Curriculum Vitae

Yuhong Huang is currently a doctoral student at the Chair of Robotics, Artificial Intelligence and Real-time Systems, Department of Informatics, Technical University of Munich. He received his M.Eng degree in Integrated Circuit Engineering at the National University of Defense Technology, China, in 2020, and his B.Eng degree in Software Engineering at Sun Yat-sen University, China, in 2017.

His research interests include autonomous control of mobile robots, and machine learning (especially learning in small simple).

Thesis Topic

Topic 1: Spine-based Adaptive Locomotion of the Rat Robot

The rat robot is a bionic quadruped robot with soft actuated spine.  The soft spine is able to enhance the quadrupedal locomotion of the rat robot, including balance, velocity, and direction.
In our following research, we will focus on generating spine-based pathological gait for the rat robot with CPG-rewiring.

• Topic 1.1: Pathological gait generation for rat robot with spine-based damage control
• Topic 1.2: Rewiring the CPG controller during rat robot error behaviors
• Topic 1.3: Spine-based damage control for the rat robot with CPG-rewiring.

Topic 2: Reinforcement learning for the small-size quadruped robot

Reinforcement learning approaches, which learn from the interaction between robots and their environments, can autonomously generate adaptable quadrupedal locomotion based on real-time feedback. However, these approaches often rely on the variability of data obtained from the sensors of full-size quadruped robots and the execution of suitable behaviors through powerful actuators. In contrast, small-sized quadruped robots face volumetric constraints that limit the configuration of sensors and actuators, making it challenging for them to understand and adapt to environmental changes.

• Topic 2.1: Extract effective environmental differences based on little perception information
• Topic 2.2: Robot gait generation via Reinforcement Learning
• Topic 2.3: Adaptive control of robots in different environments through hierarchical reinforcement learning

Topic 3: Swarm Intelligence for Group Navigation

In real life, humans always work in an environment with other people (avoid interference from others or collaborate with others). This is a great challenge to robot decisions. For example, in obstacle avoidance of dense crowd navigation, the robot is required to predict the uncertain movement path of other people and avoid them. At the same, the robot needs to make the optimal decision to arrive at the target quickly.

• Topic 3.1: Dense-Crowd Navigation via Reinforcement Learning
• Topic 3.2: Control Method Based on Neighbor Reward for Robot Swarm

If you are interested in one of the topics, please feel free to contact me.