About

Fengda Zhu is a researcher specializing in embodied artificial intelligence, with a particular focus on vision-and-language navigation, 3D indoor navigation, and robotic perception. His work sits at the intersection of deep learning, reinforcement learning, and autonomous agent behavior, addressing some of the most challenging problems in intelligent robotics. Zhu's most impactful contribution is "SOON: Scenario Oriented Object Navigation with Graph-based Exploration" (2021), which has garnered 115 citations and tackles the ambitious goal of enabling robots to navigate freely within 3D environments guided by natural language — moving beyond the limitations of fixed starting-point benchmarks. His earlier work on sim-to-real reinforcement transfer (2019) addressed the critical challenge of bridging the gap between simulated training environments and real-world deployment, reducing the costly burden of physical data collection. Beyond navigation benchmarks, Zhu has advanced dialogue-driven navigation through self-motivated communication agents and contributed a comprehensive survey on deep learning for embodied visual navigation, underscoring his role as both a practitioner and a synthesizer of the field. His more recent work on robotic manipulation with world models signals a broadening research agenda. Collectively, his publications reflect a coherent vision: building robots that understand, communicate, and act intelligently in complex, human-centered environments.

Research Focus

Key Achievements

5
H-Index
7
Papers
192
Total Citations
27
Avg Citations/Paper
🏆 Most Cited Paper
SOON: Scenario Oriented Object Navigation with Graph-based Exploration
115 citations · 2021
📈 Most Prolific Year: 2021 (4 Papers)
🤝 Key Collaborators: 17
🏛 Institutions: Australian Regenerative Medicine Institute, Southern University of Science and Technology, Monash University

Top Papers

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 14 days ago