Future Ideas

SocRob Sponsors and Social Media

• Insta, here we go ;) (ask Barbara's help on this);

Hardware

• Install 3D LIDAR;
• Gimbal Arm;
• Kinova Arm;
• New compliant gripper that can wrap around objects (soft robotics);
• Bigger gripper to facilitate grabbing bigger objects (ex: cereal box);

Software/Engineering

• Optimize deep learning models (e.g. TensorRT)
• Implement CI/CD practice
• Migrate to ROS2;

Perception Pipeline

• Dataset

  • Buy YCB objects or ask Vislab to use theirs;
  • Add YCB objects to simulation + Train YCB objects in simulation. Eventually try to make the simulation photorealistic (Uni Bremen);
  • Do testbed dataset (RoboCup oriented);

• REID

  • 3D LIDAR reID;
  • Improve the current face recognition algorithm (DeepFace currently being tested);
  • Test Vicente's REID approach (pointcloud);

• Object Detection

  • 6 DOF pose estimation;

Navigation

• 3D LIDAR SLAM + 3D Localization + 3D Navigation;
• People tracker (Trajectory People Follow MPC + local planner to avoid obstacles) (Vicente fez reid usando a pointcloud da pessoa);

Manipulation

• Implement force feedback on the gripper;
• Pregrasp and Grasp pose dictionary for each object;

Speech

• Add voice commands running by default for safety: stop to stop robot, home to send the robot to home motion, ...;
• Chatbot built using chatgpt so the robot can interact with people in a non scripted conversation (the robot would need to know facts about himself and the environment. This is the thesis of Veronica Spelbrink but she is using a grammar based chatbot);

Semantic Map

• Improve semantic map by adjusting drawn surfaces when they are perceived;

Interfaces

• Explainable robot actions: GUI + sound that clarifies what the robot is doing;