Title: Online Dynamic Trajectory Optimization and Control for a Quadruped Robot Abstract: Legged robot locomotion requires the planning of stable reference trajectories, especially while traversing uneven terrain. The proposed trajectory optimization framework is capable of generating dynamically stable base and footstep trajectories for multiple steps. The locomotion task can be defined with contact locations, base motion or both, making the algorithm suitable for multiple scenarios (e.g., presence of moving obstacles). The planner uses a simplified momentum-based task space model for the robot dynamics, allowing computation times that are fast enough for online replanning. This fast planning capability also enables the quadruped to accommodate for drift and environmental changes.
Hanz Cuevas Velasquez
Title: Joint semantic segmentation and disparity learning
Abstract: Disparity estimation and semantic segmentation are two fundamental problems in computer vision. The goal of disparity estimation is to obtain the corresponding pixels from a rectified pair of images. On the other hand, semantic segmentation is used to assign class labels to each pixel in the image. These two methods are highly used in scene understanding, autonomous driving and robotics. However, executing the state-of-the-art methods of both areas together involves a long computation run-time, which is prohibitively expensive in these systems. Therefore, a method that combines both tasks into a single model reduces computation and allows it to be embedded in portable devices or executed in real-time. We approach this problem through multi-task learning, our method focuses on sharing the specific feature knowledge between tasks in a progressive manner, which improves learning efficiency and prediction accuracy for each task.