Wheelchair Robot Navigation in Outdoor Environments
We are developing a robotic system which can be automatically moved in outdoor environments. The system uses a wheelchair robot as the base platform. The robot is mainly navigated through images taken from the attached camera. Different sensors such as laser rangefinder, GPS, and compass are also attached to the robot to improve navigation performances. Deep learning algoorithms are implemented to the robot navigation system. Initial prototypes include the robot navigation using convolutional neural network (CNN) with robot odomery, and outdoor localization using Faster Regional CNN (Faster R-CNN).
Deep Learning Applications for Robot Manipulator
Industrial robot manipulator can be useful in many different ways. We add instances of deep learning algorithms to the robot manipulator for different tasks. One in particular is the use of deep belief neural network (DBNN) to detect the desire object within a pile of objects. The coordinates and orientations of the detected object are output and sent to the robot manipulator. Robot manipulator can move to the coordinates and rotate to the orientation received from DBNN. Currently, we are working on further implementtions of different algorithms, with several human cooperating tasks as additions.
Neural Network Based Robot Navigation in Indoor Environments
We are developing a navigation for mobile robots in indoor environments. The humanoid robot is chosen as the base for this research. We attached and tested different sensors such as laser rangefinder, camera, and Kinect camera. We applied depth camera and Kinect camera for indoor navigation. Images with depth information are used to train the neural network for obstacle avoidance. Images and depth information are used as the input while the robot movements are output from the neural network. Further works include different applications of deep neural networks and depth camera images.
3D Printed Prostheses
Low cost prosthesic limbs were designed and printed by a 3D printer. The aim is to produce a controllable prosthetic limbs at the low price cost. The prototype can be moved by electromyogram signals from arm muscles. We are working on deep learning implementations for controlling the 3D-printed prostheses.
Arm Rehabilitation System
We have developed a robot for human arm rehabilitation. The robot assists arm movements in different modes. The robot can be conveniently controlled by a smartphone application which connected to the robot via bluetooth connection.
Ankle Rehabilitation System
We have developed a robot for human ankle rehabilitation. The robot moves human ankle in different patterns. The robot can be controlled manually by human or move automatically by the programmed patterns. Now we are working on learning by immitation, where the robot learns the best training technics.
Brain Machine Interface
The mobile robot ability to navigate autonomously in its environment is very important. Even though the advances in technology, robot self-localization and goal directed navigation in complex environments are still challenging tasks. We propose a novel method for robot navigation based on rat`s brain signals (Local Field Potentials). We developed an algorithm by which the robot learned to imitate the rat`s decision-making by mapping the rat`s brain signals into its own actions. Finally, the robot learned to integrate the internal states as well as external sensors in order to localize and navigate in the complex environment.