Papers
8
Total Citations
126
H-Index
7
About
Lifan Pan is an emerging researcher specializing in autonomous mobile robotics, with a particular focus on deep reinforcement learning (DRL) for robot navigation and multi-robot collision avoidance. His work addresses some of the most pressing challenges in robotics: enabling robots to navigate complex, dynamic environments safely and efficiently without relying on inter-robot communication. Pan's most significant contributions center on map-based deep reinforcement learning frameworks, where convolutional and recurrent neural networks are trained to derive intelligent navigation policies directly from egocentric environmental maps. His 2020 paper on distributed, non-communicating multi-robot collision avoidance — garnering 35 citations — tackled the particularly difficult problem of coordinating robots of varying shapes in communication-free settings, a scenario highly relevant to real-world deployments. Complementary work on single-robot navigation and 3D obstacle avoidance using depth cameras with limited fields of view further demonstrates the breadth of his contributions. Pan has also investigated sparse reward challenges in navigation training, applying techniques like Random Network Distillation to improve learning efficiency. With a growing citation record across multiple venues and a consistent research trajectory from 2020 to 2021, Pan represents a promising voice in the intersection of reinforcement learning and practical autonomous robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2Robot Navigation with Map-Based Deep Reinforcement Learning34 citations · 2020
- 3
- 4Multi-Robot Collision Avoidance with Map-based Deep Reinforcement Learning10 citations · 2020
- 5DRQN-based 3D Obstacle Avoidance with a Limited Field of View9 citations · 2021
- 6
- 7Robot Navigation with Map-Based Deep Reinforcement Learning8 citations · 2020
- 8DRQN-based 3D Obstacle Avoidance with a Limited Field of View2 citations · 2021