Mahdi Abolfazli Esfahani

About

Mahdi Abolfazli Esfahani is a leading researcher at the intersection of deep learning, autonomous navigation, and inertial sensing. His work addresses fundamental challenges in enabling mobile robots and autonomous vehicles to perceive, localize, and navigate complex environments with minimal sensor reliance. Esfahani’s most significant contributions include the development of **AbolDeepIO** (113 citations), a pioneering deep inertial odometry network that overcomes the drift and noise inherent in IMU sensors, and **OriNet** (98 citations), which achieves robust 3D orientation estimation from a single IMU—a critical requirement for accurate trajectory tracking. He has also advanced autonomous steering and path planning through deep reinforcement learning, introducing novel architectures like **TDPP-Net** (59 citations) for 3D path planning and a DRL-based navigation policy (57 citations) that learns to steer safely from raw depth data. His work on **BND*-DDQN** (52 citations) further refines end-to-end learning for collision-free navigation. By combining deep neural networks with reinforcement learning and GPU-accelerated techniques (e.g., Parallel RANSAC, 23 citations), Esfahani has pushed the boundaries of how robots understand and move through their surroundings, making his research essential for the next generation of autonomous systems.

Research Focus

Key Achievements

Top Papers

1. 1
   AbolDeepIO: A Novel Deep Inertial Odometry Network for Autonomous Vehicles

   113 citations · 2019
2. 2
   OriNet: Robust 3-D Orientation Estimation With a Single Particular IMU

   98 citations · 2019
3. 3
   TDPP-Net: Achieving three-dimensional path planning via a deep neural network architecture

   59 citations · 2019
4. 4
   Learn to Navigate Autonomously Through Deep Reinforcement Learning

   57 citations · 2021
5. 5
   BND*-DDQN: Learn to Steer Autonomously Through Deep Reinforcement Learning

   52 citations · 2019
6. 6
   Learn to Steer through Deep Reinforcement Learning

   39 citations · 2018
7. 7
   iTD3-CLN: Learn to navigate in dynamic scene through Deep Reinforcement Learning

   25 citations · 2022
8. 8
   Parallel RANSAC: Speeding up plane extraction in RGBD image sequences using GPU

   23 citations · 2015
9. 9
   Depth-based Obstacle Avoidance through Deep Reinforcement Learning

   15 citations · 2019
10. 10
    From Local Understanding to Global Regression in Monocular Visual Odometry

    14 citations · 2019

Key Collaborators

Contact & Links