Perception, learning, planning, control, and hardware experiments for field, industrial, and assistive robotics.
This work covers object grounding, 3D reconstruction, grasp selection, collision-aware planning, locomotion, and real-time control across manipulators, legged robots, and wheeled platforms. Experiments address clutter, occlusion, incomplete depth, calibration error, and changing environments.
Navigation models account for morphology and physical capability when estimating traversability and selecting motions.
The operator provides motion, target, or task commands through vision, gaze, muscle signals, and natural language. Robot assistance covers collision avoidance, local planning, grasp execution, and final alignment.
EMG and lightmyography support gesture, dexterous-motion, and force decoding. RGB-D tracking, mixed reality, and open-vocabulary perception support camera-based and assistive interfaces.
Hardware implementations, experimental results, and system components.
The operator selects a target with a free-form prompt and controls Spot through arm motion and hand gestures. The robot tracks the target, avoids collisions, and assists the final approach.
The operator names an object in the scene. The system segments it, completes the partial RGB-D geometry, selects a collision-free 6-DoF grasp, and executes it on Spot.
An RGB-D camera maps the operator's wrist motion to Spot's arm. Trajectory planning prevents collisions, and hand gestures trigger grasping.
EMG and lightmyography armbands decode gestures, in-hand motion, and grasping force for robot and prosthesis control.
The traversability model uses the robot's physical capabilities as an input. The same scene produces different maps for legged and wheeled platforms.
I work in the group led by Assoc. Prof. Marcelo Becker at the USP Center for Robotics.