Projects

Deep Visual Odometry for Robot Navigation

Visual OdometryDeep LearningROSNavigation

We implemented DeepVO, a deep visual odometry model (paper 1, paper 2), to estimate a robot’s motion from a sequence of camera images, and integrated it with the ROS navigation stack.

Code on GitHub

To test its effectiveness we added Gaussian noise to the wheel odometry and compared navigation performance with and without DeepVO in the localization loop. In the end DeepVO was added as a measurement model to the EKF localization algorithm.

Robot hardware configuration.
Robot hardware configuration.
Navigation software architecture.
Navigation software architecture.
Experiment environment, scene 1.
Experiment environment, scene 1.
Experiment environment, scene 2.
Experiment environment, scene 2.

Result

The red dot is the target destination.

Navigation without DeepVO.
Navigation with DeepVO.