abmorobotics/maps

Hello 👋

• 👨‍💼 I am a robotics master student from Aalborg University in Denmark, currently working on my master thesis in learning-based mapless navigation using reinforcement learning.
• 📖 I'm a passionate about artificial intelligence, reinforcement learning, and scaling these systems to generalize to real-world robotic applications.

📫 Contact with me:

📚 Main Programming Languages:

Currently Working on:

Currently, I am working on migrating a reinforcement learning environment I've developed previously to ORBIT, and improving the scalability and generalizability of the system. The goal is to significantly enhance real-world performance and bridge the gap between simulation and reality (Sim2Real). To achieve these objectives, I am employing a range of state-of-the-art methods, including:

• Learning By Cheating.
• Offline Reinforcement Learning
• Combining Real World Datasets with Synthetic Datasets

Learning by Cheating is used to train the agent using a simple and computationly fast observation space (heightmap), and transfer the learned policy to a model that uses RGB-D information instead of a heightmap. The video below shows the environment i am migrating to ORBIT from Isaac Gym.

Rover.1.mp4

maps's Issues

Functions remake create_publicher_array

These two function need to be remade, so they are capable of take in multiple names assigned for each robot ID for better control. Be aware that it should also work for the shuttles

MAPs/ros2_ws/src/orchestrator/orchestrator/manipulator.py

Lines 49 to 55 in 4dcafb1

	self.publisher_array.append(create_publisher_array(self,
	len(self.name_of_manipulators),
	topic_prefix=name,
	topic_name='/joint_command',
	msg_type=JointState))
	self.msg.append(create_joint_state_message_array(joint_names,