cost, power, and space. This was recently achieved due to memristor advances and availability. The researchers believe that memristor-based hybrid computing could improve mobile robots' robustness and performance with higher degrees of freedom.

Related articles
Sponsored articles
Related articles