An article by Hafiq Anas, SDS Phd Student
The MSc research project was a thorough exploration into the field of Reinforcement Learning (RL), to solve a challenging problem in mobile robot navigation called Crowd Robot Navigation. This is a critical research area that can contribute to safe and socially compliant navigation for mobile robots sharing the same living space as humans. Mobile robots are often deployed in crowded public spaces such as airports and shopping areas to provide service and these places present a complex navigation environment for the robot due to the high uncertainty of different crowd-moving behaviors. We have explored various classical and deep RL algorithms and developed a novel approach that incorporates a sense of danger level of the moving crowd into the robot’s perception based on Collision Probability which we call Risk Perception. The results show that our method achieved a 100% success rate in all test settings. We compared our approach with a current state-of-the-art DRL-based approach, and our approach has performed significantly better, especially in terms of social safety. Importantly, our method can navigate in different crowd behaviors and requires no fine-tuning after being trained once. We further demonstrated the crowd navigation capability of our model in real-world tests.
Throughout the MSc journey, we believe we have made significant contributions to the advancement of knowledge in the AI domain for intelligent robotics. We are pleased to say that all of our work in implementing the method and system has been properly documented on our lab’s (Robolab, AI Lab) GitHub. For transparency, the source codes have also been made publicly available so anyone can attempt to replicate the approach and results. This MSc research has been a challenging yet fascinating work that originates from the Robolab, The School of Digital Science at Universiti Brunei Darussalam where the insights that we have uncovered have been shared for the world to see through our contributions at Robotics-related academic conferences. The first half of the work was accepted and presented at the 10th International Conference on Robot Intelligence Technology and Applications (RiTA2021) held in Daejeon, South Korea. The second half of the work was accepted to be presented as a workshop paper for the 2nd Workshop on Social Robot Navigation: Advances and Evaluation at the 36th International Conference on Intelligent Robots and Systems (IROS2023) held in Detroit, Michigan, USA.
Please check out Hafiq's work here:
An implementation of ROS Autonomous Navigation on Parallax Eddie platform, August 2021 (Paper link)
Comparison of Deep Q-Learning, Q-Learning and SARSA Reinforced Learning for Robot Local Navigation, April 2022, RiTA 2021 (Paper link)
Deep Reinforcement Learning-Based Mapless Crowd Navigation with Perceived Risk of the Moving Crowd for Mobile Robots, April 2023, IROS 2023 (Paper Link)
Hafiq’s other contributions can be found here (Researchgate) and here (Github)
Hafiq Anas is currently a first year PhD student at SDS. Hafiq continues to work under supervisors, Dr Ong Wee Hong and Dr Owais A. Malik
—
Universiti Brunei Darussalam
Jalan Tungku Link, BE1410
Negara Brunei Darussalam
office.sds@ubd.edu.bn