Legged Robot Locomotion
 |
Figure 1. Timmy, the legged robot. |
Biological systems are able to perform complex movements that adapt to environmental changes elegantly and with a higher energy-efficiency than traditionally-engineered mechanical systems. We are interested in studying the underlying control principles that produce these movements by experimenting with robotic systems that demonstrate biological similarity both in their observable output and in their underlying control system.
One robotic system we use for this research is Timmy (at right). This is a commercially available, six degree-of-freedom (DOF) biped robot that uses servo actuators. A drawback to the use of servos in robots is that they control joint angles by using as much force as possible to acquire and maintain the desired angle. An active topic of this research area is to improve the biological similarity of servo-controlled robots by moving the control paradigm away from overcontrolled, feedforward systems and toward systems with closed loop neural control and improved mechanical dynamics. The mechanical improvements are important because they allow some of the "computation" of the robot's movements to be performed by mechanical devices (as with muscles) rather than forcing the neural control architecture to handle all perturbations. By creating a more biological robot, we can more appropriately pair it with biologically inspired control systems.
 |
Figure 2. Modeled neural network. |
Another active research topic is the development and analysis of models of the neural networks used in the control of locomotion. The neuron models are implemented using a Field Programmable Gate Array (FPGA), which is essentially an array of millions of logic gates that can be easily reconfigured to produce custom, high-speed, parallel-computing circuits. The output of the modeled neural network is used to control the robot's motors while sensors on the robot provide feedback signals to the network. The goal of this research is to test the major theories of motor control and neural connectivity to determine their ability to produce stable locomotion in a real-world, mechanical system.